Commission, June 27, 2017, No 39740

THE EUROPEAN COMMISSION,

Having regard to the Treaty on the Functioning of the European Union, Having regard to the Agreement on the European Economic Area,

Having regard to Council Regulation (EC) No 1/2003, of 16 December 2002 on the implementation of the rules on competition laid down in Articles 81 and 82 of the Treaty1,  and in particular Articles 7(1), 23(2) and 24(1) thereof,

Having regard to the Commission Decisions of 30 November 2010 and of 14 July 2016 to initiate proceedings in this case,

Having given the undertaking concerned the opportunity to make known its views on the objections raised by the Commission pursuant to Article 27(1) of Regulation (EC) No 1/2003 and Article 12 of Commission Regulation (EC) No 773/2004 of 7 April 2004 relating to the conduct of proceedings by the Commission pursuant to Articles 81 and 82 of the Treaty2,

After consulting the Advisory Committee on Restrictive Practices and Dominant Positions, Having regard to the final report of the Hearing Officer in this case,

Whereas:

1. INTRODUCTION

(1)     This Decision is addressed to Google Inc. (“Google”) and to Alphabet Inc. (“Alphabet”).

(2)     The Decision establishes that the more favourable positioning and display  by Google, in its general search results pages, of its own comparison shopping service compared to competing comparison shopping services (the “Conduct”) infringes Article 102 of the Treaty and Article 54 of the Agreement on the European Economic Area (“EEA Agreement”).3

(3)     This Decision is structured as follows. Section 2 provides an overview of Google’s business activities. Section 3 summarises the procedure relating to the proceedings in this case to date. Section 4 addresses Google's allegations that the Commission's investigation has suffered from procedural irregularities. Sections 5 to 11 set out the Commission’s conclusions regarding the relevant product and geographic markets, Google's dominant position, Google's abuse of that dominant position, the Commission's jurisdiction, the effect of the abuse on trade between Member States and between Contracting Parties to the EEA Agreement, the duration of the abuse  and the addressees of this Decision. Sections 12 to 14 conclude by describing the remedies and periodic penalty payments necessary to bring the infringement to an end and explaining the amount of the fine.

2. GOOGLE'S BUSINESS ACTIVITIES

2.1. The undertaking

(4)     Google is a multinational technology company based in the United States of America (“the US”), specialising in internet-related services and products that include online advertising technologies, search, cloud computing, software and hardware. It offers various services in the territories of all the Contracting Parties to the EEA Agreement.

(5)     In August 2015, Google announced its intention to create a new holding company, Alphabet. The reorganisation was completed on 2 October 2015. Since that date, Google has been a wholly-owned subsidiary of Alphabet, which has continued to be the umbrella company for the internet interests of Alphabet.

(6)     According to the consolidated financial statements of Alphabet, its turnover was USD 90 272 million (approximately EUR 81 597 million4) for the year running from 1 January 2016 to 31 December 2016.5

2.2. Overview of Google's business activities

(7)     Google’s business model is based on the interaction between the online products and services it offers free of charge and its online advertising services from which it generates the main source of its revenue.6

2.2.1. Google Search

(8)     Google’s flagship online service is its general search7 engine, Google Search, which is accessible either through Google’s main website in the US (www.google.com), or through localised websites. Google also powers the search functions of certain third party websites.

(9)     Google Search allows users to search for information across the internet. Google Search exists for static devices (personal computers and laptops) and for mobile devices (smartphones and tablets). While the user interface may vary depending on the type of device, the underlying technology is essentially the same.

(10)    When a user enters a keyword or a string of keywords (a “query”) in Google Search, Google’s general search results pages return different categories of search results, including generic search results8 (as described in section 2.2.2) and  specialised search results9 (as described in section 2.2.4). In addition, Google Search may return a third category of results, namely online search advertisements (as described in section 2.2.3).

(11)    When a user enters a query, Google’s programmes essentially run two sets of algorithms: generic search algorithms and specialised search algorithms.10

(12)    Google’s generic search algorithms are designed to rank pages containing any possible content. Google applies these algorithms to all types of pages, including the web pages of competing specialised search services. By contrast, specialised search algorithms “are specifically optimized for identifying relevant results for a particular type of information”,11 such as news, local businesses or product information.

(13)    The results of these two sets of algorithms – the generic search results and the specialised search results – appear together on Google’s general search results pages.

2.2.2. Generic search results

(14)    Generic search results typically appear on the left side of Google’s general search results pages in the form of blue links with short excerpts (“snippets”) in order of their “web rank”. Generic search results can link to any page on the internet, including web pages of specialised search services competing with Google's specialised search services.

(15)    The delivery of generic search results involves three automated processes: crawling, indexing and serving.12 Crawling is the process by which Google discovers new and updated web pages. Indexing is the process by which web pages and their content are catalogued and added to the Google index. Serving is the process by which, when a user enters a query, Google’s programmes check the index for web pages that match the query, determine their relevance to the query and “serve” the results to the user.

(16)    To rank generic search results in response to a query, Google uses algorithms. It relies in particular on an algorithm called PageRank, which is “a method for rating Web pages objectively and mechanically, effectively measuring the human interest and attention devoted to them”.13 PageRank essentially measures the importance of a web page based on the number and quality of links to that page, the underlying assumption being that more important websites are likely to receive more links  from

other websites. Google applies a variety of adjustment mechanisms to the results of PageRank to improve the relevance of the generic search results on its general search results pages. PageRank and the adjustment mechanisms together determine the rank of a web page in generic search results on Google's general search results pages.

(17)    Google does not charge websites ranked in generic search results on its general search results pages and does not accept any payment that would allow websites to rank higher in these results.

2.2.3. Online search advertising results

(18)    In response to a user’s query on Google Search, Google’s general search results pages may also return search advertisements drawn from Google’s auction-based online search advertising platform, AdWords (“AdWords results”).

(19)    AdWords results are not limited to specific categories of products, services or information. They typically appear on general search results pages above or below generic search results with a label informing users of their nature as advertisements (for example, “Ads”).14 AdWords results can be purchased by any advertiser and are not limited to particular categories of advertisers.

(20)    The appearance of AdWords results in response to a user’s query involves two main elements. First, AdWords identifies a pool of relevant search advertisements by matching the keywords on which advertisers have associated their search advertisements with the keywords used in the query. Second, AdWords ranks the relevant search advertisements within the pool based on their “Ad Rank”. The  ranking of a search advertisement depends on two factors: the maximum price an advertiser has indicated it is willing to pay for each click on its search advertisement in a second-price auction15 and the quality rating of that search advertisement  (known as “Quality Score”). The Quality Score is based among other things on a search advertisement’s predicted click-through rate.16 AdWords results that appear the most visibly on Google’s general search results pages are those with the highest Ad Rank scores.17

(21)    When a user clicks on an AdWords result, Google receives remuneration for that click from the advertiser owning the website to which the user is directed (known as the “pay per click” system).

(22)    AdWords results allow advertisers to lead interested users entering queries on Google Search to their websites, including in circumstances where these websites would otherwise not rank highly in generic search results on Google's general search results pages. Specialised search services competing with services provided by Google often also purchase AdWords results.

2.2.4. Specialised search results

(23)    In response to a user query, Google’s general search results pages may also return specialised search results from Google’s specialised search services. In most instances, specialised search results are displayed with attractive graphical features, such as large scale pictures and dynamic information. Specialised search results in a particular category are positioned within sets referred to by Google as “Universals”  or “OneBoxes”.18 They are in most instances positioned above generic search results, or among the first of them.

(24)    Google operates several search services that can be described as “specialised” because they group together results for a specific category of products, services or information (for example, “Google Shopping”, “Google Finance”, “Google Flights”, “Google Video”).19 In addition to the results returned in “Universals”  or  “OneBoxes”, Google's specialised search services can also be accessed through menu-type links displayed at the top of Google's search results pages.

(25)    Certain of Google’s specialised search services are based on paid inclusion. Third party websites have to enter into an agreement with Google in order to be listed in  the search results of such a specialised search service. In most instances, such an agreement provides for a payment based on a pay per click system. This is the case for instance for Google Shopping (see section 2.2.5).

2.2.5. Google's comparison shopping service

(26)    Google’s comparison shopping service is one of Google’s specialised search  services. In response to queries, it returns product offers from merchant websites, enabling users to compare them.

(27)    Google launched the first version  of  its  comparison  shopping  service  in  December 2002 in the US under the brand name “Froogle”. Froogle operated as a standalone website. Merchants did not have to pay to be listed in Froogle as it was monetised by advertisements. Google launched Froogle in the United Kingdom in October 200420 and in Germany in November 2004.21

(28)    In April 2007, Google renamed Froogle as “Google Product Search”22 and subsequently launched along the standalone Google Product Search website a dedicated “Universal” or “OneBox” for Google Product Search, referred to as the “Product Universal”. Google did not, however, change the business model of its comparison shopping service: like Froogle, merchants did not have to pay to be listed in Google Product Search as it was monetised by advertisements.

(29)    The Product Universal comprised specialised search results from Google Product Search, accompanied by one or several images and additional information such as the price of the relevant items. The results within the Product Universal, including the clickable images, in most cases led the user to the standalone Google Product Search websites. There was also a header link leading to the main website of Google Product Search.

(30)    The Product Universal was launched in October 2007 in the US, in January 2008 in the United Kingdom and Germany, in October 2010 in France and in May 2011 in Italy, the Netherlands and Spain.23

(31)    In May 2012, Google renamed Google Product Search as “Google Shopping” and revamped the Product Universal which was renamed first “Commercial Unit” and then “Shopping Unit”. At the same time, Google also changed the business model of its comparison shopping service (both the standalone website and the Universal) to a “paid  inclusion”  model,  in  which  merchants  pay  Google  when  their  product  is clicked on in Google Shopping, to more closely reflect the industry standard.24

(32)    In the same way as the Product Universal comprised specialised search results from Google Product Search, the Shopping Unit comprises specialised search results from Google Shopping, as illustrated by the screenshot below25. Those results are commercially  named  “Product  Listing  Ads”  –  PLAs.  Unlike  for  the Product Universal, however, the results within the Shopping Unit generally lead users directly to the pages of Google's merchant partners on which the user can purchase the relevant item.

(33)    The Shopping Unit and the standalone Google Shopping website were first launched in the US in May 2012, with the transition completed by autumn 2012.

(34)    Subsequently, the Shopping Unit was launched on Google's domains in the EEA as follows: (i) in February 2013 in the Czech Republic, France, Germany, Italy, the Netherlands, Spain and the United Kingdom; and (ii) in November 2013 in   Austria, Belgium, Denmark, Norway, Poland and Sweden.26 Google also started running a Shopping Unit experiment in Ireland in September 2016.27

(35)    As for the standalone Google Shopping website, it was launched (i) in February 2013 in the Czech Republic, France, Germany, Italy, the Netherlands, Spain and  the United Kingdom;28 (ii) in September 2016 in Austria, Belgium, Denmark, Norway, Poland and Sweden; and (iii) in Ireland in January 2017.29

2.2.6. Other Google products and services

(36)    Google offers a number of other online products and services free of charge, including an open-source operating system for smartphones and tablets (“Android”), an app store for the Android operating system (“Google Play”), an internet browser (“Google Chrome”), a web-based email account service (“Gmail”), a web mapping service (“Google Maps”), a file storage and editing service offering a suite of office applications (“Google Drive”), an instant messaging and video chat platform (“Google Hangouts”), a set of search tools (“Google Toolbar”) and a video-sharing website (“YouTube”).

(37)    Google also sells hardware products such as digital media players (“Chromecast”), laptops (“Chromebooks”), smartphones and tablets (“Google Nexus”).

3. PROCEDURE

(38)    The Commission has received a number of complaints against Google pursuant to Article 7(2) of Regulation (EC) No 1/2003. In addition, a number of cases pending before competition authorities of the Member States have been re-allocated to the Commission.

(39)    On 3 November 2009, Infederation Ltd. (“Foundem”) lodged a complaint against Google with the Commission. That complaint was replaced by a new version submitted by Foundem on 2 February 2010. On 10 February 2010, the Commission sent Google a non-confidential version of the complaint. On 3 May 2010, Google provided comments on the complaint.

(40)    On 22 January 2010, pursuant to Article 12 of Regulation (EC) No 1/2003, the Bundeskartellamt (Germany) exchanged information with the Commission on a complaint against Google lodged by Ciao GmbH (“Ciao”). Ciao’s complaint was re- allocated to the Commission on 22 January 2010 in accordance with the Commission’s Notice on cooperation within the Network of Competition Authorities.30 On 10 February 2010, the Commission sent a non-confidential version of the complaint to Google. On 20 March 2010, Google provided comments on the complaint.

(41)    On 2 February 2010, eJustice.fr (“eJustice”) lodged a complaint against Google with the Commission. On 10 February 2010, the Commission sent Google a non- confidential version of the complaint. On 3 May 2010, Google provided comments on the complaint. On 22 February 2011, eJustice’s parent company 1plusV lodged with the Commission a supplement to eJustice’s complaint and joined eJustice’s complaint. On 1 April 2011, the Commission sent a non-confidential version of the supplement to Google. On 16 September 2011, Google provided comments on the supplement.

(42)    On 26 October 2010, the Verband freier Telefonbuchverleger (“VfT”) lodged a complaint against Google with the Commission. On 1 December 2010, the Commission  sent  a  non-confidential  version  of  the  complaint  to  Google.  On  14 January 2011, Google provided comments on the complaint. On 24 January 2011, VfT submitted a supplement to its complaint.

(43)    On 30 November 2010, the Commission initiated proceedings against Google pursuant to Article 2(1) of Regulation (EC) No 773/2004 in relation to a number of practices. The initiation of proceedings relieved the competition authorities of the Member States of their competence to apply Articles 101 and 102 of the Treaty to those practices.

(44)    As a result, on 14 December 2010, the Bundeskartellamt transferred to the Commission a joint complaint against Google lodged by the Bundesverband Deutscher Zeitungsverleger (“BDZV”) and the Verband Deutscher Zeitschriftenverleger (“VDZ”), and complaints against Google lodged by Euro-Cities AG (“Euro-Cities”) and Hot Maps Medien GmbH (“Hot Maps”). Google had already provided comments on these complaints,  except for that lodged  by Hot  Maps. On  1 April 2011, the Commission sent a non-confidential version of Hot Maps’ complaint to Google. On 16 September 2011, Google provided comments on the complaint.

(45)    On 17 January 2011, the Autorità Garante della Concorrenza e del Mercato (Italy) also transferred to the Commission a complaint against Google lodged by Mr. Sessolo (“nntp.it”). On 13 April 2011, the Commission sent a non-confidential version of the complaint to Google. On 17 September 2011, Google provided comments on the complaint.

(46)    On 31 January 2011, Elf B.V. (“Elf”) lodged a complaint against Google with the Commission. On 1 April 2011, the Commission sent a non-confidential version of  the complaint to Google. On 16 September 2011, Google provided comments on the complaint.

(47)    On 31 March 2011, Microsoft Corporation (“Microsoft”) lodged a complaint against Google with the Commission. On 1 April 2011, the Commission sent a non- confidential version of the complaint to Google. On 16 September 2011, Google provided comments on the complaint.

(48)    On 27 June 2011, Twenga SA (“Twenga”) submitted an application to be heard as a third party within the meaning of Article 13 of Regulation (EC) No 773/2004 and Article 5 of Decision 2011/695/EU of the President of the European Commission of 13 October 2011 on the function and terms of reference of the hearing officer in certain competition proceedings (“Decision 2011/695”).31 The Hearing Officer approved Twenga’s application on 11 July 2011.

(49)    On 19 December 2011, La Asociación de Editores de Diarios Españoles (“AEDE”) lodged a complaint against Google with the Commission. On 25 January 2012, the Commission sent a non-confidential version of the complaint to Google. On 6 March 2012, Google provided comments on the complaint.

(50)    On 23 January 2012, Twenga lodged a complaint against Google with the Commission. On 7 February 2012, the Commission sent a non-confidential version  of the complaint to Google. On 28 February 2012 and 5 March 2012, Google provided comments on the complaint.

(51)    On 20 February 2012, Company E, a company wishing to remain anonymous, submitted an application to be heard as a third party within the meaning of Article 13 of Regulation (EC) No 773/2004 and Article 5 of Decision 2011/695. The Hearing Officer rejected Company E’s application on 28 February 2012.

(52)    On 26 March 2012, Yelp Inc. (“Yelp”) submitted an application to be heard as a third party within the meaning of Article 13 of Regulation (EC) No 773/2004 and Article  5 of Decision 2011/695, in the event that the Commission issued a statement of objections. The Hearing Officer acknowledged on 28 March 2012 that Yelp met the criteria to be heard as an interested third party in the proceedings.

(53)    On 29 March 2012, Streetmap EU Ltd (“Streetmap”) lodged a complaint against Google with the Commission. On 12 April 2012, the Commission sent a non- confidential version of the complaint to Google. On 23 May 2012, Google provided comments on the complaint.

(54)    On 30 March 2012, Expedia Inc. (“Expedia”) lodged a complaint against Google with the Commission. On 12 April 2012, the Commission sent a non-confidential version of the complaint to Google. On 24 May 2012, Google provided comments on the complaint.

(55)       In April 2012, Cases 39.768 (Ciao), 39.775 (eJustice/1PlusV), 39.845 (VfT),  39.863 (BDZV&VDZ), 39.866 (Elfvoetbal), 39.867 (Euro-Cities/HotMaps), 39.875 (nntp.it), 39.897 (Microsoft) and 39.975 (Twenga) were merged with Case 39.740 (Foundem). The documents that had been registered in the aforementioned case files were subsequently put into a single file under case number COMP/C-3/39.740 – Google Search. The Commission continued the procedure under case number COMP/C-3/39.740.

(56)    On 1 April 2012, Odigeo Group (“Odigeo”) lodged a complaint against Google with the Commission. On 12 April 2012, the Commission sent a non-confidential version of the complaint to Google. On 24 May 2012, Google provided comments on the complaint.

(57)    On 2 April 2012, TripAdvisor Inc. (“TripAdvisor”) lodged a complaint against Google with the Commission. On 12 April 2012, the Commission sent a non- confidential version of the complaint to Google. On 24 May 2012, Google provided comments on the complaint.

(58)    On 27 June 2012, Nextag Inc. (“Nextag”) and Guenstiger.de GmbH (“Guenstiger”) lodged a complaint against Google with the Commission. On 2 August 2012, the Commission  sent  a  non-confidential  version  of  the  complaint  to  Google.  On    8 September 2012, Nextag and Guenstiger submitted a supplement to their complaint. On 22 September 2012, Google provided comments on the complaint.

(59)    On 17 July 2012, the Hearing Officer approved an application from MoneySupermarket.com Group PLC (“MoneySupermarket”) to be heard as a third party within the meaning of Article 13 of Regulation (EC) No 773/2004 and Article  5 of Decision 2011/695.

(60)    On 11 January 2013, Visual Meta GmbH (“Visual Meta”) lodged a complaint against Google with the Commission. On 22 February 2013, the Commission sent a non- confidential version of the complaint to Google. On 4 June 2013, Google provided comments on the complaint.

(61)    On 30 January 2013, the Initiative for a Competitive Online Marketplace (“ICOMP”) lodged a complaint against Google with the Commission. On 22 February 2013, the Commission  sent  a  non-confidential  version  of  the  complaint  to  Google.  On    1 June 2013, Google provided comments on the complaint.

(62)    On 6 March 2013, Bureau Européen des Unions des Consommateurs AISBL (“BEUC”) submitted an application to be heard as a third party within the meaning of Article 13 of Regulation (EC) No 773/2004 and Article 5 of Decision 2011/695. The Hearing Officer approved BEUC’s application on 26 March 2013.

(63)    On 13 March 2013, the Commission adopted a Preliminary Assessment addressed to Google under Article 9 of Regulation (EC) No 1/2003 (“Preliminary Assessment”).  In the Preliminary Assessment, the Commission took the view that Google engages in the following business practices that may infringe Article 102 of the Treaty and Article 54 of the EEA Agreement:

–     The favourable treatment, within Google’s general search results pages, of  links to Google’s own specialised search services as compared to links to competing specialised search services (“first business practice”);

–     The copying and use by Google without consent of original content from third party websites in its own specialised search services (“second business practice”);32

–     Agreements that de jure or de facto oblige websites owned by third parties (referred to in the industry as “publishers”) to obtain all or most of their online search advertisement requirements from Google (“third business practice”);  and

–     Contractual restrictions on the management and transferability of online search advertising campaigns across online search advertising platforms (“fourth business practice”).

(64)    On 25 March 2013, Organización de Consumidores y Usuarios (“OCU”) submitted an application to be heard as a third party within the meaning of Article 13 of Regulation (EC) No 773/2004 and Article 5 of Decision 2011/695. The Hearing Officer approved OCU’s application on 8 April 2013. On 23 February 2017, OCU informed the Commission that it no longer wished to be considered as an interested third party on its own, but would only be involved in the case through BEUC.

(65)    Google did not agree with the legal analysis in the Preliminary Assessment and contested the assertion that any of the business practices described in it infringe Article 102 of the Treaty and Article 54 of the EEA Agreement. It nevertheless offered three sets of commitments pursuant to Article 9 of Regulation (EC) No 1/2003 to address the Commission’s competition concerns regarding the four business practices identified in the Preliminary Assessment. The first set of commitments was submitted to the Commission on 3 April 2013, the second set of commitments  on  21  October   2013   and   the   third   set   of   commitments   on  31 January 2014.

(66)    On 31 March 2014, BEUC lodged a complaint against Google with the Commission. On 2 April 2014, the Commission sent a non-confidential version of the complaint to Google.

(67)    On 15 April 2014, Open Internet Project (“OIP”) lodged a complaint against Google with the Commission. On 1 July 2014, the Commission sent a non-confidential version of the complaint to Google. On 18 August 2014, Google provided comments on the complaint.

(68)    On 16 May 2014, Deutsche Telekom AG (“Deutsche Telekom”) lodged a complaint against Google with the Commission. On 27 May 2014, Deutsche Telekom  submitted further information  on  the  allegations  covered  by  its  complaint.  On  25 June 2014, the Commission sent a non-confidential version of the complaint to Google. On 22 September 2014, Google provided comments on the complaint.

(69)    On 2 June 2014, Yelp lodged a complaint against Google with the Commission. On 25 June 2014, the Commission sent a non-confidential version of the complaint to Google. On 4 December 2014, Google provided comments on the complaint.

(70)    On 24 July 2014, HolidayCheck AG (“HolidayCheck”) lodged a complaint against Google with the Commission. On 1 September 2014, the Commission sent a non- confidential version of the complaint to Google. On 8 November 2014, Google provided comments on the complaint.

(71)    Between 27 May 2014 and 11 August 2014, the Commission sent letters pursuant to Article 7(1) of Regulation (EC) No 773/2004 (“the Article 7(1) letters”) to all complainants that had lodged  a  complaint under the  relevant proceedings before   27 May 2014.33 The letters outlined the Commission’s preliminary view that the third set of commitments offered by Google on 31 January 2014 could address the Commission’s competition concerns identified in the Preliminary Assessment. The Commission also informed the complainants in the Article 7(1) letters that it  intended to reject their complaints, to the extent that they related to the competition concerns identified in the Preliminary Assessment.

(72)    19 complainants submitted written observations in response to the Article 7(1) Letters (“the Article 7(1) replies”).34 Streetmap and nntp.it did not submit written observations to the Article 7(1) letters addressed to them by the Commission within the time-limit set within the letters. Streetmap’s and nntp.it's complaints were therefore deemed to have been withdrawn pursuant to Article 7(3) of Regulation (EC) No 773/2004.

(73)    Having received and analysed the written observations of the 19 complainants, the Commission considered that it was not in a position to adopt a decision under Article 9 of Regulation (EC) No 1/2003 making binding Google’s third set of commitments in relation to the four business practices covered by the Preliminary Assessment. The Commission brought this to the attention of Google on 4 September 2014.35

(74)    On 26 January 2015, Trivago GmbH (“Trivago”) lodged a complaint against Google with the Commission. On 16 September 2015, the Commission sent a non- confidential version of the complaint to Google.

(75)    On 13 April 2015, Company AC, a company that wishes to remain anonymous, submitted an application to be heard as a third party within the meaning of Article 13 of Regulation (EC) No 773/2004 and Article 5 of Decision 2011/695. The Hearing Officer approved Company AC’s application on 21 April 2015. The Hearing Officer also accepted Company AC's request for anonymity.

(76)    On 15 April 2015, the Commission reverted to the procedure under Article 7 of Regulation (EC) No 1/2003 in relation to the Conduct36 and adopted a Statement of Objections addressed to Google (“the SO”). In the SO, the Commission reached the provisional conclusion that the Conduct constitutes an abuse of a dominant position and, therefore, infringed Article 102 of the Treaty.

(77)    On 15 April 2015, News Corporation (“News Corp”) lodged a complaint against Google with the Commission. On 7 October 2015, the Commission sent a non- confidential version of the complaint to Google.

(78)    On 24 April 2015, FairSearch Europe (“FairSearch”) submitted an application to be heard as a third party within the meaning of Article 13 of Regulation (EC) No 773/2004 and Article 5 of Decision 2011/695. The Hearing Officer approved FairSearch’s application on 8 May 2015.

(79)    On 27 April 2015, the Commission granted Google access to file.

(80)    On 8 May 2015, SARL Acheter moins cher (“Acheter moins cher”) submitted an application to be heard as a third party within the meaning of Article 13 of Regulation (EC) No 773/2004 and Article 5 of Decision 2011/695. The Hearing Officer approved Acheter moins cher’s application on 11 May 2015.

(81)    On 20 May 2015, S.A. LeGuide.com (“LeGuide”) submitted an application to be heard as a third party within the meaning of Article 13 of Regulation (EC) No 773/2004 and Article 5 of Decision 2011/695. The Hearing Officer approved LeGuide’s application on 26 May 2015.

(82)    On 21 May 2015, Kelkoo SAS (“Kelkoo”) submitted an application to be heard as a third party within the meaning of Article 13 of Regulation (EC) No 773/2004 and Article 5 of Decision 2011/695. The Hearing Officer approved Kelkoo’s application on 26 May 2015.

(83)    On 1 June 2015, Getty Images, Inc. (“Getty”) submitted an application to be heard as a third party within the meaning of Article 13 of Regulation (EC) No 773/2004 and Article 5 of Decision 2011/695. The Hearing Officer approved Getty’s application on 4 June 2015.

(84)    On 10 June 2015, Myriad International Holdings B.V. (“MIH”) submitted an application to be heard as a third party within the meaning of Article 13 of Regulation (EC) No 773/2004 and Article 5 of Decision 2011/695. The Hearing Officer approved MIH’s application on 18 June 2015.

(85)    On 18 June 2015, ICOMP submitted to the Commission a supplement to its complaint of 30 January 2013. On 16 September 2015, the Commission sent a non- confidential version of the supplement to Google.

(86)    On 2 July 2015, Tradecomet.com Ltd and its parent company Tradecomet LLC (“TradeComet”) lodged a  complaint  against  Google  with  the  Commission.  On  18 September 2015, the Commission sent a non-confidential version of the complaint to Google. On 20 November 2015, Google provided comments on the complaint.

(87)    On 22 July 2015, the European Technology & Travel Services Association (“ETTSA”) submitted an application to be heard as a third party within the meaning of Article 13 of Regulation (EC) No 773/2004 and Article 5 of Decision 2011/695. The Hearing Officer approved ETTSA’s application on 23 July 2015.

(88)    On 20 August 2015, VG Media Gesellschaft zur Verwertung der Urheber- und Leistungsschutzrechte von Medienunternehmen mbH (“VG Media”) lodged a complaint against Google with the Commission. On 13 November 2015, the Commission  sent  a  non-confidential  version  of  the  complaint  to  Google.  On  22 January 2016, Google provided comments on the complaint.

(89)    Between June and September 2015, the Commission sent a non-confidential version of the SO to 24 complainants and 10 interested third parties.37 20 complainants and 7 interested third parties submitted written comments on the SO.38

(90)    On 27 August 2015, Google submitted its response to the SO (“the SO Response”). Google did not request the opportunity to express its views at an oral hearing pursuant to Article 12(1) of Regulation (EC) No 773/2004.

(91)    Between October and November 2015, the Commission sent a non-confidential version of the SO Response to 23 complainants and 9 interested third parties.39 14 complainants and 7 interested third parties submitted written comments on the SO Response.40

(92)    On 17 April 2016, News Corp lodged an additional complaint against Google with the Commission. On 3 May 2016, the Commission sent a non-confidential version of the additional complaint to Google. On 25 July 2016, Google provided comments on News Corp's complaint of 15 April 2015 and additional complaint of 17 April 2016.

(93)    On 21 April 2016, Microsoft and Ciao informed the Commission that they were withdrawing the complaints they had lodged against Google.

(94)    On 26 April 2016, Getty lodged a complaint against Google with the Commission. On 3 May 2016, the Commission sent a non-confidential version of the complaint to Google. On 5 August 2016, Google provided comments on the complaint.

(95)    On 4 May 2016, Promt GmbH (“Promt”) lodged a complaint against Google with the Commission. On 14 June 2016, the Commission sent a non-confidential version of the complaint to Google. On 10 August 2016, Google provided comments on the complaint.

(96)    On 14 July 2016, the Commission initiated proceedings against Alphabet pursuant to Article 2(1) of Regulation (EC) No 773/2004 and adopted a Supplementary Statement of Objections (“the SSO”) addressed to Google and Alphabet in relation to the Conduct. The SO was annexed to the SSO and was therefore also addressed to Alphabet.

(97)    On 27 July 2016, the Commission granted Google and Alphabet access to file.

(98)    On 18 August 2016, BDZV and VDZ lodged a supplement to their initial complaints with the Commission. On 16 December 2016, the Commission sent a non- confidential version of the supplement to Google and Alphabet.

(99)    Between September and October 2016, the Commission sent a non-confidential version   of   the   SSO   to   20   complainants   and   6   interested   third   parties.41   9 complainants and 3 interested third parties submitted written comments on the SSO.42

(100)    On 3 November 2016, Google and Alphabet submitted their response to the SSO (“the SSO Response”). Google and Alphabet did not request the opportunity to express their views at an oral hearing pursuant to Article 12(1) of Regulation (EC) No 773/2004.

(101)    On 28 February 2017, the Commission sent Google and Alphabet a letter of facts (“the LoF”) drawing their attention to pre-existing evidence that was not expressly relied on in the SO and the SSO, but which, on further analysis of the file, could be potentially relevant to support the preliminary conclusion reached in the SO and the SSO. The Commission also informed Google and Alphabet about additional  evidence brought to the Commission's attention after the adoption of the SSO that could also be potentially relevant to support the preliminary conclusion reached in  the SO and the SSO.

(102)    On 1 March 2017, the Commission granted Google and Alphabet access to file.

(103)    On 18 April 2017, Google and Alphabet submitted their response to the LoF (“the LoF Response”).43

(104)    On 19 May 2017, the Commission sent Google and Alphabet a letter providing updated versions of Tables 37 and Table 64 of the LoF.

(105)    On 31 May  2017,  Google  and  Alphabet  replied  to  the  Commission's  letter  of  19 May 2017.

In a submission of 26 June 2017, sent by e-mail from […] to […] at 23:40, Google and Alphabet provided the Commission with additional evidence based on Nielsen user studies allegedly showing "strong competition exercised by merchant platforms in product search and price comparison" essentially on the basis that "For the large majority of product searches, merchant platforms are users' preferred search option and most users start their product searches there." That submission was received after the Advisory Committee on Restrictive Practices and Dominant Positions had given its opinion on the draft Decision in two meetings held on 20 June 2017 and in the morning of 26 June 2017. The Nielsen studies are based on data  collected between October and December 2016. The Commission considers it highly unlikely, both given this and Google's general awareness that a Decision pursuant to Article 7 of Regulation (EC) No 1/2003 was a possibility, that Google could not have submitted the studies well in advance of the late evening of 26 June 2017. In accordance with settled case-law, the consultation of the Advisory Committee represents the final stage of the procedure before the adoption of the decision (Joined Cases 100 to 103/80, Musique Diffusion française and others v Commission, EU:C:1983:158, paragraph 35; Joined Cases T-213/01 and T-214/01 Österreichische Postsparkasse and Bank für Arbeit und Wirtschaft v Commission, EU:T:2006:151, paragraph 149). In addition, it follows from Article 11(1) of Regulation (EC) No 773/2004 that the undertaking concerned should in principle exercise its right to be heard before the Advisory Committee is consulted.

4. GOOGLE'S ALLEGATIONS THAT THE COMMISSION'S INVESTIGATION SUFFERS FROM PROCEDURAL ERRORS

(106)    Google alleges that the Commission's investigation suffers from a number of procedural errors. First, the Commission has breached Google's right to a sound administrative  process  by  failing  to  assess  the  evidence  properly.  Second, the Commission has breached Google's rights of defence by failing to explain its preliminary conclusions in the SO and the SSO properly. Third, the Commission has breached Google's rights of defence by not providing Google with adequate access to minutes of meetings with third parties. Fourth, the Commission has breached Google's rights of defence by failing to provide adequate reasons as to why  it reverted to the Article 7 procedure in 2014. Fifth, the Commission has breached Google's rights of defence by failing to provide sufficient information regarding the envisaged remedies.

(107)    For the reasons set out in recitals (110) to (111), (114) to (117), (120) to (122) and

(125) to (137), the Commission concludes that Google's right to a sound administrative process and its rights of defence have been respected throughout the investigation.

4.1. The Commission's alleged failure to assess the evidence properly

(108)    Google claims that the Commission has failed to assess the facts and evidence properly. In particular, the evidence relied upon by the Commission is of lower probative value than the evidence relied upon by Google.44 Google also claims that the Commission has used outdated information for the finding of dominance.

(109)    This claim is unfounded.

(110)    First, Google's claim is in effect a challenge to the merits of the Commission's assessment of the Conduct and is therefore irrelevant to the question of whether the Commission has complied with Google's right to a sound administrative process.

(111)    Second, the Commission's finding of an infringement does not rely “largely on blog posts, news articles, incomplete traffic data, and unverified assertions by complainants”. Rather, the finding is largely based on a body of evidence of high probative value, including traffic data from Google's logs, user surveys, contemporaneous internal Google documents, experimental analyses and responses to requests of information under Article 18(2) and 18(3) of Regulation (EC) No 1/2003.

4.2. The Commission's alleged failure to explain its preliminary conclusions  properly

(112)    Google claims that its rights of defence have been infringed because, in the SSO, the Commission failed to explain how and why additional items of evidence supported the preliminary conclusions expressed in the SO.45 In addition, the SO and the SSO relied on “vague and obscure terminology” such as “extracted from” and “emanation”, without explaining what those words mean or what they relate to.46

(113)    Both claims are unfounded.

(114)    First, for each additional piece of evidence, the SSO set out the precise conclusion of the SO that it further supported.

(115) Second, it is apparent from Google's submissions that it well understood the terminology relied on by the SO and the SSO and what that terminology related to.

(116)  In the first place, Google explained in the SO Response why it considered that “[i]t is wrong to say that product ads shown in Shopping Units are 'extracted […] from Google’s own comparison shopping service'”.47

(117)  In the second place, Google devoted three pages of the SSO Response to explaining why it considered that “the Shopping Unit is not an 'emanation' of a Google comparison shopping unit”.48

4.3. The Commission's alleged failure to provide adequate access to minutes of meetings with third parties

(118)  Google claims that its rights of defence have been infringed because the minutes of meetings with third parties that have been provided to it “only list the topics discussed, without recording their substance”.49

(119)  This claim is unfounded.

(120)  First, in competition proceedings, the Commission is under no general duty to establish records of the discussions that it has had with complainants or other parties during the meetings or telephone conversations held with them.50 The Commission is also not required to make all interviews subject to the formal requirements provided for in Article 19(1) of Regulation (EC) No 1/2003, read in conjunction with Article 3 of Regulation (EC) No 773/2004.51 The Commission is required to  establish  adequate documentation of incriminatory or exculpatory information that is of a certain importance and which bears an objective link with the subject-matter of an investigation.52 Such documentation can take the form of a succinct note in the file to which the undertakings concerned have access.53

(121)  Second, the Commission has met that obligation in this case by providing Google with succinct notes containing the minutes of the meetings and telephone conversations held between representatives of the Commission and complainants and other third parties in the context of the investigation. Such minutes mention the undertaking(s) attending the meeting or participating in the telephone conversation (save for certain anonymity requests) and the timing and topic(s) covered by the meeting or telephone conversation.

(122) Third, the discussions that the Commission has had with third parties during the meetings or telephone conversations took place at the request of the third parties and were used by those third parties as an opportunity to urge the Commission to take action on the case quickly. This is confirmed by the fact that: (i) this Decision does not rely on any incriminating information provided by third parties during meetings or telephone conversations; and (ii) Google has not identified, on the basis of the timing  and  topic(s)  covered  by  the  meetings  or  telephone  conversations, any potentially exculpatory information of which the Commission should have made an adequate record accessible.

4.4. The Commission's alleged failure to explain adequately why it reverted to the Article 7 procedure

(123)  Google claims that the Commission has failed to provide adequate reasons why it reverted to the Article 7 procedure in 2014.54 Google also claims that the  Commission failed to provide it with an opportunity to modify the third set of commitments.55

(124)  Both claims are unfounded.

(125)  First, the Commission is not required to give reasons as to why it reverted to the Article 7 procedure in 2014. Such an argument was already rejected by the Court of Justice in Alrosa, where it held that “the General Court based its reasoning on the incorrect proposition that the Commission was required to give [Alrosa] reasons for rejecting the joint commitments”.56

(126)  Second, and in any event, the Commission has provided adequate reasons as to why  it reverted to the Article 7 procedure in 2014. These reasons were already referred to in the SSO.57

(127)  In the first place, notwithstanding the third set of commitments, other comparison shopping services may have been unable to compete with Google's own comparison shopping service on the merits, because their results would still have been displayed after results from Google's comparison shopping service.

(128)  In the second place, if the auction mechanism for the display of results from competing comparison shopping services foreseen by the third set of commitments had been adopted, this might have led to Google receiving the majority of any profits generated by clicks on these results, whereas Google's own comparison shopping service would not have had to incur any cost for the display of its results.

(129)  In the third place, in light of the Article 7(1) replies, the Commission had concerns about the effectiveness of the third set of commitments to address the competition concerns […].

(130)  Third, the adequacy of the reasons is not called into question by the position outlined in the press material of 5 February 2014 and in the Article 7(1) letters.

(131)  In the first place, the position outlined in the press material of 5 February 2014 and the Article 7(1) letters was necessarily preliminary and potentially subject to change.58

(132)  In the second place, a letter of Commission Vice President Almunia to Google of 22 July 2014 anticipated the possible need for Google to offer revised commitments in the light of the replies to the Article 7(1) letters.

(133)  Fourth, the Commission was not required to provide Google with an opportunity to modify the third set of commitments. Such an argument was also rejected by the Court of Justice in Alrosa, where it held that “the General Court based its reasoning on the incorrect proposition that the Commission was required to suggest to Alrosa that it offer new joint commitments with De Beers”.59

(134)  Fifth, and in any event, the Commission did provide Google with an opportunity to modify the third set of commitments.

(135)  In the first place, Google was informed on 4 September 2014 of the reasons why the Directorate-General for Competition considered that the third set of commitments was insufficient to address adequately the Commission's concerns.

(136)  In the second place, those reasons were further discussed and explained during a meeting on 8 September 2014 and in various follow-up conversations.

(137)  In the third place, in its submission of 7 October 2014, Google indicated that it did not intend to offer revised commitments because to do so “would  considerably change the nature of the rival link remedy compared to Google’s February 2014 proposal and […] would necessitate a deeper integration of Google and rival search results”.

4.5. The Commission's alleged failure to provide sufficient information regarding the envisaged remedies

(138 Google claims that the Commission has failed to provide it with all the information  to enable it to defend itself as to the necessity and proportionality of the envisaged remedies.60

(139) This claim is unfounded.

(140) First, a statement of objections must set out, even briefly, but in sufficiently clear terms, the measures which the Commission intends to take in order to bring an end to an infringement and give the undertaking concerned all the information necessary to enable  it  properly  to   defend  itself.61    The  SO  and  the  SSO   provided  all the information necessary to enable Google to defend itself properly regarding the envisaged remedies.

(141)  In the first place, the SO and the SSO indicated that Google may be required to position and display competing comparison shopping services in the same way as it positions and displays its own comparison shopping service in its general search results pages.62

(142) In the second place, the SO identified three specific ways in which Google may be required to implement such a requirement.63

(143) Second, the positioning and display of competing comparison shopping services in Google's general search results pages had already been extensively discussed during the Article 9 procedure.64

(144) Third, the Commission was not required to specify further “how its requested remedy is meant to work”. It is for Google and Alphabet, and not the Commission, to make a choice between the several possible lawful ways of positioning and displaying competing comparison shopping services in the same way as Google positions and displays its own comparison shopping service in Google's general search results pages, thereby bringing the infringement to an end.65

5. MARKET DEFINITION

5.1. Principles

(145) For the purposes of investigating the possible dominant position of an undertaking on a given product market, the possibilities of competition must be judged in the context of the market comprising the totality of the products or services which, with respect to their characteristics, are particularly suitable for satisfying constant needs and are only to a limited extent interchangeable with other products or services.66

(146) Moreover, since the determination of the relevant market is useful in assessing whether the undertaking concerned is in a position to prevent effective competition from being maintained and to behave to an appreciable extent independently of its competitors and its customers, an examination to that end cannot be limited solely to the objective characteristics of the relevant services, but the competitive conditions and the structure of supply and demand on the market must also be taken into consideration.67

(147) A relevant product market comprises all those products and/or services which are regarded as interchangeable or substitutable by the consumer, by reason of the products' characteristics, their prices and their intended use.68

(148) The relevant geographic market comprises the area in which the undertakings concerned are involved in the supply and demand of products or services, in which the conditions of competition are sufficiently homogeneous and which can be distinguished from neighbouring areas because the conditions of competition are appreciably different in those areas.69

(149) Undertakings are subject to three main sources of competitive constraints: demand substitutability, supply substitutability and potential competition. From an economic point of view, for the definition of the relevant market, demand substitution constitutes the most immediate and effective disciplinary force on the suppliers of a given product.70

(150) Supply side substitutability may be taken into account in situations in which its effects are equivalent to those of demand substitution in terms of effectiveness and immediacy. There is supply side substitution when suppliers are able to switch production to the relevant products and market them in the short term without incurring significant additional costs or risks in response to small and permanent changes in relative price.71

(151) The distinctness of products or services for the purpose of an analysis under Article 102 of the Treaty has to be assessed by reference to customer demand.72 Factors to  be taken into account include the nature and technical features of the products or services concerned, the facts observed on the market, the history of the development of the products or services concerned and also the undertaking’s commercial practice.73

(152) The fact that a product or service is provided free of charge does not prevent the offering of such a service from constituting an economic activity for the purposes of the competition rules of the Treaty.74 This is simply a factor to be taken into account in assessing dominance.

(153) The definition of the relevant market does not require the Commission to follow a rigid hierarchy of different sources of information or types of evidence. Rather, the Commission must make an overall assessment and can take account of a range of tools for the purposes of that assessment.75

5.2. The relevant product markets

(154) The Commission concludes that the relevant product markets for the purpose of this case are the market for general search services (section 5.2.1) and the market for comparison shopping services (section 5.2.2).

5.2.1. The market for general search services

(155)  A number of different companies offer general search services. Some of them use their own search technology, such as Google, Microsoft (Bing)76 and Seznam.77 Others show results of a third party general search engine (often Google or Bing) with which they have an agreement. This is the case for instance, for Yahoo78 and America Online79, both of which return Bing generic search results, and Ask80, which returns Google generic search results.

(156) The Commission concludes that the provision of general search services constitutes a distinct product market. First, the provision of general search services constitutes an economic activity (section 5.2.1.1). Second, there is limited demand side substitutability (section 5.2.1.2) and limited supply side substitutability (section 5.2.1.3) between general search services and other online services. Fourth, this conclusion does not change if general search services on static devices versus mobile devices are considered (section 5.2.1.4).

5.2.1.1. The provision of general search services constitutes an economic activity

(157) For the reasons set out below, the provision of general search services constitutes an economic activity.

(158)  First, even though users do not pay a monetary consideration for the use of general search services, they contribute to the monetisation of the service by providing data with each query. In most cases, a user entering a query enters into a contractual relationship with the operator of the general search service. For instance, Google’s Terms of Service provide: “By using our Services, you agree that Google can use such data in accordance with our privacy policies”.81 In accordance with its privacy policies, Google can store and re-use data relative to user queries.82  The terms and conditions of other providers of general search services contain similar provisions.83 The data which users agree to allow a general search engine to store and re-use is of value to the provider of the general search service as it is used to improve the relevance of the search service and to show more relevant advertising.84

(159) Second, offering a service free of charge can be an advantageous commercial strategy, in particular for two-sided platforms such as a general search engine platform that connect distinct but interdependent demands. In two-sided platforms, two distinct user groups interact. At least for one of these users groups, the value obtained from the platform depends on the number of users of the other class. General search services and online search advertising constitute the two sides of a general search engine platform. The level of advertising revenue that a general search engine can obtain is related to the number of users of its general search service: the higher the number of users of a general search service, the more the online search advertising side of the platform will appeal to advertisers.85

(160) Third, even though general search services do not compete on price, there are other parameters of competition between general search services.86 These include the relevance of results, the speed with which results are provided87, the attractiveness of the user interface88 and the depth of indexing of the web.89

5.2.1.2. Limited demand side substitutability with other online services

(161) General search services are not the only means by which users can explore the web. Alternative ways of discovering content include content sites, specialised search services and social networks.

(162) For the reasons set out below, there is, however, limited demand side substitutability between general search services and other online services.

5.2.1.2.1. General search services versus content sites

(163) There is limited substitutability between general search services and content sites.

(164) First, the two types of services serve a different purpose. On the one hand, a general search service primarily seeks to guide users to other sites. As Google indicates on its website: “[O]ur goal is to have people leave our website as quickly as possible”.90 On the other hand, while content sites may contain references to other sites, their primary purpose is to offer directly the information, products or services users are looking for. Well-known examples of content sites include Wikipedia, IMDb,91 and websites of newspapers and magazines such as The New York Times or Nature.92

(165) Second, content sites that offer sophisticated content search functionality on their websites93 are not substitutable for general search services. Such content search functionality remains limited to their own content or content from partners and does not allow users to search for all content over the internet, let alone all information on the web.

5.2.1.2.2. General search services versus specialised search services

(166) There is also limited substitutability between general search services and specialised search services.

(167) First, the nature of specialised search services and general search services is  different. Specialised search services do not aim to provide all possible relevant results for queries; instead, they focus on providing specific information or purchasing options in their respective fields of specialisation. A specialised search service will also often cover a content category which is monetisable.94 By contrast, general search services search the entire internet and therefore generally return different, more wide-ranging results. Their services are not limited beforehand to one of several particular content categories.95

(168) Second, there are a number of differences in the technical features of specialised and general search services. In the first place, specialised search services and general search services often rely on different sources of data: the main input for general search services originates from an automated process called “web crawling”96, whereas many specialised search services rely on user input or information supplied by third parties.97 In the second place, specialised search services are usually monetised in a different way; in addition to relying on online search advertising, they generate revenue from, for instance, paid inclusion, service fees or commissions (pay-per-acquisition fees).98

(169) Third, the facts observed in the market, the history of the development of the products concerned and Google’s commercial practice further support the conclusion that specialised search services and general search services are different.

(170) In the first place, a wide variety of specialised search services have been offered on a standalone basis for several years. Examples include services specialised in search  for products such as Shopzilla, LeGuide, Idealo, Beslist, Kelkoo and Twenga, in search for local businesses such as Yelp, in search for flights such as Kayak and EasyVoyage, and in search for financial services such as MoneySupermarket or confused.com.99 None of these companies offers a general search service.

(171) In the second place, Google offers and describes its specialised search services as a service distinct from its general search service. Google has a help page on its website which purports to list its different products and services. The page comprises a list of Google products and services. It distinguishes between, on the one hand, “Web Search - Search billions of web pages”, with a link to Google's general search service,  and,  on  the  other  hand  “Specialized  Search”,  which  comprises   several different services, including for instance “Google Shopping - Search for stuff to buy”.100

(172) Regarding specifically Google’s comparison shopping service, Google offers that service as a separate standalone service, and describes its functionality and purpose differently to how it describes its general search service. For example, Google Shopping UK, Google’s UK comparison shopping service, is described within Google UK’s general search results pages following a query for “Google Shopping” as “Google's UK price comparison service, with searches by keyword, which can  then be broken down by category and ordered by price”. There is also a dedicated information page about Google Shopping, entitled “About Google Shopping”. It describes Google Shopping as a distinct “product discovery experience”: “Google Shopping is a new product discovery experience. The goal is to make it easy for  users to research purchases, find information about different products, their features and prices, and then connect with merchants to make their purchase”.101 The page then elaborates on the many search tools that are specific to Google Shopping:  “When you do a search within Google Shopping you'll see a variety of filters (like price, size, technical specifications, etc.) on the left side of the page that can help you quickly narrow down to the right product. You can also choose to display items in either a list or grid view by selecting one of these options in the top right hand  corner above the results. When viewing certain apparel product detail pages (like dresses, coats and shoes), you'll also see items that are “visually similar” to the item you've selected. These are just a few of the many tools within Google Shopping and we look forward to providing more in the future”.102 By contrast, Google’s general search  service  is  described  within  Google  UK’s  general  search  results pages following a query for “Google” in a more general manner: “The local version of this pre-eminent search engine, offering UK-specific pages as well as world results”.

(173) In the third place, reports of specialised market observers such as Nielsen and comScore distinguish between general search services and other search services.103

(174) Fourth, contrary to what Google claims,104 even though search results provided by a general search service may sometimes overlap with the results provided by a specialised search service, the two types of search services act as complements rather than substitutes.105

(175) In the first place, a general search service is the only online search service on which users can seek potential relevant results from all categories at the same time.

(176) In the second place, specialised search services offer certain search functionalities that do not exist, or not to the same extent, on general search services. For instance, on search services specialised in travel, users may look for hotels with a certain number of stars, or within a certain range of a city, or they may read user reviews of these hotels. These functionalities are unavailable to the same extent on a general search service for the same queries.

(177) In the third place, a substantial number of users reach to specialised search services only after having first entered a query in a general search service.106

5.2.1.2.3.  General search services versus social networking sites

(178) There is also limited substitutability between general search services and social networking sites.

(179) First, general search services and social networking sites perform different functions. While general search services help users to find content they are looking for, social networks lead users to content they might be interested in by offering a means for users to connect and interact with people who, for instance, share interests or activities.

(180) Second, while certain social networks offer a general search function on their websites, so that users do not need to leave the sites to perform a general search, none of these sites uses its own general search technology. Instead, they rely on existing third party search services to power these searches. For example, Facebook previously relied on Microsoft’s Bing to provide search results.

(181) Third, the volume of general searches performed on social networks represents only  a small share of the total volume of general searches. For example, in 2011, the number of general searches performed via Facebook in Europe107 was equivalent to only 3.2% of the number of general searches performed on Google Search, even though Facebook is by far the largest social network.108 […]109

(182)     Fourth, […]110

(183)    Fifth, regarding Google's claim that social networking sites are “increasingly active in product search”,111 this can indicate at most that comparison shopping services  and social networking sites are substitutable, not that general search services and social networking sites are substitutable. Similarly, regarding Google's claim that “merchants and other business set up pages on social media sites with links to their websites and actively post on these sites”,112 this indicates at most that merchants and other businesses use social media sites as a tool to reach potential customers, not that these social media sites offer a comparison shopping service (as defined in section 5.2.2).

5.2.1.3. Limited supply side substitutability with other online services

(184)    Supply side substitutability is also limited.

(185)    In order to offer general search services, providers of other online services would need to make significant investments in terms of time and resources, particularly the initial costs associated with the development of algorithms and the costs of crawling and indexing the data. These barriers to entry are described in more detail in section 6.2.2.

5.2.1.4. General search services on static devices versus mobile devices

(186)    For the reasons set out in recitals (187) to (189), general search services offered on static devices such as desktop and laptop PCs and on mobile devices such as smartphones and tablets belong to the same relevant product market.

(187)    First, general search services are active on both types of devices. Although the user interface is different, the underlying technology is the same.113

(188)    Second, general search services on static and mobile devices are offered by the same undertakings.

(189)    Third, Google does not contest that general search services offered on static devices and on mobile devices are part of the same relevant product market.

(190)    In any event, even if general search services on static devices were to constitute a distinct product market from general search services on mobile devices, this would not alter the Commission’s view as regards dominance (see section 6.2.6).

5.2.2. The market for comparison shopping services

(191)    Comparison shopping services are specialised search services that: (i) allow users to search for products and compare their prices and characteristics across the offers of several different online retailers (also referred to as online merchants) and merchant platforms (also referred to as online marketplaces);115 and (ii) provide links that lead (directly or via one or more successive intermediary pages) to the websites of such online retailers or merchant platforms.116

(192)    The Commission concludes that the provision of comparison shopping services constitutes a distinct relevant product market. This is because comparison shopping services are not interchangeable with the services offered by: (i) search services specialised in different subject matters (such as flights, hotels, restaurants, or  news); (ii) online search advertising platforms; (iii) online retailers; (iv) merchant platforms; and (v) offline comparison shopping tools.

5.2.2.1.Comparison shopping services versus other specialised search services

(193)    There is limited substitutability between comparison shopping services and other specialised search services.

(194)    From the demand side perspective, each type of service focuses on providing specific information from different sources in its respective field of specialisation.117 Thus, comparison shopping services provide users that are looking for information on a product with a selection of existing commercial offers available on the internet for that product, as well as tools to sort and compare such offers based on various criteria.118 From the perspective of those users, such a service is not substitutable  with that offered by search services specialised in different subject matters such as flights,119 hotels, locals (such as restaurants),120 and news.

(195)    From the supply side perspective, each specialised search service selects and ranks results through specific criteria that rely on dedicated signals and formulas designed to determine the relevance of a particular information type (e.g. price, product information,   merchant   rating,   product   popularity,   stock   level for comparison shopping services; freshness for news aggregators; proximity to the user for specialised search services focused e.g. on restaurants).121 Moreover, specialised search services mostly select content from a pool of relevant suppliers with whom they have contractual relationships and those suppliers must provide input to databases and other data infrastructure operated by specialised search service providers.122 Each specialised search service therefore needs to develop and maintain a dedicated data infrastructure and structured relationships with relevant suppliers. Comparison shopping services typically employ a commercial workforce whose role is to enter into agreements with online retailers, pursuant to which these retailers  send them feeds of their commercial offers. These services are only partially automated and involve commercial relationships with online retailers.123 Likewise, flight search services use proprietary databases of content that are usually updated in real-time to ensure that they provide the most current possible information and have contractual arrangements with the booking websites, which remunerate them either  by paying a commission per flight ticket booked or on a cost-per-click basis.124 From the supply side perspective, substitutability is therefore also limited because providers of other specialised search services are not in a position to start providing comparison shopping services in the short term and without incurring significant additional costs.

5.2.2.2. Comparison shopping services versus online search advertising platforms

(196)    There is also limited substitutability between comparison shopping services and online search advertising platforms.

(197)    From the demand side perspective, while online retailers generally promote their offers through both comparison shopping services and online search advertising platforms, the latter do not provide services that are interchangeable from the perspectives of users and online retailers (and other advertisers).

(198)    First, users perceive comparison shopping services as a service to them and  navigate

– either directly (albeit to a limited extent) or (mostly) through a general search service (see section 7.2.4.1) – to a comparison shopping website (including Google Shopping) to search for a product and receive specialised search results. By contrast, users do not perceive online search advertising as a service to them and do not enter  a query in a general search engine specifically in order to receive search advertising results. Indeed, Google does not offer a standalone service which users can navigate directly to in order to receive search advertising. Online search advertising is therefore not a service that users seek, but rather a compensation for the free service offered by general search engines (similar to the advertising on free-to-air TV).

(199)    Second, comparison shopping services and online search advertising platforms are also complementary and not substitutable from the perspective of online retailers and other advertisers.

(200)    In the first place, only specific subsets of advertisers (i.e. online retailers and merchant platforms) can bid to be listed in comparison shopping services whereas  any advertiser can bid to be listed in online search advertising results.125

(201)    In the second place, participation in comparison shopping services involves different conditions than in online search advertising results, including the provision to comparison shopping services of structured data in the form of feeds. For instance, online retailers and merchant platforms wishing to be listed in Google Shopping need to give Google dynamic access to structured information on the products that can be purchased on their websites, including dynamically adjusted information on prices, product descriptions and the number of items available in their stock (see also recital (432)).126

(202)    In the third place, comparison shopping services display their results in richer  formats than online search advertising results.127

(203)    In the fourth place, the results of comparison shopping services are ranked based on different algorithms that take into account different parameters, tailored to the relevant specialised search category (see also recital (429)).128

(204)    In the fifth place, unlike for online search advertising platforms, in order to appear in comparison shopping services such as Google Shopping, third party websites bid on products and not on keywords (see also recital (426)).129

(205)    In the sixth place, when Google’s own comparison shopping service appears in Google's general search results page, AdWords results may also appear, thus  showing that the two services are complementary from Google's perspective.

(206)    From the supply side perspective, the functionalities and infrastructures required for the provision of comparison shopping services (see recital (195)) are different from those required for the provision of online search advertising. In particular, the provision of online search advertising services requires a company to invest in the development of a general search engine with a technology allowing users to search for keywords that can be matched with online search advertisements, as well as in a search advertisement technology to match keywords entered by users in their queries with relevant online search advertisements. According to Google, this is “[t]he most significant task that  [a company]  wishing to provide a search advertising solution might consider undertaking […]”.130

5.2.2.3. Comparison shopping services versus online retailers

(207)    There is also limited substitutability between comparison shopping services and online retailers.

(208)    From the demand side perspective, comparison shopping search services serve a different purpose than online retailers.

(209)    On the one hand, comparison shopping services act as intermediaries between users and online retailers. They allow users to compare offers from different online retailers in order to find the most attractive offer. They also do not offer users the possibility to purchase a product directly on their websites but rather seek to refer users to third-party websites where they can buy the relevant product.131  Indeed, far from being competitors, comparison shopping services consider online retailers as business partners or customers, with which they enter into contractual relationships.132

(210)    On the other hand, online retailers do not allow on their websites the possibility to compare their own offers with the offers for the same or similar products on the websites of other online retailers.133 Online retailers also do not seek to refer users to third-party websites; rather, they want users to buy their products without leaving their own websites; accordingly, they also offer after-sale support, including product return functionality.

(211)    From the supply side perspective, the provision of comparison shopping services and online retailers requires different functionalities and infrastructures. Comparison shopping services have to retrieve the relevant information in response to each query by analysing the real-time feeds from as many online retailers as possible. By contrast, online retailers host and manage the inventory of their selected manufacturers and provide “check-out” and payment functionalities.

(212)    The limited demand side and supply side substitutability between comparison shopping services and online retailers is not called into question by Google's claims that (i) online retailers also offer product search functionalities on their websites; and (ii) many users navigate directly to the websites of online retailers to find products, thereby bypassing comparison shopping services.134

(213)    First, the search functionality offered by online retailers is limited to products on  their own website(s) and does not offer the possibility to compare the offers on their website with the offers for the same or similar products on the websites of other online retailers.

(214)    Second, the fact that users may navigate directly to the websites of certain online retailers indicates that the service offered by online retailers is complementary and not substitutable with comparison shopping services from the perspective of users.

(215)    On the one hand, users navigate directly to the websites of online retailers when they want to buy a product and are not particularly interested in comparing offers and prices from different sellers. On the other hand users visit a comparison shopping service when they want to search and compare offers from different online retailers in order to find the most attractive offer. As acknowledged by Google,  online retailers do not offer this service, because “they typically do not list offers from other retailers”.135

5.2.2.4. Comparison shopping services versus merchant platforms

(216)    Contrary to what Google claims,136 there is also limited substitutability between comparison shopping services and merchant platforms, such as Amazon Marketplace and eBay Marketplaces.

(217)    From the demand side perspective, while comparison shopping services and  merchant platforms both aggregate offers from different sellers and provide a search functionality to search and filter those offers based on certain criteria,137 they serve a different purpose for users and for online retailers.

(218)    First, regarding the different purpose served by comparison shopping services and merchant platforms for users, on the one hand, as noted in recital (209), comparison shopping services: (i) act as intermediaries between users and online retailers/merchant platforms, allowing users to compare offers from different online retailers/merchant platforms in order to find the most attractive offer; (ii) do not offer users the possibility to purchase a product directly on their websites but rather seek  to refer users to third-party websites where they can buy the relevant product;138 (iii) do not offer after-sale support, including product return functionality;139 (iv) typically list offers only from professional sellers for new products.140

(219)    On the other hand, like online retailers, merchant platforms: (i) act as a place where retailers and consumers can conclude sales.141 Indeed, a number of merchant platforms (e.g. Amazon, Fnac, Rue Du Commerce) are online retailers that have also decided to include and sell on their websites third party products, mainly to complement  their  own  offering;142   (ii)  are  perceived  by  users  (and   comparison shopping services) as multi-brand retailers, i.e. a final destination where users can  buy products;143 (iii) offer after-sale support (including product return functionality and in some cases indemnification against any problem with the retailers);144 and (iv) list offers for second-hand products from non-professional sellers.145

(220)    The different purpose served by comparison shopping services and merchant platforms for users is confirmed by the following evidence:

(1)             The   fact   that   Google    itself   distinguishes the     different     purpose   and characteristics of, respectively, Google Shopping and of merchant platforms:

–    on its webpage “About Google Shopping”, Google indicates that “Google Shopping does not sell products directly to shoppers, instead we collect product information from participating merchants and make those products searchable for you. Our job is to find the product you want and point to the store that sells it. […] You can't buy products directly on www.google.com/shopping. We help you find the items you're looking for and  point  you  to  stores  (both  online  and  offline)  that  offer  them for sale”.146

–    on its webpage “Marketplaces, Marketplace sellers and Aggregators”, Google defines “marketplaces” (i.e. merchant platforms) as: “commerce sites that host items and/or websites of multiple individual sellers on the same domain [and] must provide a way for users to purchase the product online through either a payment service or directly from [the merchant platforms] website”.147

(2)             The fact that Google allows merchant platforms, but not competing comparison shopping services, to participate in Google Shopping:148

–    The lists of the top fifty customers of Google Shopping in the EEA based on 2015 revenues include several merchant platforms ([…]).149 This is true both for the whole EEA and in each of the thirteen EEA countries in which the Conduct was taking place on 3 May 2016 (Austria, Belgium, the Czech Republic, Denmark, France, Germany, Italy, the Netherlands, Norway, Poland, Spain, Sweden, and the United Kingdom). […]150

–       […]151

(3)             The fact that a majority of comparison shopping services152 and merchant platforms153  that replied to the Commission's requests for information indicated that they are – and consider themselves as – rather business partners in a vertical relationship, not competitors.

(4)             The fact that comparison shopping services list offers from merchant platforms based on the same terms and conditions applied to online retailers154 and that eBay and Amazon (followed by other merchant platforms) are consistently among the top online retailers in terms or revenues for many comparison shopping services.155

(5)             Internal Google documents indicating that Google and Amazon are business partners in a vertical relationship, not competitors:

(a)     […]156

(b)    […]157

(c)     […]158

(d)    […]159

(e)     […]160

(f)     […]161

(g)    […]162

(h)    […]163

(i)     […]164

(j)     […]165

(k)    […]166

(l)     […]167

(6)             The responses to a survey carried out in the context of the Commission's 2014 market study on “Comparison Tools and Third-Party Verification Schemes”168 that focussed, inter alia, on consumers' perceptions of, respectively: (i) “price comparison websites” (comparison shopping services); and (ii) “multi-trader e- commerce websites” defined as “online marketplaces selling products from a range of different retailers, for example Amazon, eBay, and Allegro” (merchant platforms).169 On the one hand, 91% of the respondents indicated that comparison shopping services allow customers to compare prices, with 79% indicating that the price comparison aspect is by far their most valued characteristic for these services. On the other hand, “[r]espondents also had a very specific perception of multi-trader e-commerce platforms (e.g. Amazon, eBay or Allegro)”: 62% of the respondents answered that  “these platforms  were mainly dedicated to buying products”.

(221)    Second, regarding the different purpose served by comparison shopping services and merchant platforms for online retailers, on the one hand, comparison shopping services offer online retailers the opportunity to promote their offerings to a large audience of users in search of a specific product. This allows online retailers both to increase brand awareness and to attract user traffic to their own websites, while retaining full control over their retail activities (including the merchandising strategy, the relationships with customers and the handling of the transactions). Comparison shopping services therefore tend to list offers from larger retailers that do not want to cede the customer interaction and data about their business and their customers to merchant platforms, such as Amazon, which they view as competitors.170

(222)    On the other hand, merchant platforms offer a full service to retailers that want to concentrate on sourcing and logistics only. Indeed, unlike comparison shopping services (including Google Shopping), merchant platforms also list offers from brick- and-mortar retailers which do not have their own website. Merchant platforms therefore take care of all or most of the other aspects associated with online retailing, including customer relationships, online store design, the maintenance and handling of online transactions, and customer care (including, as noted above, after-sale services and handling of complaints).171 Merchant platforms therefore tend to list offers mostly from certain small and medium-sized professional retailers that have limited brand awareness and/or are unable or unwilling to develop and maintain their own online stores (or even a website) or even non-professional sellers (including for second-hand products).172

(223)    The different purpose served by comparison shopping services and merchant platforms for online retailers is confirmed by the following evidence:

(1)             A number of internal Google documents:

(a)     […]173

(b)     […]174

(2)             The reply of […] to the Commission’s request for information of […]175 which confirms that out of the top […] Google Shopping customers176 based on 2015 revenues in each of the thirteen EEA countries in which the Conduct takes place […] a limited number had a registered account177 on […]  in  2015: […].178

(3)             The reply of […] to the Commission’s request for information […]179 which indicates that out of the top […] Google Shopping customers180 based on 2015 revenues in each of the thirteen EEA countries in which the Conduct takes place, […] a limited number had an active account181 on […] in 2015: […].182

(4)             The fact that in 2015, […] listed offers from thirteen of the top fifteen online retailers in France, in terms of revenues, whereas none of these top fifteen online retailers were present on Amazon.183

(5)             A comparison submitted by […] of the top ten sellers listed by Amazon and […] for the same product. Of the top ten sellers listed by Amazon, only three had their own standalone website and each of those websites had less than  1000 visits per month. By contrast, each of the top ten sellers listed by […] had their own standalone website.184

(6)             A study submitted by […] conducted in November 2015 regarding how many of the top twenty online retailers of the main comparison shopping services in Germany and France (in terms of visits) were registered at that time with a merchant platform. The study indicates that, in Germany, all of the top twenty online retailers of the selected comparison shopping services were listed on at least two of the four selected comparison shopping services, whereas only three of them also listed their products on at least one marketplace, i.e. Amazon or eBay (of which, only two list products on both Amazon and eBay). The analysis for France produced equivalent results: eighteen out of the top twenty online retailers of the selected comparison shopping services were listed on at least two of the three selected comparison shopping services, whereas only two of them also listed their products on at least one marketplace, i.e. Amazon or eBay (of which, only one list products on both Amazon and eBay).185

(224)    From the supply side perspective, the services provided by comparison shopping services and merchant platforms require different functionalities.

(225)    First, on the one hand, comparison shopping services collect and select the relevant information in response to each user query by analysing the real-time feeds from as many online retailers as possible186 and by providing users with such information. As noted (see recital (218)) they do not, however, sell the products directly on their websites but rather seek to refer users to third-party websites where they can buy the relevant product and, therefore, are not required to offer after-sale support, including product return functionality. On the other hand, merchant platforms manage the inventory of their retailer partners and sell the products directly on their website; these activities require specific functionalities (e.g. check-out and payment functionalities) and additional services (e.g. after-sale support)187 and are subject to specific   regulatory frameworks (e.g. concerning online payments and dispute resolution; see recital (240)).

(226)    Second, comparison shopping services and merchant platforms are generally remunerated in different ways. On the one hand, comparison shopping services are generally remunerated based on a cost-per-click model, with the retailer paying a fee for each visitor sent to its website, regardless of whether the user eventually makes a purchase. On the other hand, merchant platforms, are generally remunerated by a commission on the transaction performed on the platform or a nominal listing fee or, in case they also sell their own products, by the actual price of the products sold.188

(227)    The Commission's conclusions regarding the limited demand side and supply side substitutability between comparison shopping services and merchant platforms are not called into question by Google's claims189 that:

(a)   merchant platforms provide a product search and comparison function that is substitutable from the perspective of users and online retailers to the product search and comparison function provided by comparison shopping services.190 In support of this claim, Google relies on: (i) screenshots comparing the search results  pages  of  certain  comparison  shopping  services  (LeGuide,  Twenga, Nextag, Kelkoo, Ciao!, Idealo.fr, and prix.net) and of certain merchant platforms (Amazon, eBay, Fnac);191  (ii) user surveys conducted by Google and from third parties;192 (iii) replies from certain respondents to the Commission's requests for information;193 and (iv) a number of internal Google documents;194

(b)   the fact that merchant platforms are customers of comparison shopping  services (including Google's own comparison shopping service) is not a reason to exclude them from the relevant product market because firms can simultaneously be each other's customers and competitors;195

(c)   the fact that merchant platforms allows users to purchase products directly on their websites is not a reason to exclude merchant platforms from the relevant product market because the possibility to conclude sales represent an added value that makes merchant platforms a greater, not a lesser, competitive force vis-à-vis comparison shopping services;196 and

(d)   the Commission should have carried out a SSNIP test.197

(228)    First, the product search and comparison function of merchant platforms are not substitutable, from the perspective of users, with that of comparison shopping services, because only a limited number of the largest retailers present on comparison shopping services have chosen also to appear on merchant platforms.198 Moreover, comparison  shopping  services  generally  display  a  range  of  offers  wider      than merchant platforms, also because comparison shopping services list offers from several merchant platforms (so their catalogues comprise those of merchant platforms), whereas each merchant platform only lists offers from the retailers that have decided to work with it.199 This is confirmed by the screenshots on which Google relies in support of its claim:

(a)   The screenshots of the search results pages from LeGuide (a comparison shopping service) and Amazon (a merchant platform) indicate that  LeGuide lists offers from larger online retailers such as Conrad and merchant platforms such as Fnac and PriceMinister whereas none of these large online retailers and merchant platforms is included in the Amazon list; and

(b)   The screenshots of the search results pages of eBay and Fnac (merchant platforms)200 mainly list offers from smaller retailers whereas the screenshots  of the search results pages of Twenga, Nextag, Kelkoo, Ciao!, Idealo.fr and prix.net (comparison shopping services)201  list offers from larger retailers such as the Apple store and Darty, and from merchant platforms such as Amazon, Fnac, Priceminister and Laredoute.202

(229)    Second, none of the user surveys conducted by Google and by third parties support Google's claim that the product search and comparison function of merchant platforms and comparison shopping services are substitutable from the perspective of users.

(230)     In the first place, […]203  […]204 […].205

(231)  In the second place, the presentation by Bloomreach entitled “State of Amazon 2016”206 does not indicate that a large proportion of users in the US have recourse to the product search functionality of Amazon to compare offers from different sellers. Rather, the presentation reports that “7 in 10 [consumers] check another retailer even if they found a product they want on Amazon” (page 2).

(232) In the third place, the report “Online search: Consumer and firm behaviour – A review of the existing literature” by the Competition and Markets Authority,207 simply refers, among the “existing literature” to the findings illustrated by the presentation by Bloomreach entitled “State of Amazon 2016” (see recital (231)). Moreover,  in  the  same  report  (paragraph  4.3),  the  Competition  and       Markets Authority observes that the various search tools that consumers can use to search for products online “are by no means mutually exclusive: consumers might search for 'blue jeans' on a search engine and compare the results to the prices on their favourite brand's website. Also, a consumer might type 'high resolution camera' in Amazon's search box, read the product reviews contained there to select a model, and finally go to a price comparison website to find out which retailer is willing to offer the lowest price for that specific model”.

(233) In the fourth place, the paper “The Evolution of Product Search”, published by researchers from the University of Indiana208 is not probative as it ignored the majority of product searches started on general search engines, such as Google.com and Bing.com, and assessed only the substitutability of the product search function of merchant platforms with searches carried out directly on the standalone Google Product Search (now Google Shopping) domain.209

(234) In the fifth place, the studies by ECC Koln210 and the survey conducted by the Raymond James211 indicate only that Amazon is a website that many users visit when they want to purchase a product and says nothing about whether users have  recourse to the product search functionality of Amazon to compare offers from different sellers.

(235) Third, regarding certain replies to the Commission's requests for information, as noted above (recital (220)(3)), a majority of the comparison shopping services and merchant platforms that replied to the Commission's requests for information indicated that they do not consider themselves as competitors, but rather as business partners. This is further confirmed by certain of the replies on which Google relies in support of its claim:

(a)   […] (a comparison shopping service): […]212 […]213

(b)   […] (a comparison shopping service): Google quotes a statement by the […] referring to Amazon as a “large competitor” of comparison shopping services. Google refers to an interview in which […] was asked to provide his views on the future of comparison shopping services in five years214; it is in that context that […] mentioned, in addition to Google, also Amazon and Apple among the current or potential competitors of comparison shopping services.215 However, in its replies to the Commission's requests for information […] indicated that comparison shopping services and merchant platforms provide different services to users and retailers.216

(c)   […] (a merchant platform): Google quotes the following statement from […] reply to […] the Commission's request for information […]: “[…] competes with comparison shopping services”.217 However, in the same reply, […] clarified that it competes with comparison shopping services only “to the extent that both services try to promote third party products and generate revenue by commissioning sales and providing additional services for extra costs. However, […] rather competes with other online marketplaces offering similar services […]. Comparison shopping services are also used by […] to lead potential customers to […] marketplace (see also questions 14 and 15)”.218

(d)   […] (a merchant platform): Google quotes the following statement from […] reply to […] the Commission's request for information […]: “in […]'s view, comparison shopping services in the EEA such as those offered by Google would by their nature compete with in-platform search functionality available on […] of other online platforms”.219 However, […] also clarified in the same response that “due to the relatively small volume of business that […] carries out in the EEA, […] has a limited understanding of online markets in this region, and is not yet in a position to take a view regarding the level of competition  between  merchant  platform  services  and  comparison shopping services in the EEA”.220;

(e)   […] (a comparison shopping service): Google quotes the following statement from […] reply to […] the Commission's request for information […]: “Merchant platforms are competition to our shopping services”.221  However,  in the same reply, […] also explained that “we work with [merchant platforms] extensively as they are attractive partners for us” and that comparison  shopping services and merchant platforms list different online retailers: “Merchant platforms are still limited in the catalogue size that they are able to offer - they can only have products from retailers who have opted to work with them, which means the other 99% of retailers aren’t on their platform. For us, it means that when we can aggregate their catalogues with other merchant platforms and merchant feeds, we can offer a wider choice of products to our users and make sure that we maintain a whole of market view. As outlined in our answer to Question 4, our largest Merchants (affiliates) are Merchant Platforms”.222

(f)   […] (a comparison shopping service): Google quotes the following statement from […] reply to […] the Commission's request for information […]: “[…] is operating price comparison websites for searching products similar to online market places. Thereby the user refers to a broad variety of products, extensive search queries and many filter possibilities similar to online market places”.223 However, in the same reply, […] indicated that it competes with online  retailers and merchant platforms only to the extent that they might bid for the same AdWords' keywords224 and that “the advantage of price comparison websites is that both products of online marketplaces and products of smaller merchants are listed”.225

(g)   […] (which operates the merchant platform […]): Google quotes the following statement from […] reply to […] the Commission's request for information […]: “A user can search and find products on both [merchant platforms and comparison shopping services”.226 However, in the same reply, […] indicated that “No […] products compete with Google” and that “No […] products compete with comparison shopping services”.227

(236)     Fourth, […]228  […]229 […]230

(237)    Fifth, none of the past merger decisions of the Commission and the UK Office of  Fair Trading (“OFT”) relied on by Google supports its claim that merchant platforms and comparison shopping services are part of the same relevant product market. Rather, the customer relationship between comparison shopping services and merchant platforms shows that they are in different markets:

(1)      in   its   decision   in   case   COMP/M.6163   Axa/Permira/Opodo/Go Voyages /eDreams, the Commission found that online travel agents (“OTAs”) and  airline websites compete in a possible market for online distribution of flight tickets because they both sell tickets to end users. This is consistent with the finding that merchant platforms compete with online retailers, not with comparison shopping services. Indeed OTAs can be compared to merchant platforms, whereas airline websites can be compared to online retailers and the Commission also noted that both OTAs and airline website  promote  their offers through price comparison services dedicated to flights, which display offers from OTAs and airline websites “indistinguishably”;

(2)      in its Hotel online booking decision,231 the OFT found that price comparison services are advertising tools, not competitors, for OTAs (recital 1.9);

(3)      in its Priceline/Kayak decision,232 the OFT “did not need to conclude on the precise market definition” (recital 15) and analysed competition between OTAs and travel search services only “on a cautious basis” (recital 45). In this context, the OFT found that, even in a potential market for travel search services including both OTAs and travel search services, the parties, i.e. an OTA and a travel search service “are not in fact close competitors” (recital 66);

(4)    all the other Commission merger decisions quoted by Google233 concern manufacturers and intermediaries (such as wholesalers, retailers, spare parts distributors) that compete in the sale of the same products. This is different from the relationship between comparison shopping services and merchant platforms because, unlike merchant platforms, comparison shopping services do not sell products to end users; instead they provide information to users and traffic to merchant platforms and online retailers.

(238)    Sixth, […]234 as noted in recitals (220)(3) and (220)(4) comparison shopping services are an important source of traffic for merchant platforms and merchant platforms are among the main customers of comparison shopping services, including of Google's comparison shopping service.

(239)    Seventh, the introduction of direct purchase functionality by certain of comparison shopping services does not indicate that merchant platforms and comparison  shopping services are part of the same relevant product market.

(240)    In the first place, as noted in recitals (218)-(225) a direct purchase functionality distinguishes merchant platforms from comparison shopping services from the perspective of users and retailers:

(1)     a direct purchase functionality can changes the business model and nature of the service provided by comparison shopping services to users and retailers to such an extent that the service may no longer be considered to constitute a comparison shopping service, especially if the direct purchase functionality introduced systematically for all (or the majority of) merchants and offers;235

(2)     a direct purchase functionality changes the regulatory framework applicable to comparison shopping services. For instance, a direct purchase functionality triggers the application of the regulatory framework concerning online payments and the EU Regulation on online dispute resolution for consumer disputes;236

(3)    a direct purchase functionality has an impact on the relationship of comparison shopping service with customers. For example, as noted in recital […];237

(4)    a direct purchase functionality requires significant and time-consuming investments. […]238 […]239

(241)    In the second place, only a limited number of comparison shopping services have introduced such a function. Of the 361 competing comparison shopping services in the EEA identified by Google in the SO Response (the “361 SO Response Aggregators”),240 only seven have introduced a direct purchase functionality on their websites (i.e. “Buy button” of “Shopping cart”).241 Moreover, those seven services have introduced such a functionality for only a limited number of merchants and offers.242  For example, Idealo, the largest comparison shopping service in   Germany after Google Shopping, had managed in 2015 to convince less than [<20%] of its online retailers to introduce a “Shopping cart” option.243

(242)    Eighth, the Commission was not required to carry out a SSNIP test.

(243)    In the first place, the SSNIP test is not the only method available to the Commission when defining the relevant market.244

(244)    In the second place, the Commission is required to make an overall assessment of all the evidence and there is no hierarchy between the types of evidence that the Commission can rely upon.245

(245)    In the third place, the SSNIP test would not have been appropriate in the present case because Google provides its search services for free to users.246

(246)    The Commission thus concludes that comparison shopping services constitute a distinct relevant product market, which does not include merchant platforms. Google claims247 that the relevant product market comprises both comparison shopping services and merchant platforms. The consequences of following such an alternative relevant product market definition are examined in section 7.3.2.

5.2.2.5. Comparison shopping services versus offline comparison shopping tools

(247)    There is also limited substitutability between comparison shopping services and offline shopping comparison tools such as print catalogues, consumer magazines, shopping TV channels and programmes.

(248)    From the demand side perspective, an offline comparison shopping tool cannot provide users with the same amount of information, level of reactivity and service as a comparison shopping service. The latter are typically able to aggregate millions of offers from thousands of online retailers and to provide instantaneous, customised, updated answers and other information such as user reviews to real-time requests.248

(249)    Moreover, comparison shopping services provide users with links that lead (directly or via one or several successive intermediary pages) to the websites of retailers where, if they wish to, users can further explore an offer and make a purchase. By contrast, if users want to continue their shopping experience based on the offers presented by offline comparison shopping tools, users must perform different  actions, such as visiting a brick-and-mortar shop, calling a dedicated phone number, or retrieving and manually recording the address of the website of the relevant retailer.

(250)    From the supply side perspective, comparison shopping services differ from offline comparison shopping tools in terms of functionalities and business model.249 As noted  in  recital  (195),  the  provision  of  comparison  shopping  services requires developing and maintaining a dedicated data  infrastructure and structured relationships with online retailers.

5.3. The relevant geographic markets

(251)    The Commission concludes that the relevant geographic markets for general search services (section 5.3.1) and comparison shopping services (section 5.3.2) are all national in scope.

5.3.1. The national markets for general search services

(252)    For the reasons set out in recitals (253) to (255), the Commission concludes that the markets for general search services are national in scope.

(253)    First, even though general search services can be accessed by users anywhere in the world, the main general search services offer localised sites in different countries and in a variety of language versions. For instance, Google has national sites for each EEA country and in nearly every official language of the Union. Moreover, […].250

(254)    Second, there are barriers to the extension of search technology beyond national and linguistic borders.251 These barriers are one of the reasons why certain  smaller general search services in the EEA use their own search technology mainly for websites from their own country and in their own native language, while returning the search results of Google or Bing for their websites in other countries or different languages.252 Moreover, even for large, multinational companies, the costs associated with upsizing search technology to cover sites in other countries and in different languages can be prohibitive.253

(255)    Third, Google does not contest that the relevant geographic markets for general search services are national in scope.

5.3.2. The national markets for comparison shopping services

(256)    For the reasons set out in recitals (257)-(263), the Commission concludes that the markets for comparison shopping services are national in scope.

(257)    First, even though comparison shopping services can be accessed by users anywhere in the world, the main comparison shopping services offer localised sites in different countries and in a variety of language versions. For instance, Google has national comparison shopping services for each EEA country and in nearly every official language of the Union.

(258)    Second, language is a particularly important aspect of comparison shopping services. This is for a number of reasons.

(259)    In the first place, comparison shopping services typically provide language-specific results, including links to the relevant national websites of the relevant online retailers, and design their services and websites on a country-by-country basis to address their target audiences in their native language.

(260)    In the second place, the provision of comparison shopping services requires a commercial workforce whose role is to enter into agreements with online retailers, most of which are active at a national level or run their business on a country-by- country basis.

(261)    In the third place, a product that sells well in one country may not necessarily have the same success in other countries.

(262)    In the fourth place, a product that may be promoted or sold without limitation in one country may be subject to promotion and/or marketing restrictions in another country.

(263)    Third, Google does not contest that the relevant geographic markets for comparison shopping services are national in scope.

6. DOMINANT POSITION

6.1. Principles

(264)    The dominant position referred to in Article 102 of the Treaty relates to a position of economic strength enjoyed by an undertaking which enables it to prevent effective competition being maintained on the relevant market by affording it the power to behave to an appreciable extent independently of its competitors, its customers and ultimately of its consumers.254

(265)    The existence of a dominant position derives in general from a combination of several factors which, taken separately, are not necessarily determinative.255

(266)    One important factor is the existence of very large market shares, which are in themselves, save in exceptional circumstances, evidence of the existence of a dominant position.256 An undertaking which holds a very large market share for some time, without smaller competitors being able to meet rapidly the demand from those who would like to break away from that undertaking, is by virtue of that share in a

position of strength which makes it an unavoidable trading partner and  which, already because of this, secures for it, at the very least during relatively long periods, that freedom of action which is the special feature of a dominant position.257 That is the case where a company has a market share of 50% or above.258 Likewise, a share of  between  70%  and  80%  is,  in  itself,  a  clear  indication  of  the  existence of a dominant position in a relevant market.259  The ratio between the market share held  by the dominant undertaking and that of its nearest rivals is also a highly significant indicator.260

(267)    In fast-growing sectors characterised by short innovation cycles, large market shares may sometimes turn out to be ephemeral and not necessarily indicative of a dominant position.261 However, this fact cannot preclude application of the competition rules,  in particular Article 102 of the Treaty, especially if a fast-growing market does not show signs of marked instability during the period at issue and, on the contrary, a rather stable hierarchy is established.262

(268)    The fact that a service is offered free of charge is also a relevant factor to take into account in assessing dominance. In so far as users expect to receive a service free of charge, an undertaking that decides to stop innovating may run the risk of reducing  its attractiveness, depending on the level of innovation on the market in question. In this respect, another relevant factor is whether there are technical or economic constraints that might prevent users from switching providers.263

(269)    Another important factor when assessing dominance is the existence of barriers to entry or expansion, preventing either potential competitors from having access to the market or actual ones from expanding their activities on the market.264

(270)    Such barriers may result from a number of factors, including exceptionally large capital investments that competitors would have to match, network externalities that would entail additional cost for attracting new customers, economies of scale from which newcomers to the market cannot derive any immediate benefit and the actual costs of entry incurred in penetrating the market.265 Switching costs are therefore  only one possible type of barrier to entry and expansion.

6.2. Google's dominant position in the national markets for general search services

(271)    The Commission concludes that Google holds a dominant position in each national market for general search services since 2008, apart from in the Czech Republic, where Google holds a dominant position since 2011.

(272)    This conclusion is based on Google’s market shares (section 6.2.1), the existence of barriers to expansion and entry (section 6.2.2), the infrequency of user multi-homing and the existence of brand effects (section 6.2.3) and the lack of countervailing buyer power (section 6.2.4). The conclusion holds notwithstanding the fact that general search services are offered free of charge (section    6.2.5), and regardless of whether general search on static mobile devices constitutes a distinct market from general search on mobile devices (section 6.2.6).

6.2.1. Market shares

(273)    General search services have been offered in the EEA since the middle of the 1990s.266 Google has been active in each national market since September 1998.267

(274)    The Commission concludes that Google has enjoyed strong and stable market shares by volume across the EEA since 2008, and there has been no effective entry in any EEA country during that period. Contrary to Google's claim268, this provides a good indication of Google’s competitive strength in the national markets for general search services.

(275)    The Commission has used market shares by volume as a proxy for four reasons.  First, market shares by value cannot be computed because general search services are provided free of charge to the user. Second, despite its best efforts, the Commission has been unable to obtain precise and verifiable values regarding the Revenue Per

Search (“RPS”) of the main general search services. Third, advertisers look at usage shares when deciding where to place their search advertisements.269 Fourth, […].270

(276)    There are several methods to calculate market shares by volume in the EEA.271 All the methods indicate that since 2008, Google has enjoyed high market shares in all the relevant general search markets across the EEA, except for the Czech Republic.

(277)    2010 data by Nielsen (based on page views) indicates that Google’s share of the national markets for general search services in 2010 was 84.6% in France, 85.3% in Germany, 85.9% in Italy, 91.3% in Spain and 81.3% in the United Kingdom. None  of Google’s competitors had a market share exceeding 4.1% in any of these five countries.272

(278)    November 2014 data by AT Internet (based on site visits) similarly indicates that Google’s share of the national markets for general search services in November 2014 was 93.5% in France, 94.1% in Germany, 96.6% in Spain and 92.6% in the United Kingdom.273 None of Google’s competitors had a market share exceeding 3.9% in any of these four countries.

(279)    AT Internet data does not cover EEA countries other than France, Germany, Spain and the United Kingdom. The Commission has therefore also looked at data from StatCounter, which covers all EEA countries, for the period 2008-2016.274

(283)    Apart from the Czech Republic and Slovenia, Google’s market shares in all EEA countries since 2008 have thus always been above 85%.

(284)    In the Czech Republic and Slovenia, Google's market share has been above 70% since 2011.282

6.2.2. Barriers to entry and expansion

(285)    The Commission concludes that the national markets for general search services are characterised by the existence of a number of barriers to entry and expansion.

(286)    First, the establishment of a fully-fledged general search engine requires significant investments in terms of time and resources. For example, each year since at least 2009, Microsoft has invested [a significant amount] in R&D and capital expenditure in the development and maintenance of the latest version of its general search engine launched in June 2009 under the brand name “Bing”.283 Other companies indicate  that the costs associated with the establishment of a fully-fledged general search engine constitute a barrier to entry. For example, […] argues that it only operates its own general search technology for […] language websites because “investments are too large to develop such technology for […] language websites”.284 […] says that “the incremental costs of converting […] into a viable, competitive broad search service would […] require years of development”.285 As for […], a company that already has substantial server capacities in place […], it says: “Investment in equipment and personnel would likely be the primary costs, and they would be very high. In addition, obtaining the large quantity of data necessary to develop an effective [general] search engine (e.g., the information upon which relevancy algorithms can be built and improved) would be a significant barrier to entry”.286

(287)    Second, because a general search service uses search data to refine the relevance of  its general search results pages, it needs to receive a certain volume of queries in order to compete viably. The greater the number of queries a general search service receives, the quicker it is able to detect a change in user behaviour patterns and update and improve its relevance. This is supported by internal Google documents287 and by evidence from a number of other general search services.288

(288)    A general search service also needs to receive a certain volume of queries in order to improve the relevance of its results for uncommon (“tail”) queries. Tail queries are important because users evaluate the relevance of a general search service on a holistic basis and expect to obtain relevant results for both common (“head”) and uncommon tail queries.289  The greater the volume of data a general search service possesses for rare tail queries, the more users will perceive it as providing more relevant results for all types of queries.290

(289)    In that regard, there may be diminishing returns to scale in terms of improvements in relevance once the volume of queries a general search service receives exceeds a certain volume.291 It may also be that the lower success and relevance of a general search service can be explained by other reasons, such as the fact that it does not localise its search results in different countries, that its web index is more limited in depth, or that it is slower in updating its index in order to deliver fresh content to users.292 Regardless of the veracity of such arguments, however, they remain of limited relevance for the assessment of barriers to entry and expansion on  the national markets for general search services because of the underlying fact that a general search service has to receive at least a certain minimum volume of queries in order to compete viably.

(290)    The relevance of scale is also not called into question by the fact that in the late 1990s, Google was able to overtake the former market leaders, AltaVista and Lycos. At that time, scale was less of a critical factor because the indexing technology of general search engines was not yet able to assess user behaviour.293

(291)    Third, general search services constantly invest to improve their product and a new entrant would have no choice but to attempt to match these investments. The following chart shows worldwide capital investments made by Google294 and Yahoo295 in their general search services between 2006 and 2015.

revenue which can be reinvested in the maintenance and improvement of the general search service so as to attract more users (see also recital (293)).300 Google generates substantial revenues from its online search advertisement business. In the years 2013-2016 its advertising revenues rose from USD 50.6 billion to USD 79.4 billion.301

(297)  The existence of barriers to entry and expansion is supported by a number of additional factors.

(298)  First, since 2007, a number of companies have exited the national markets  for general search services, either completely or by abandoning their general search technology in favour of third party technology. For example, Yahoo abandoned its general search technology, including in the EEA, in 2009 and now relies on Bing’s general search technology to power its portal.302 Equally, Ask.com abandoned its general search technology, including in the EEA, in November 2010 and started  using Google’s general search technology to power its portal.303

(299)  Second, a number of smaller players still present on the national markets for general search services have been unable to expand and are contemplating interrupting their general search services in the EEA in the near future. This is the case of […], one of the first general search services in […], which currently offers general search services restricted to […] language websites.304

(300)  Third, since 2007, there has been only one significant entrant in the national markets for general search services, Microsoft, which launched the latest version of  its general search service, Bing, in 2009.305 Since 2009, however, Bing’s market shares have never exceeded 10% in any EEA country.306 […]307

(301) A number of start-ups have also attempted to launch competing general search services since 2007. None of these companies have, however, been able to establish a significant market presence. Many have either stopped providing general search services or chosen instead to provide complementary types of services that do not compete with Google’s general search service in the EEA.

(302) In June 2008, Kosmix, at the time a search service specialised in health matters, was reported to have attempted to start providing a general search service.308 Soon after, however, it changed its business model and became a platform integrating a number of specialised search services.309

(303)  In July 2008, former employees of Google launched a general search service called Cuil. At launch, Cuil claimed that its index of the web was deeper than Google’s310 and certain industry observers considered that Cuil had the potential to compete against Google.311 Within a few weeks of launch, however, Cuil’s share of a hypothetical global general search services market decreased from 0.11% to 0.01%,312 and on 17 September 2010, Cuil went offline.313

(304) In April 2010, another general search service called DuckDuckGo, offered only in English, was launched. Its traffic remains, however marginal. In December 2014, it processed only 221 million searches.314 This represented less than 0.8% of all general searches performed in the US in December 2014. Because of its US origin and the fact that it is only offered in English, it is likely that DuckDuckGo's shares of the national markets for general search services are even lower. In addition, DuckDuckGo is dependent on third party technology for its general search services. As DuckDuckGo explains on its website: “While our indexes are getting bigger, we do not expect to be wholly independent from third-parties. Bing and Google each spend hundreds of millions of dollars a year crawling and indexing the deep Web. It costs so much that even big companies like Yahoo and Ask are giving up general crawling and indexing. Therefore, it seems silly to compete on crawling and, besides, we do not have the money to do so”.315

(305) In October 2010, a general search service called Blekko, offered only in English, was launched. Blekko’s distinctive feature was that it allowed users to mark the sites they visit with special attributes (“slashtags”) which other users could use to focus their queries. Its traffic, however, remained marginal. In June 2013, it reported an average of only 5 million searches per day.316 This represented less than 0.8% of all general searches performed in the US in June 2013.317 Because of its US origin and the fact that it was only offered in English, it was likely that Blekko's shares of the national markets for general search services were even lower. In addition, because it relied on human input to generate and curate slashtags, Blekko could not generate large-scale real time search results comparable to entirely automated general search engines such as Google or Bing. This is supported by an internal Google document.318 Blekko stopped providing general search services at the end of March 2015.319

6.2.3. The infrequency of user multi-homing and the existence of brand effects

(306) Contrary to what Google claims,320 the Commission concludes that notwithstanding the technical ability of users to switch between different general search services,321 only a minority of users in the EEA that use Google’s general search service as their main general search service use other general search services (a behaviour known as “multi-homing”).

(307) First, a survey commissioned by […] quantified how often Google users in five EEA countries (France, Germany, Italy, Spain and the United Kingdom) use other general search services.322 The survey was performed between December 2010 and April 2011 and defined as a multi-homer a user that conducts at least 5% of all its queries on at least two distinct general search services. Based on that definition, the survey found that only 12% of users in Germany, Italy and Spain multi-home. As for France and the United Kingdom, the percentage of users that multi-home was respectively 15% and 21%.

(308) The survey also found that users that use Google as their primary general search service in these five EEA markets are significantly less likely to multi-home than users that use Bing or Yahoo as their primary general search service in  those markets. The chart below summarises the proportion of users that multi-home, depending on their primary general search service.

confirm that a large number of users outside the EEA (76% in case of the Performics and ROI Research study) multi-home at best only occasionally.

–    A study of 2010 referred to by […]328 confirms that more than two thirds of users did not use general search services other than Google (“nearly a third of [beta] users were aware of, and used, alternative web services made available by default”);

–    An eBay blogpost of 2013329 states only that users generally access ten  websites during a purchase process, without specifying whether one or more of those sites are general search services (“consumers can access many more options, with the average shopper visiting 10.4 websites during their purchasing process”).

(312)    Fourth, because of the strength of the Google330 brand,331 users trust in the relevance of search results provided by Google. Consequently, as supported by internal Google documents332 and two third party studies,333 a significant number of users  are  unlikely to multi-home even if Google were to degrade the quality of its general search service.

(313)    Fifth, contrary to what Google claims,334 the Commission is not required to ascertain the number of users that would need to multi-home in order to render a possible degradation by Google in the quality of its general search service unprofitable.335 There is no legal requirement on the Commission to quantify how much a dominant undertaking would have to alter its product for it to no longer to be dominant.

(314)    Sixth, even if a significant number of users of Google’s general search service were to multi-home, switching costs are in any event only one possible type of barrier to entry and expansion. Barriers to entry and expansion can also derive for instance from the need for large investments or network effects and such barriers exist in this case (see recitals (285) to (305)).

(315)    Seventh, even if users were to prefer Google's general search service because of the perceived relevance of the search results that it returns, this would not preclude a finding of dominance. The reasons or the causes for which Google has a dominant position on the national markets for general search services are irrelevant.336 Google does not contest this.

6.2.4. Lack of countervailing buyer power

(316)    The Commission concludes that the users are unable to exert any meaningful countervailing buyer power vis-à-vis Google.

(317)    This is because the national markets for general search services are characterised by the fact that each user only represents a tiny fraction of the volume of total search queries.

(318)    Google does not contest this.

6.2.5.Google's argument that Google’s general search services are offered free of charge and the Commission's response

(319)    The Commission concludes that a finding of dominance is not precluded by Google's claim337 that it offers its general search services free of charge.

(320)    First, that claim is misleading. While users do not pay a monetary consideration for the use of general search services, they contribute to the monetisation of the service by providing data with each query.

(321)    Second, and in any event, the free nature of a service is only one “relevant factor in assessing [...] market power”.338 Other equally, if not more, relevant factors in this case include the following.

(322)    In the first place, Google has enjoyed strong and stable market shares by volume across the EEA since 2008, and there has been no effective entry in any EEA country during that period (see section 6.2.1).

(323)    In the second place, the national markets for general search services are characterised by the existence of a number of barriers to entry and expansion (see section 6.2.2).

(324)    In the third place, because of the infrequency of user multi-homing and the existence of brand effects (see section 6.2.3), Google could alter the quality of its general search service to a certain degree without running the risk that a substantial fraction of its users would switch to alternative general search engines.

6.2.6. Static devices versus mobile devices

(325)    The Commission's conclusion on dominance would hold even if general search services on static devices were to constitute a distinct market from general search services on mobile devices.

(326)    First, Google’s share of general searches on mobile devices was above 95% in 2016 in all EEA countries, except the United Kingdom and the Czech Republic, where it was 94.9% and 84.4% respectively.

(327)    Second, Google’s share of general searches on mobile devices has been above 90% since 2009 in almost all EEA countries (with the exception of the Czech Republic  and Slovenia where the lowest annual average was respectively 80.2% and 74.4%). The chart below shows for each country the share of general searches on mobile devices performed on Google in 2016 as well as the lowest yearly share of general searches  on  mobile  devices  performed  on  Google  since  2009,  as  reported by StatCounter. Before 2009, general search on mobile devices was marginal. It represented less than 1% of all general searches.339 […]340

the internal market.343 A system of undistorted competition can be guaranteed only if equality of opportunity is secured as between the various economic operators.344 The scope of the special responsibility of the dominant undertaking has to be considered in light of the specific circumstances of the case.345

(332)   Article 102 of the Treaty and Article 54 of the EEA Agreement prohibit abusive practices which may cause damage to consumers directly, but also those which harm them indirectly through their impact on an effective competition structure.346

(333)   The concept of abuse is an objective concept relating to the behaviour of an undertaking in a dominant position which is such as to influence the structure of a market where, as a result of the very presence of the undertaking in question, the degree of competition is weakened and which, through recourse to methods different from those which condition normal competition on the merits, has the effect of hindering the maintenance of the degree of competition still existing in the market or the growth of that competition.347

(334)   Article 102 of the Treaty and Article 54 of the EEA Agreement prohibit not only practices by an undertaking in a dominant position which tend to strengthen that position,348 but also the conduct of an undertaking with a dominant position in a  given market that tends to extend that position to a neighbouring but separate market by  distorting  competition.349   Therefore,  the  fact  that  a  dominant    undertaking’s abusive conduct has its adverse effects on a market distinct from the dominated one does not preclude the application of Article 102 of the Treaty or Article 54 of the EEA Agreement.350 It is not necessary that the dominance, the abuse and the effects  of the abuse are all in the same market.

(335)   Article 102 of the Treaty and Article 54 of the EEA Agreement list a number of abusive practices. These are merely examples, not an exhaustive enumeration of the practices that may constitute abuses of dominant position prohibited by the Treaty or the EEA Agreement.351 For example, Article 102 of the Treaty and Article 54 of the EEA Agreement generally prohibit any abusive practice capable of limiting markets.352 The legal characterisation of an abusive practice does not depend on the name given to it, but on the substantive criteria used in that regard.353 The specific conditions to be met in order to establish the abusive nature of one form of conduct covered by Article 102 of the Treaty and Article 54 of the EEA Agreement must not necessarily also apply when assessing the abusive nature of another form of conduct covered by those articles.354

(336)   In order to determine whether the undertaking in a dominant position has abused  such a position, it is necessary to consider all the circumstances and to investigate whether the practice tends, for example, to bar competitors from access to  the market, to apply dissimilar conditions to equivalent transactions with other trading parties, thereby placing them at a competitive disadvantage, or to strengthen the dominant position by distorting competition.355

(337)   It follows from the nature of the obligations imposed by Article 102 of the Treaty  and Article 54 of the EEA Agreement that, in specific circumstances, an undertaking in a dominant position must refrain from adopting a course of conduct which would be unobjectionable if adopted by non-dominant undertakings.356

(338)   Similarly, the Court of Justice held that an abuse of a dominant position is prohibited under Article 102 of the Treaty and Article 54 of the EEA Agreement “regardless of the means and procedure by which it is achieved”, and “irrespective of any fault”.357 In the same vein, the Commission is under no obligation to establish the existence of an abusive intent on the part of the dominant undertaking in order to render Article 102 of the Treaty and Article 54 of the EEA Agreement applicable. While intent is not a necessary prerequisite to show an abuse, it is, however, one of the criteria which can be used for assessing the abusive nature of behaviour under Article 102 of the Treaty.358

(339)   Concerning the effects of the dominant undertaking's conduct, Article 102 of the Treaty and Article 54 of the EEA Agreement prohibit behaviour that tends to restrict competition or is capable of having that effect,359 regardless of its success.360 This occurs not only where access to the market is made impossible for competitors, but also where the conduct of the dominant undertaking is capable of making that access more difficult, thus causing interference with the structure of competition on the market.361 Customers and users should have the opportunity to benefit from whatever degree of competition is possible on the market and competitors should be able to compete on the merits for the entire market and not just for a part of it.362 An undertaking in a dominant position may not therefore justify abusive conduct in a certain segment of a market by the fact that its competitors remain free to compete in other segments.363

(340)   It is open to a dominant undertaking to provide a justification for conduct that is liable to be caught by the prohibition under Article 102 of the Treaty and Article 54 of the EEA Agreement. It may demonstrate, for that purpose, either that its conduct  is objectively necessary, or that the exclusionary effect produced may be counterbalanced, outweighed even, by advantages in terms of efficiency gains that also benefit consumers.364

7.2. The abusive conduct: the more favourable positioning and display, in Google’s general search results pages, of Google's own comparison shopping service compared to competing comparison shopping services

(341)   The Commission concludes that the Conduct constitutes an abuse of Google's dominant position in each of the thirteen national markets for general search services where Google either launched the Product Universal or, if the Product Universal was never launched in that market, the Shopping Unit. The Conduct is abusive because it constitutes a practice falling outside the scope of competition on the merits as it: (i) diverts traffic in the sense that it decreases traffic from Google's general search results pages to competing comparison shopping services and increases traffic from Google's general search results pages to Google's own comparison shopping service; and (ii) is capable of having, or likely to have, anti-competitive effects in the national markets for comparison shopping services and general search services.

(342)   To demonstrate why the Conduct is abusive and falls outside the scope of competition on the merits, the Commission first describes how Google positions and displays more favourably, in its general search results pages, its own comparison shopping service compared to competing comparison shopping services (section 7.2.1). Second, it illustrates the importance of traffic for comparison shopping services (section 7.2.2) and how the Conduct diverts traffic in the sense that it decreases traffic from Google's general results pages to competing comparison shopping services and increases traffic from Google's general search results pages to Google's own comparison shopping service (section 7.2.3). Third, it shows that the generic search traffic from Google's general search results pages accounts for a large proportion of traffic for competing comparison shopping services and cannot be effectively replaced by other sources of traffic currently available to competing comparison shopping services (section 7.2.4). Fourth, it explains how the Conduct is capable of extending Google’s dominant position in the national markets for general search services to the national markets for comparison shopping services (section 7.3.1), and of protecting Google's dominant position in the national markets for general search services (section 7.3.3). It also explains that, even if the alternative product market definition proposed by Google, comprising both comparison  shopping services and merchant platforms, were to be followed, the Conduct would be capable of having, or likely to have, anti-competitive effects in at least the comparison shopping services segments of the possible national markets comprising both comparison shopping services and merchant platforms (section 7.3.2). Finally, it rejects Google's arguments with regard to the applicable legal test (section 7.4) and  to potential objective justifications for the Conduct (section 7.5).

(343)   In summary, Google has artificially reaped the benefits of the Conduct. Google did not invent comparison shopping. Google's first comparison shopping service, Froogle, was not gaining traffic as it did not appear visibly in Google's general search results pages. It was only after Google started the Conduct in each of the thirteen national markets for general search services that traffic to Google's comparison shopping service from Google's general search results pages began to increase on a lasting basis whereas traffic to almost all competing comparison shopping services began to decrease on a lasting basis.

7.2.1. Google positions and displays, in its general search results  pages,  its  own  comparison shopping service more favourably compared to competing comparison shopping services

(344)   The Commission concludes that notwithstanding Google's arguments to the contrary (section 7.2.1.3), Google positions and displays, in its general search results pages,  its own comparison shopping service (section 7.2.1.2) more favourably compared to competing comparison shopping services (section 7.2.1.1). While competing comparison shopping services can appear only as generic search results and are  prone to the ranking of their web pages in generic search results on Google's general search results pages being reduced (“demoted”) by certain algorithms, Google's own comparison shopping service is prominently positioned, displayed in rich format and is never demoted by those algorithms.

7.2.1.1. The way competing comparison shopping services are positioned and displayed in Google’s general search results pages

7.2.1.1.1. The way competing comparison shopping services are positioned in Google's general search results pages

(345)   In response to a user query in Google’s general search engine, Google uses generic search algorithms to rank web pages, including those of competing comparison shopping services. These algorithms include the PageRank algorithm (see recital (16)).365 Google also applies a variety of adjustment mechanisms to the results of the PageRank algorithm “to improve the user experience”.366

(346)   Google has issued Webmaster Guidelines that outline the “core concepts for maintaining a Google-friendly site”.367 Websites that do not comply with Google’s Webmaster Guidelines are prone to being demoted or even removed altogether from Google’s index, in which case a website’s pages will no longer appear at all in generic search results on Google's general search results pages.368

(347)   Google’s Webmaster Guidelines warn in particular against hidden text or hidden links, automated queries to Google, pages loaded with irrelevant keywords, duplicate content, malware (e.g. viruses), and pages with little or no original content.369 They also note that “Google may respond negatively to other misleading practices not listed here”.370

(348)   Google uses dedicated algorithms to identify and demote automatically websites that do not comply with its Webmaster Guidelines.371 In a minority of cases, Google employees that are part of its “Webspam Team” and “Bad Urls Team” also identify and apply manually demotions to websites that do not comply with the Webmaster Guidelines.372

(349)   Competing comparison shopping services in the EEA are prone to being demoted by […] dedicated algorithms: […] and “Panda” (which, since May 2014, is known as “Panda 4.0” […]).

(350)     […],373 […].374 […]375 […]  376

(351)     […]377

[…] 378  […] 379 […]380

[…]

(352)   Comparison shopping services are prone to being demoted by the […] algorithm due  to the characteristics of those services.

(353)     First, […] 381 […].382

(354)     Second, […].

(355)     Third, […] 383 […].

(356) The Panda algorithm was introduced in the United States in February 2011. It was subsequently extended to all English language queries worldwide on 11 April 2011 and to all general search queries across the EEA on 12 August 2011.384 Regular updates of the Panda  algorithm  have  also  subsequently  been  rolled  out.  As  of 20 May 2014, 27 updates of the Panda algorithm had been implemented.385 […].386

(357)     […] 387 […].388

(358)               Like in the case of […], set out in recitals (353)-(355), comparison shopping services are prone to being demoted by the Panda algorithm due to the characteristics inherent to those services. This is confirmed by the following Google documents:

(a)   A Google blogpost of 21 January 2011 regarding the introduction of the Panda algorithm, which stated that: “we’re evaluating multiple changes that should help drive spam levels even lower, including one change that primarily affects sites that copy others’ content and sites with low levels of original content”.389

(b)   A blogpost by Matt Cutts of 28 January 2011 regarding the introduction of the Panda algorithm which stated that: “the change was approved at our weekly quality launch meeting last Thursday and launched earlier this week. […] The net effect is that searchers are more likely to see the sites that wrote the original content rather than a site that scraped or copied the original site’s content”.390

(c)   A Google blogpost of 24 February 2011 (updated in April 2011) in relation to the introduction of the Panda algorithm for all English language queries, which stated that: “This update is designed to reduce rankings for low-quality sites— sites which are low-value add for users, copy content from other websites or sites that are just not very useful. At the same time, it will provide better rankings for high-quality sites—sites with original content and information such as research, in-depth reports, thoughtful analysis and so on”.391

(d)   A Google document of 6 May 2011 entitled “More guidance on building high- quality sites”, which stated that one of the elements that Google takes into account to assess the quality of a website is: “original content or information, original reporting, original research, or original analysis; whether the page provides substantial value when compared to other pages in search results; whether the content mass-produced by or outsourced to a large number of creators, or spread across a large network of sites, so that individual pages or sites don’t get as much attention or care”.392

(359)               Contrary to Google's claim that the Commission overstates the impact of the Panda algorithm on competing comparison shopping services and their positioning in Google's general search results pages393, the fact that competing comparison  shopping services are prone to being demoted in Google's general search results pages by the […] and Panda algorithms is further supported by evidence on (i) the number of comparison shopping services to which those algorithms have been applied; (ii) the impact of the application of the Panda algorithm on the visibility394 of competing comparison shopping services; (iii) the impact of the application of the Panda algorithm on the trigger rates395 of competing comparison shopping services; and (iv) the impact of the application of the Panda algorithm on the average ranking of competing comparison shopping services.

(360)               First, since their rollout, the […] and Panda algorithms have been applied […] to over […] of the 361 SO Response Aggregators. In particular, the […] algorithm have been applied to […] of the […] of the 361 SO Response Aggregators ([…]), the Panda algorithm396  to […], and […] algorithm to […].397

(361)               Second, the weekly Visibility Index398 calculated by the independent market monitoring company Sistrix in the United Kingdom, Germany, France, Italy and Spain, as well as analyses carried out by the Commission based on data from Sistrix, indicates that during the period between 2 August 2010 and 2 December 2016, for the vast majority of the most important comparison shopping services in terms of traffic, their visibility in Google's general search results pages (i) was at its highest at the end of 2010 and beginning of 2011, (ii) was followed by a sudden drop after the launch of the Panda algorithm in the respective EEA country and (iii) no sustainable recovery occurred afterwards:

(a)   In the United Kingdom, the Sistrix Visibility Index indicates that: (i) the visibility of the most important comparison shopping services in terms  of traffic in Google's general search results pages was at its highest at the end of 2010 and the beginning of 2011; (ii) this was followed by a sudden decline, after April 2011 when the Panda algorithm was extended to all English language queries worldwide. The visibility of competing comparison shopping services has not since recovered; (iii) even when a limited number of comparison shopping services subsequently recovered some visibility (e.g. twenga.co.uk in April 2013) this was limited, and did not result in visibility returning to the level prior to the end of 2010/beginning of 2011; and (iv) such limited recovery was followed by another decline, which persisted at least until 2 December 2016.

The above trends are illustrated by Graphs 1 and 2.

(b)   In Germany, the Sistrix Visibility Index indicates that: (i) the visibility of the majority of the most important comparison shopping services in terms of traffic in Google's general search results pages declined after August 2011 when the Panda algorithm was extended to all queries across the EEA; and (ii) the visibility of the vast majority of competing comparison shopping services had not returned to the level prior to August 2011 by 2 December 2016.

In the case of ciao.de, the Sistrix Visibility Index indicates a similar trend: (i)  its visibility level declined suddenly after August 2011; and (ii) that decline persisted until 2 December 2016.

In the case of idealo.de, the Sistrix Visibility Index indicates: (i) an overall increase in visibility until mid-2013; (ii) followed by a sudden decline, which persisted at least until 2 December 2016. […] 401

The above trends are illustrated by Graphs 3 and 4.

(c)   In France, the Sistrix Visibility Index indicates that: (i) the visibility of the majority of the most important comparison shopping services in terms of traffic in Google's general search results pages declined after August 2011 when the Panda algorithm was extended to all queries across the EEA; and (ii) the visibility of the vast majority of competing comparison shopping services had not returned to the level prior to August 2011 by 2 December 2016.

In the case of Twenga, the Sistrix Visibility Index indicates: (i) a sudden decline in visibility after August 2011; (ii) a recovery in visibility at the beginning of 2013; and (iii) a subsequent constant decline in visibility, which persisted until 2 December 2016.

In the case of LeGuide, the Sistrix Visibility Index indicates: (i) a sudden decline in visibility around August 2011; (ii) a slight recovery of visibility until the second half of 2012; and (iii) a subsequent constant decline in visibility, which persisted at least until 2 December 2016.

The above trends are illustrated by Graphs 5 and 6.

(363)    As indicated in recitals (475) to (477), the visibility trends for the competing comparison shopping services in Google's general search results pages closely match the evolution of generic search traffic from Google's general search results pages, indicating that the lower their visibility in Google's general search results pages, the less traffic competing comparison shopping sites get.

(364)    Third, data provided by Google shows  that,  between  November  2014  and  October 2015, the trigger rates411 of the 361 SO Response Aggregators in the: (i) first generic search result; (ii) first four generic search results; and (iii) all generic search results on the first page of Google's general search results412 was low in the vast majority of the thirteen EEA countries in which the Conduct takes place.

(365)    This is true regarding both (i) the trigger rates of any of the 361 SO Response Aggregators in each of the thirteen EEA countries in which the Conduct takes place and (ii) the individual trigger rates of the most important of the 361 SO Response Aggregators in the United Kingdom, Germany and France.

(366)    In the first place, regarding the trigger rates of any of the 361 SO Response Aggregators in each of the thirteen EEA countries in which the Conduct takes place:

(a)   In the United Kingdom, for only […] of the product queries was there a SO Response Aggregator triggered in the first generic search result. At the same time, in only […] of the product queries was there at least one SO Response Aggregator triggered in the first four generic search results and in […] of product queries was at least one comparison shopping service displayed on the first general search results page.

(b)   In Germany, for only […] of the product queries was there a SO Response Aggregator triggered in the first generic search result. At the same time, in only […] of the product queries was there at least one SO Response Aggregator triggered in the first four generic search results, and in […] of product queries was at least one comparison shopping service displayed on the first general search results page.

(c)   In France, for only […] of the product queries was there a SO Response Aggregator triggered in the first generic search result. At the same time, in only […] of the product queries was there at least one SO Response Aggregator triggered in the first four generic search results, and in […] of product queries was at least one comparison shopping service displayed on the first general search results page.

(d)   In the Netherlands, for only […] of the product queries was there a SO Response Aggregator triggered in the first generic search result. At the same time, in […] of the product queries there was at least one SO Response Aggregator triggered in the first four generic search results, and in […] of product queries at least one comparison shopping service was displayed on the first general search results page.

(e)   In Italy, for only […] of the product queries was there a SO Response Aggregator triggered in the first generic search result. At the same time, in […] of the product queries there was at least one SO Response Aggregator triggered in the first four generic search results, and in […] of product queries at least  one comparison shopping service was displayed on the first general search results page.

(f)   In Spain, for only […] of the product queries was there a SO Response Aggregator triggered in the first generic search result. At the same time, in only […] of the product queries was there at least one SO Response Aggregator triggered in the first four generic search results, and in […] of product queries at least one comparison shopping service was displayed on the first general search results page.

(g)   In Austria, for only […] of the product queries was there a SO Response Aggregator triggered in the first generic search result. At the same time, in only […] of the product queries was there at least one SO Response Aggregator triggered in the first four generic search results, and in […] of product queries at least one comparison shopping service was displayed on the first general search results page.

(h)   In Belgium, for only […] of the product queries was there a SO Response Aggregator triggered in the first generic search result. At the same time, in only […] of the product queries was there at least one SO Response Aggregator triggered in the first four generic search results, and in […] of product queries at least one comparison shopping service was displayed on the first general search results page.

(i)    In the Czech Republic, for only […] of the product queries was there a SO Response Aggregator triggered in the first generic search result. At the same time, in only […] of the product queries was there at least one SO Response Aggregator triggered in the first four generic search results, and in […] of product queries at least one comparison shopping service was displayed on the first general search results page.

(j)    In Denmark, for only […] of the product queries was there a SO Response Aggregator triggered in the first generic search result. At the same time, in only […] of the product queries was there at least one SO Response Aggregator triggered at the first four generic search results, and in […] of product queries at least one comparison shopping service was displayed on the first general search results page.

(k)   In Norway, for only […] of the product queries was there a SO Response Aggregator triggered in the first generic search result. At the same time, in only […] of the product queries was there at least one SO Response Aggregator triggered in the first four generic search results, and in […] of product queries at least one comparison shopping service was displayed on the first general search results page.

(l)    In Poland, for only […] of the product queries was there a SO Response Aggregator triggered in the first generic search result. At the same time, in only […] of the product queries was there at least one SO Response Aggregator triggered in the first four generic search results, and in […] of product queries at least one comparison shopping service was displayed on the first general search results page.

(m)   In Sweden, for only […] of the product queries was there a SO Response Aggregator triggered in the first generic search result. At the same time, in only […] of the product queries was there at least one SO Response Aggregator triggered in the first four generic search results, and in […] of product queries at least one comparison shopping service was displayed on the first general search results page.

(367)    In the second place, regarding the individual trigger rates413 of the most important of the 361 SO Response Aggregators in the United Kingdom, Germany and France:414

The above-mentioned trends are illustrated by the tables below.420

7.2.1.1.2.    The way competing comparison shopping services are displayed in Google's general search results pages

(371)    Competing comparison shopping services can be displayed only as generic search results in Google's general search results pages. They cannot therefore be displayed  in rich format, with pictures and additional information on the products and prices.

(372)    Adding images, price and merchant information to product search results increases click-through rates. This is confirmed by the following evidence.

(373)     First, […] 424

(374)    Second, Google indicates on its website that search results from Google Shopping displayed within general search results (which it sells under the brand “Product Listing Ads”) attract “[m]ore traffic and leads”, and that “[m]any businesses experience significantly higher clickthrough rates (CTR) with Product Listing Ads compared to text ads shown in the same location for Google Shopping searches  […] [i]n some cases, advertisers have experienced double or triple standard clickthrough rates”.425

(375)    Third, several eye-tracking studies and research indicate that such features have a considerable impact on user behaviour and consequently on click-through rates.

(376)    In the first place, a 2013 study submitted by […] that tested in a lab-controlled context the click-through rates on comparison shopping results displayed in different graphical formats on Google’s general search results pages indicates that the click- through rate on a link is multiplied by a factor of between 2.2 and 3.7 if, instead of being displayed in the form of text accompanied by a static small icon, it is displayed in a form that includes a larger picture representing the relevant product.426

(377)    In the second place, two 2013 eye-tracking studies reported in a submission by Company X that assessed, among other things, the effects of including images on click-through rates on links to competing comparison shopping services displayed below the Shopping Unit in Google's first and second set of commitments indicate that the addition of images to links to competing comparison shopping services significantly increase the proportion of study participants clicking on those links.427

7.2.1.2. The way that Google’s comparison shopping service is positioned and displayed in  its general search results pages

(378)    Google's comparison shopping service is both positioned and displayed differently to competing comparison shopping services in its general search results pages, despite Google having similar characteristics to those of competitors.

7.2.1.2.1. The different ways that Google’s comparison shopping service is positioned in its general search results pages

(379)    There are two main differences in the way that Google’s own comparison shopping service and competing comparison shopping services are positioned in Google's general search results pages, despite Google’s comparison shopping service having similar characteristics to competing comparison shopping services: (i) Google's own comparison shopping service is not subject to the same ranking mechanisms as its competitors, including adjustment algorithms such as […] and Panda; and (ii) when triggered, Google positions results from its own comparison shopping service on its first general results page in a highly visible place (i.e. either above all generic search results or, in the majority of cases, within or at the level of the first generic few search results).

(380)    First, Google’s own comparison shopping service is not subject to the same ranking mechanisms as competing comparison shopping services, including adjustment algorithms such as […] and Panda. This is despite the fact that Google’s own comparison shopping service exhibits several of the characteristics that make competing comparison shopping services prone to being demoted by the […] and Panda algorithms.

(381)    In the first place, Google's own comparison shopping service also includes […], […] for the […] and Panda algorithms. This is confirmed by the following:

(a)   An internal Google email of 31 March 2006, in which its generic search team indicated that pages from Froogle were “unlikely to appear high in the search results”, because “most of the information that is on [Froogle's] pages that we create is aggregated from various sources, and those sources often have that material online already”.428

(b)     […] 429

(c) Replies from a number of respondents to the Commission's requests for information indicating that Google's comparison shopping service contains content from competing comparison shopping services: “Google published consumer reviews on Google Shopping that had been taken […] from our […] websites”;430 “Google used reviews from […] from both products and shops in Google Shopping”;431 and “the core of the Google Shopping content beyond basic product information is scraped content that Google takes from other sites, including […] […]”.432

(382)    In the second place, Google was aware that its comparison shopping service would not rank highly in Google's general search pages if it were subject to the same ranking mechanisms that apply to competing comparison shopping services. This is confirmed by an internal Google email of 6 May 2009 by […] related to Google Product Search, which stated that: “From a principal perspective it would be good if we  could  actually  just  crawl  our  product  pages  and  then  have  the[m] rank organically (…) Problem is that today if we crawl it will never rank”.433

(383)    In the third place, […],434 […] ComScore data indicates that the […] to certain comparison shopping services, including the standalone Google Shopping website, is similar. This is illustrated by Table 17.435

(389)    Google was, however, aware that, if the Product Universal was positioned at the bottom of the general search results page, it would attract limited traffic: “Putting [OneBox] results on bottom will get no clicks”.441 Google was also aware that positioning the Product Universal in the middle (above the fourth generic search result) instead of at the top (above the first generic search result) would result in a loss of traffic, despite there being no change in the content displayed: “the change to move all product results from pos 1 to 4 is on the current […] release train and will be live as soon as the current […] release is out (i.e. in several weeks). Second, in our Search Properties OCQ review this week, we made a brief mention of the upcoming change (since it will decrease product search traffic by about 20% we wanted to give a heads-up). […] objected to the notion of this much demotion for products, and asked for a followup discussion in a product review”.442

(390)    As of 2009, Google began to ensure that the Product Universal would always be positioned at the top of the first Google general search results page when a competing comparison shopping service would be ranked as the first generic search result. This is confirmed by an internal Google email dated 9 July 2009 by […]:

“Here is what we all agreed to: * The PS onebox [Product Universal] should trigger at the top any time the top result is from another comparison shopping engine (shopping.com, pricegrabber, nexttag [sic – Nextag], etc.)”.443

(391)    A similar policy remained in place at least until 7 September 2010. Indeed, at that time, in order to trigger the appearance of the Product Universal, and/or to be positioned in the middle or top position of the first general search results page, Google was using in particular the following signals: “[t]he number of stores and the number of shopping comparison sites in the top results from Google’s generic search algorithms” and “the presence of a shopping comparison engine in the top-3 generic search results”.444

(392)    Between October 2007 and June 2010,445 more than 80% of all Product Universals appeared in the top or middle position (i.e. above the fourth generic search result) of the first general search results page.

(393)    As of June 2010, Google made it progressively possible for the Product Universal to appear anywhere among generic search results on the first general search results page.446 It nevertheless continued to be positioned prominently on the first general search results page. For instance, between April and September 2010, more than 70% of all Product Universals were positioned on the first general search results page above the first, second, or third generic search result. Between October 2010 and April 2011, the proportion was approximately 60%.447 Furthermore, between April 2010 and April 2011, between 76.7% and 85% of the Product Universals were positioned on the first general search results page before the first, second, third or fourth generic search result.448

(394)    As of May 2012, Google began to transition Google Product Search and the Product Universal to Google Shopping and the Shopping Unit as outlined in recital (31).

(395)    The Shopping Unit is always positioned at the top of the first Google general search results page.449 It can be positioned either on the left hand side of the page, that is, in the same column as generic search results, in which case it is positioned above the first generic search result, or on the right hand side of the page, in which case it is positioned above the AdWords results, if there are any. In instances where it is positioned on the left hand side of the first general search results page, it generally features five items spread across one row. In instances where it is positioned on the right hand side of the first general search results page, it generally features eight items distributed over two rows.

(396)    Moreover, contrary to Google's claim that the Shopping Unit is triggered in the general search results pages for a limited percentage of product queries450, the trigger rate of the Shopping Unit exceeds: (i) in most instances the trigger rate of all 361 SO Response Aggregators (taken together) in the first four generic search results; (ii) in all instances the trigger rate of all 361 SO Response Aggregators taken together in the first generic search result.451 This is illustrated by the following:

(a)                            In the United Kingdom, in November-December 2014, for […] of the product queries was there a SO Response Aggregator triggered in the first generic search result of Google's general search results pages and for […] of  the product queries was there at least one SO Response Aggregator triggered in the first four generic search results. The Shopping Unit was triggered for […] of the product queries in 2014.

(b)                           In Germany, in November-December 2014, for […] of the product queries was there a SO Response Aggregator triggered in the first generic search result of Google's general search results pages and for […] of the product queries there was at least one SO Response Aggregator triggered in the first four generic search results. The Shopping Unit was triggered for […] of the product queries in 2014.

(c)                            In France, in November-December 2014, for […] of the product queries was there a SO Response Aggregator triggered in the first generic search result of Google's general search results pages and for […] of the product queries was there at least one SO Response Aggregator triggered in the first four generic search results. The Shopping Unit was triggered for […] of the product queries in 2014.

(d)                           In the Netherlands, in November-December 2014, for […] of the product queries was there a SO Response Aggregator triggered in the first generic search result of Google's general search results pages and for […] of  the product queries there was at least one SO Response Aggregator triggered in the first four generic search results. The Shopping Unit was triggered for […] of the product queries in 2014.

(e)                            In Italy, in November-December 2014, for […] of the product queries was  there a SO Response Aggregator triggered in the first generic search result of Google's general search results pages and for […] of the product queries there was at least one SO Response Aggregator triggered in the first four generic search results. The Shopping Unit was triggered for […] of the product queries in 2014.

(f)                              In Spain, in November-December 2014, for […] of the product queries was there a SO Response Aggregator triggered in the first generic search result of Google's general search results pages and for […] of the product queries was there at least one SO Response Aggregator triggered in the first four generic search results. The Shopping Unit was triggered for […] of the product queries in 2014.

(g)                           In Austria, in November-December 2014, for […] of the product queries was there a SO Response Aggregator triggered in the first generic search result of Google's general search results pages and for […] of the product queries there was at least one SO Response Aggregator triggered in the first four generic search results. The Shopping Unit was triggered for […] of the product queries in 2014.

(h)                           In Belgium, in November-December 2014, for […] of the product queries was there a SO Response Aggregator triggered in the first generic search result of Google's general search results pages and for […] of the product queries there was at least one SO Response Aggregator triggered in the first four generic search results. The Shopping Unit was triggered for […] of the product queries in 2014.

(i)                               In the Czech Republic, in November-December 2014, for […] of the product queries was there a SO Response Aggregator triggered in the first generic search result of Google's general search results pages and for […] of  the product queries there was at least one SO Response Aggregator triggered in the first four generic search results. The Shopping Unit was triggered for […] of the product queries in 2014.

(j)                               In Denmark, in November-December 2014, for […] of the product queries was there a SO Response Aggregator triggered in the first generic search result of Google's general search results pages and for […] of the product queries there was at least one SO Response Aggregator triggered in the first four generic search results. The Shopping Unit was triggered for […] of the product queries in 2014.

(k)                           In Norway, in November-December 2014, for […] of the product queries was there a Response Aggregator triggered in the first generic search result of Google's general search results pages and for […] of the product queries there was at least one SO Response Aggregator triggered in the first four generic search results. The Shopping Unit was triggered for […] of the product queries in 2014.

(l)                               In Poland, in November-December 2014, for […] of the product queries was there a SO Response Aggregator triggered in the first generic search result of Google's general search results pages and for […] of the product queries there was at least one SO Response Aggregator triggered in the first four generic search results. The Shopping Unit was triggered for […] of the product queries in 2014.

(m)                     In Sweden, in November-December 2014, for […] of the product queries was there a SO Response Aggregator triggered in the first generic search result of Google's general search results pages and for […] of the product queries there was at least one SO Response Aggregator triggered in the first four generic search results. The Shopping Unit was triggered for […] of the product queries in 2014.

7.2.1.2.2.   The different way that Google’s comparison shopping service is displayed in its general search results pages

(397)    The main difference between the way that Google’s own comparison shopping service and competing comparison shopping services are displayed in Google's general search results pages is that specialised search results from Google’s comparison shopping service are displayed with richer graphical features, including pictures and dynamic information.

(398)    Those richer graphical features lead to higher click-through rates. This is confirmed by the following.

(399)     First, […].452

(400)    Second, in response to the Commission's requests for information, Google has stated that it has developed “the display formats of Product Universals and Shopping Units based on what it thought it was more relevant for users […] Product  images, together with information on products, prices, and merchants, let users quickly see that the results relate to product offers and products, what products are involved, their price, and merchants offering them”.453 It also mentions “the display formats in Product Universals and Shopping Units let users distinguish these results from Google's generic blue links”.454

(401)    Third, two 2013 eye-tracking studies reported in a submission by Company X that assessed, among other things, the effects of including images on click-through rates on links to competing comparison shopping services displayed below the Shopping Unit in Google's first and second set of commitments indicate that the addition of images to links to competing comparison shopping services significantly increase the proportion of study participants clicking on those links.455

7.2.1.3.  Google's arguments and the Commission's response

(402)    Google essentially relies on four arguments to contest the fact that it positions and displays its comparison shopping service more favourably in its general search results pages compared to competing comparison shopping services.

(403)    First, Google claims that the positioning and display of the Product Universal did not favour Google's comparison shopping service.456 The Commission's definition of a comparison shopping service implies a “separate stand-alone service [that] in response to queries […] returns product offers from merchants' website [and offers] search tools that are specific” to product search. The function of Product Universals was, however, to provide relevant responses to user queries, like all results on Google's general search results pages. Google also claims that the majority of links within the Product Universal led to the websites of merchants and not to Google's standalone Product Search page.

(404)    Second, Google claims that the positioning and display of the Shopping Unit does  not favour Google's comparison shopping service.457 The Shopping Unit is not a comparison shopping service but rather an improved form of AdWords results458 because: (i) the Shopping Unit does not exhibit the characteristics of a comparison shopping service; (ii) […];459  (iii) […].460

(405)    Third, Google claims that, in any event, competing comparison shopping services  can benefit from the same positioning and display as the Shopping Unit since they are eligible to participate in Google Shopping.461

(406)    Fourth, Google claims that it held Product Universals to the same relevance  standards that it applies to all of the generic search results on its general search results pages and that it holds the Shopping Unit to the same relevance standards that it applies to all of its product ads.462

(407)    Each of these four claims is unfounded.

(408)    First, the Commission's case is not that the Product Universal was in itself a comparison shopping service.463 Rather, the Commission's case is  that  the positioning and display of the Product Universal was one means by which Google favoured its comparison shopping service.

(414)    In the first place, the Shopping Unit is based on the same database of products and merchants as the standalone Google Shopping website. For both the Shopping Unit and the standalone Google Shopping website, merchants must have a Google Shopping Merchant Centre account containing information related to the products on offer. The merchants deliver the feed of information on their product through the Google Shopping Merchant Centre.

(415)    In the second place, the selection of paid product results displayed in the Shopping Unit presents many common technological features and mechanisms with the selection of results on the standalone Google Shopping website:468

(a)     […]

(b)     […]

(416)    In the third place, merchants cannot choose to have their products displayed only in the Shopping Unit or in the standalone Google Shopping website, neither in aggregate, nor for individual products. Google alone determines whether a given product of a given merchant appears in the Shopping Unit or the standalone Google Shopping website […].469  This is confirmed by the following:

(a)     […].470

(b)     […].471

(c)     […]472 […].

(d)     […].473

(417)    In the fourth place, merchants are not informed of whether they are paying for a click that happened on the Shopping Unit or on the standalone Google Shopping website.474

(418)    In the fifth place, the links in the Shopping Unit and the standalone Google Shopping website lead to the same landing page on the merchant site for a given product where users can purchase that product.

(419)    In the sixth place, there are now carousel formats for the Shopping Unit that allow  the unit to unfold up to 17 results and where both a header link and a 'view all' link at the end of the carousel lead to the standalone Google Shopping website.

(420)    In the seventh place, Google presents the Shopping Unit and the standalone Google Shopping website as a single service or experience to merchants and users. This is confirmed by the following:

(a)              In eleven of the thirteen EEA countries where Shopping Units have been launched (with the exception of the United Kingdom and Poland), the results within the Shopping Unit are labelled as “results from Google Shopping”.475

(b)             A Google blogpost of 29 April 2011, “Google Shopping Arrives in Australia, Italy, Spain, and the Netherlands”,476 refers to both the Shopping Unit and the standalone Google Shopping website as “Google Shopping”.

(c)              A Google blogpost of 21 May 2013 “Connect, Click and Convert Around the World with Google Shopping”,477 refers to both the Shopping Unit and the standalone Google Shopping website as “Google Shopping”.

(d)             An undated Google blogpost on Google Shopping (.fr),478 refers to both the Shopping Unit and the standalone Google Shopping website as “Google Shopping”.

(e)              An undated Google blogpost on Google Shopping (.fr),479 refers to both the Shopping Unit and the standalone Google Shopping website as “Google Shopping”.

(f)     […]480 […].

(g)     […].481

(h)     […]482 […].

(i)    An internal Google email exchange of […],483 which indicates that in the countries where the standalone Google Shopping website had just been launched (Austria, Belgium, Denmark, Norway, Poland and Sweden), the Shopping Unit contains a clickable header link that leads to the standalone Google Shopping website, like for all other countries with a standalone Google Shopping website.

(j)      […].484

(k) The mock-ups of the Shopping Unit provided by Google  during  the commitments discussions described the results displayed in the box as “Google Shopping results”. The accompanying explanation which would be displayed  to the user when hovering over the “i” icon read “The Google Shopping results link is inserted to show more results from Google's Shopping results. For shopping results from other relevant providers, click on the links for “Alternatives” or see Google's other search results […]”.

(421)    In the eighth place, links within the Shopping Unit fulfil the same economic function as links within Google's standalone comparison shopping website. Both lead users directly to the websites of Google's merchant partners and trigger a payment by the relevant partner to Google. Google's comparison shopping service therefore benefits economically from that click in the same manner as if the user had taken the intermediary step of going through the standalone Google Shopping website before clicking on the product of that merchant partner.

(422)    In the ninth place, the conclusion that the positioning and display of the Shopping Unit is one means by which Google favours its comparison shopping service is not called into question by the fact that during an initial period, Google Shopping existed only in the form of the Shopping Unit without an associated standalone website in  six of the thirteen EEA countries in which the Conduct takes place (Austria,  Belgium, Denmark, Norway, Poland and Sweden).485 As indicated by the evidence in recitals (420)(g)-(420)(i), the fact that Google Shopping initially existed only in the form of the Shopping Unit in these six countries was a transitional phase, due to the fact that Google had only a limited number of merchant partners in these    countries, which made it more difficult to offer a standalone website. Moreover, during that transitional period, users in those six EEA countries could visit the standalone websites in the other seven EEA countries (the Czech Republic, France, Germany, Italy, the Netherlands, Spain and the United Kingdom).

(423)    In the tenth place, the fact that certain Shopping Units in the EEA display different offers for different product items is irrelevant in relation to the Conduct, since the Commission's case is not that the Shopping Unit is in itself a comparison shopping service which displays different offers for the same product items, but rather that the positioning and display of the Shopping Unit is one means by which Google favours the standalone Google Shopping website which does show different offers for the same product items.

(424)    Third, Google Shopping and the Shopping Unit are evolutions respectively of  Google Product Search and the Product Universal. This is confirmed by the  following evidence:

(a)              A Google blogpost of 31 May 2012 announcing the launch of Google  Shopping in the US, states the following: “First […] we are starting to transition Google Product Search in the U.S. to a purely commercial model built on Product Listing Ads. This new product discovery experience will be called Google Shopping and the transition will be complete this fall. […] Second, starting today we’ve also begun to experiment with some new commercial formats on Google.com that will make it easier for users to find and compare different products. These include larger product images that give shoppers a better sense of what is available and also the ability to refine a search by brand or product type”.486

(b)             A Google blogpost of 31 May 2012 explains that “Product Listing Ads appear on  google.co.uk  search  results  in  the  commercial  unit  [the Shopping Unit] […]” and that “Search  results in Google Shopping     [google.co.uk/shopping] [the standalone Google Shopping page] are built on Product Listing Ads”.487

(c)     […].488

(d)   An undated Google blogpost, states that “Product Listing Ads appear when someone searches for your products on google.co.uk and google.co.uk/shopping”.489

(e)   An article published by Search Engine Land on 31 May 2012, states that: “rather than the Product Listing Ads and Google Product Search results being separate, both will be combined into a single Google Shopping box [the Shopping Unit]”. 490

(f)   A Search Engine News blogpost of 4 June 2012 on the SEO chat website regarding the transition from Google Product Search to Google Shopping states that while Product Listing Ads (“PLAs”) and Google Product Search were displayed separately in the past, after the transition they would be merged and displayed together in the Shopping Unit.491

(g)   An article of Search Engine Journal of 4 June 2012 explains that Google Product Search would be renamed Google Shopping and that the Shopping Unit would merge and display together PLAs and Google Product Search.492

(h)   A blogpost of 19 August 2012 by […], relating to the evolution of Google's own comparison shopping service, states that Google Product Search would be replaced by PLAs and presented an infographic to illustrate the evolution of Google's own comparison shopping service from Froogle to Google Product Search to Google Shopping.493

(i)   A blogpost of Alphr, a website publishing articles and various posts dedicated to new technology, of 6 September 2013, explains that Google Shopping would be a service that is “identical to Product Search except that shop owners must now pay to have their product included”.494

(j)   An article published by Search Engine Land495 on 5 December 2013 stated, when referring to the Shopping Unit, that: “[…] The results in the box [the Shopping Unit] are drawn from Google Shopping, which controversially  shifted to an all-ad model last year”.

(k)   A blogpost of 9 December 2013 published on the website of Universem,496 a Belgian marketing agency, referred to the Shopping Unit as an “Example of search with Google Shopping ads in the SERP”.497

(l)   An article published on the website of Hallam Internet, a digital marketing agency based in the United Kingdom, on 26 June 2014, states that Google Shopping results appear both “[…] above the search engine results or within the right hand sidebar next to the results”.498

(m)  A blogpost of BidCops, an AdWords tool of the United Kingdom digital agency Click Through Marketing, of 31 March 2015, explains the following: “In June 2012, Google made the most important change to its shopping search thus far. Along with another name change – this time to Google Shopping – the service would now follow a paid model, with advertisers bidding to list their products in favourable positions”.499

(n)   An undated blogpost on Search Engine Land, states that “Google Shopping is a dedicated shopping search engine from Google […]” and that “Google Shopping results do appear within regular Google searches […]”.500

(425)    Fourth, paid product results in the Shopping Unit are not an improved form of AdWords results.

(426)    In the first place, AdWords advertisers bid on keywords while Shopping Unit merchants bid on products they want to promote and on which they upload the information in the Merchant Centre.501

(427)    In the second place, unlike for paid product results in the Shopping Unit, there is no limited pool of advertisers who can bid for being included in AdWords results and no restrictions on the categories of participants. By contrast, comparison shopping services can bid to be displayed in AdWords and not in the Shopping Unit.

(428)    In the third place, AdWords results and paid product results in the Shopping Unit have different landing page criteria. Links in paid product results in the Shopping Unit must go to pages where users can purchase a product,502 while for AdWords ads links can also lead to comparison shopping services without a purchase functionality.

(429)    In the fourth place, […]503 […].

(430)    In the fifth place, AdWords results only contain text whereas paid product results in the Shopping Unit have a distinctive display (pictures, information on products, prices).504

(431)    In the sixth place, AdWords results are not limited in scope whereas paid product results in Shopping Units relate only to products.505

(432)    In the seventh place, participation in AdWords does not require the provision of structured data feeds from merchants who must constantly update information on their products.506

(433)    In the eighth place, the search results pages on the standalone Google Shopping website include both paid product results from Google Shopping (labelled as sponsored links with specific rules, including on ranking) and AdWords text ads (labelled with a different icon which provides different information from that if one hovers on the paid product results).

(434)    In the ninth place, different policies govern the content of AdWords results and paid product results from the Shopping Unit. This is confirmed by Google's AdWords help pages concerning ad formats, that specify that:

(a)              “Google Shopping ads are not considered ad formats”;507 and

(b)             the “[…] content for Shopping campaigns and Shopping ads needs to comply with the Google Shopping Policies, which are different from the AdWords Advertising Policies”.508

(435)    This is also confirmed […].509 […].

(436)    In the tenth place, merchants perceive AdWords results and paid product results from the Shopping Unit differently. This is confirmed by the following:

(a)              An article published by Search Engine Land on 30 April 2012 regarding the launch by Google of hybrid “comparison units” for certain of its other specialised search services such as Flight Search, Hotel Search and Google Advisor. The article reported that Google referred to these comparison units as a “third type of thing” that is neither generic search results nor AdWords paid search results. The article also stated that “The comparison units also carry a “Sponsored” disclaimer rather than an “Ads” one, as with AdWords ads. This seems part of Google’s positioning the new units as something different than ads”.510

(b)             A blogpost of SearchEngineWatch of 21 August 2014,511 which highlights a number of advantages of paid product results from the Shopping Unit as compared to AdWords results: “[…] In addition to being highly effective […], PLAs [paid product results ] offer other advantages as well. For instance, unlike text ads, PLAs give advertisers the opportunity to include multiple products in a PLA grouping, increasing your chance of getting clicked and sometimes even allowing your offers to dominate the SERP. You also don’t have to worry about copy. In addition, PLAs can give lesser-known advertisers the ability to capture consumer attention on the SERP in categories where there’s an absence of bigger brands participating. […]”.

(c)     […].512

(d)     […].513

(e)     […].514

(f) A blogpost of 17 November 2015 of  GoDataFeed,515  which  highlights  a  number of differences between paid product results from the Shopping Unit and AdWords results: “[…]For those of you still on the fence about allocating more funds to product listing ads (over text ads), here are a few key points that may persuade you: Cost-per-click rates for generic search terms can get expensive; Showing the price and style of products leads to more qualified traffic; Appearing more than once can lead to higher conversion volume; Visual  nature of product listing  ads was more  successful in getting   people’s attention; Brans search queries can trigger both search text ads and  Shopping ads […]”.

(g)     […]

(h)     […]516 […].

(437)    In the eleventh place, users perceive AdWords results and paid product results from the Shopping Unit differently. This is confirmed by a study submitted by […] in July 2013 in response to the commitments offered by Google in April 2013.517 Within the study, an online survey was conducted between 18 June and 22 June 2013 using a series of simulated searches and covering a sample of 1 888 respondents residing in the United Kingdom. The survey found that the Shopping Unit commands a high degree of consumer clicks and attention, in part because of its location on the search results page and the visually-rich display. The study also concluded that, in spite of the 'sponsored' label of the Shopping Unit, over 50% of the respondents believed that it was unpaid content, while 34% correctly indicated that its content is paid.

(438)    In the twelfth place, the conclusion that paid product results in the Shopping Unit are not an improved form of AdWords results is not called into question by the fact that, […],518 […].

(439)    Fifth, competing comparison shopping services are not eligible to participate in Google Shopping, unless they change their business model by adding a direct purchase functionality or acting as intermediaries for placing merchants' paid product results in the Shopping Unit (see also recital (220)(2) on the eligibility criteria for Google Shopping). Indeed, the examples that Google gives are only of comparison shopping services that have changed their business model in one of the ways described above.519

(440)    Sixth, it is irrelevant whether, as Google claims, it held Product Universals to the same relevance standards that it applies to all of the generic search results on its general search results pages and that it holds the Shopping Unit to the same  relevance standards that it applies to all product ads.520 The Commission does not object to Google applying certain relevance standards but to the fact that Google's own comparison shopping service is not subject to those same standards as competing comparison shopping services.

(441)    Moreover, and in any event, Google has not demonstrated it held Product Universals to the same relevance standards that it applies to all of the generic search results on  its general search results pages and that it holds the Shopping Unit to the same relevance standards that it applies to all product ads.

(442)    While Google relies on two reports of ad hoc experiments regarding whether users find the Product Universal useful compared to the first generic search results displayed at the time by Google, 521  they are not probative because: (i) users can only compare what they see and the competing comparison shopping services are rarely visible on Google's first generic results page; (ii) users may find the Google comparison shopping service useful not because of the relevance of its results, but because it is displayed with richer features; and (iii) the comparison is made between generic search results in general, not results from competing comparison shopping services.

(443)    Internal Google documents confirm the lack of probative value of the above- mentioned experiments:

(a)              An internal Google document of 30 April 2008 reporting on the results of a  […] experiment aimed at improving the Google Product Universal triggering and page placement experiment are presented concludes that “the […] score mainly tells that the proposed change is better than the status quo, but does not reveal how appropriate the proposed placement of product onebox [Product Universal] is, or how accurate the triggering is”.522

(b)             In an internal Google exchange of emails of 27 February 2009, […], confirms that users can only compare what they see (“The point is that we win not because we are so much better than an Amazon or NexTag as a store, but that in many cases there just aren't stores on the page if not for product universal”).523

7.2.2.  The importance of user traffic for comparison shopping services

(444)    The Commission concludes that user traffic is important for the ability of a comparison shopping service to compete in several ways. As […], the owner of several comparison shopping services, including […], puts it, “traffic is the most important “asset” of a [specialised] search engine; it increases the relevance of [specialised] search services for a variety of reasons”.524

(445)    First, the relevance of a comparison shopping service for users is related to the breadth and the freshness of its product offering. Traffic enhances the ability of comparison shopping services to convince merchants to provide them with data  about their products. This is confirmed by the following statements from competing comparison shopping services.

(1)   As indicated by […]: “More traffic raises the relevance and attractiveness of a vertical search service for the partners who aim to sell their products/offer  their services (i.e. online shops in the case of […], real estate agents in the  case of […]) which in turn results in more and better data and offerings that the vertical search service can provide to users”.525

(2)   As indicated by […]: “To have the largest possible number of offers, one of the key arguments with online retailers is the volume of traffic potentially  provided, hence the need for a high traffic”.526

(3)   As indicated by […]: “traffic brings revenues and credibility. Revenues and credibility make much easier the partnership with product companies that provide their product flows”.527

(4)   As indicated by […]: “The more traffic a comparison shopping website can generate, the more interesting it becomes to potential partners (shops, advertisers) and other internet users. This is due to the fact that the contents of a comparison shopping site are mainly supplied by partners (e.g. product info, prices) and internet users (e.g. user generated content such as shop and product reviews, blog entries etc.)”.528

(5)   As indicated by […]: “Traffic levels also are critical for our ability to enlist and negotiate competitive terms from merchants since merchants are less interested in relationships with parties that cannot send a significant amount of traffic. As with any search engine, category breadth – or in the case of a shopping vertical search engine, the number of products in the catalogue and merchants listing each product – is critically important for the success of the site. Catalogue breadth allows us to generate more traffic by bidding on more keywords and indexing more pages and creates a better user experience by ensuring that users will find the results that they desire, thus increasing the likelihood that the user will return to the site. Sufficient traffic is essential to allow us to enlist an adequate number of merchants, and to convince the merchants to submit an adequate number of products for inclusion on our site, so that we can establish the catalogue breadth that we need. This catalogue breadth  also results  in superior  search results  because we will have more comprehensive product results for users’ searches”.529

(446)    Second, traffic generates revenue (via either commissions from merchants or online search advertising) that can be used to invest in order to improve the usefulness of the services provided. This is confirmed by […]: “The more traffic increases, the more revenue grows and the more the publisher of the service has means to develop itself and distinguish itself from competition”.530

(447)    Third, traffic allows machine learning effects, thereby improving the relevance of the results of comparison shopping services and the usefulness of the service they offer to users. This is similar to the effects observed for general search engines (see section 6.2.2). The existence of these effects is confirmed by a number of respondents to the Commission's requests for information.

(1)   As indicated by […]: “The existence of traffic is a key factor to increase the relevance of a vertical search engine. At […], we are using notably different “signals” from visits such as click rates on our offers, the rate of click on the offers of a merchant in general, according to the positions of offers in our search results pages. Substantial investments have been made to integrate with

[…] “behavioural” signals in improving the quality of our search results”.531

(2)   As indicated by […], it “runs a popularity algorithm (the most clicked, most searched products and offers are showed in the higher ranks of our results pages) which performs better and better with increasing amounts of queries and traffic”.532

(3)   As indicated by […], “traffic levels are critical for all aspects of [[…]] business, including quality of results. (...) Traffic data allows for a calculation of the most frequently selected products for each search, thus allowing for a data-driven selection of products in response to each search. Higher traffic volume  also  allows  [[…]]  to  test  search  algorithms  to  determine which algorithm provides most useful results for users”.533

(4)   As indicated by […], “Data about user behaviour on the sites improves our ability to make improvements to our search service”.534

(5)   As indicated by Company X, “[s]earch services often employ statistically driven processes to improve the relevance of search results. Generally speaking, as the volume of data increases and accumulates, so do the relevance and accuracy of results, and errors decrease. Over the volume of data that a vertical [specialised] search service is typically able to accumulate, increases in data volume may produce significant gains in accuracy and reductions in error. To the extent that a vertical search service is impeded in attractive queries, it collects less data, and the quality of its search responses is harmed”.535

(6)   As indicated by […], “[…] there is a link between the overall volume of traffic and the relevance of results brought to internet users (not quantified). Indeed, there is a snowball effect which is inherent in the system. (...) Our search engine is based on relevance and popularity. The more an offer is clicked on, the better its position will be in the search results, and the higher its  relevance for internet users. The larger the number of offers suggested on our engines, the more internet users are likely to find an offer which is totally relevant”.536

(448)    Fourth, traffic allows comparison shopping services to carry out experiments aimed  at improving their services. This is confirmed by […], which indicated that: “Higher traffic also allows for more effective experimentation and optimization of the site, which results in creation of a better user experience, development of more engaging services and features resulting in more repeat traffic and allows for more effective monetization. As described elsewhere, we deploy new features to some users and maintain the old features as a comparison, which allows us to determine the impact of new features on site performance and user experience”.537

(449)    Fifth, in addition to providing specialised search results for given queries,  comparison shopping services can also suggest other search terms that may be of interest for users. The ability of comparison shopping services to offer such suggestions is linked to the volume of traffic that they receive.

(1)   […] stated that “[m]ore traffic generates better information about a vertical [specialised] search service's “top items” and helps to identify trends in the demands of users, which in turn allows the search service to provide better suggestions or a better display of results”.538

(2)   […] also indicated that “In addition to standard search results, we believe that valuable shopping information also needs to be personal and relevant for an individual user. As a result, quality of results depends on our ability to personalize results and, increasingly, to make recommendations based on information from a user’s friends, as described below. All of these features require a tremendous amount of traffic. Higher traffic is essential for developing “social” recommendations, which can significantly improve the quality of search results. We expect to allow users to connect with other users, and we can use data from a person’s connections to refine search results and help recommend products to users”.539

(450)    Sixth, traffic allows comparison shopping services to generate more original user reviews.540 Users often use comparison shopping services with a view to reading original reviews by other users.

(451)    Seventh, Google does not contest the importance of user traffic for comparison shopping services.

7.2.3. The Conduct decreases traffic from Google's general search results pages to competing comparison shopping services and increases traffic from Google's general search results pages to Google's own comparison shopping service

(452)    The Commission concludes that, notwithstanding Google's arguments to the contrary (section 7.2.3.4), the Conduct diverts traffic in the sense that it: (i) decreases traffic from  Google's  general  search  results  pages  to  competing  comparison shopping services; and (ii) increases traffic from Google's general search results pages to Google's own comparison shopping service.

(453)    The conclusion is supported by: (i) analysis of user behaviour that indicates that generic search results generate significant traffic to a website when they are ranked  in the first three to five generic search results on the first general search results page (section 7.2.3.1); (ii) evidence on the impact of the Conduct on generic search traffic from Google's general search results pages to competing comparison shopping services (section 7.2.3.2); and (iii) evidence on the impact of the Conduct on traffic to Google's own comparison shopping service (section 7.2.3.3).

7.2.3.1. Analysis of user behaviour

(454)    Analysis of user behaviour indicates that generic search results generate significant traffic to a website when they are ranked within the first three to five generic search results on the first general search results page.

(455)    First, users typically look at the first three to five generic search results on the first general search results page and pay little or no attention to the remaining generic search results.541

(456)    Second, on the first Google general search results page, the first three to five generic search results are generally the generic search results above the fold. When a search result is above the fold, this increases the likelihood that a user will click on it and generate traffic to the underlying website.542 Google acknowledges this in an email dated 16 October 2006: “[U]sers rarely scroll, and if they are looking for a group that is not above the fold, it is often difficult for them to find it”.543

(457)    Third, the ten highest-ranking generic search results on the first Google general search results page together generally receive approximately 95% of all clicks on generic search results. Moreover, users are more likely to click on the first generic search result than on any other generic search result.544 This is illustrated by Table 19 which the Commission has compiled.

because of a propensity to click on such a result545 or because they are able to accurately assess that the first generic search result is objectively what they are looking for.

(460)    […] has provided a study by […] which sought to do just that (“the […] study”).546 On the basis of a representative sample of “[0-5]M users in the US market and [0- 5]M users in the UK market out of approximately [80-90]M and [10-20]M, respectively”,547 the study indicates that the rank of a given link in generic search results on the first general search results page has a major impact on the click rates of that link, irrespective of the relevance of the underlying page. For instance, moving the first generic search result to the third rank on the first general search results page has led to a reduction in clicks by about 50%, which was not compensated by instances where users returned to the general search results page after first clicking on the first generic search result.548 The impact on generic search traffic was even stronger when the first generic search result on the first general search results page was moved to the tenth rank: traffic decreased by around 85%.549

(461)    Contrary to what Google claims,550 the […] study also indicates that a modification  of the rank of a generic search result on the first general search results page has a major impact on generic search traffic, even if certain users adapt their behaviour.  For example, when the third generic search result on the first general search results page was moved to the first rank, the number of users that stayed for at least 30 seconds on the landing page (satisfied click-through rate551) after clicking on the result in the first rank still increased by [150 – 200]%.552

7.2.3.2. Impact of the Conduct on generic search traffic from Google's general search results pages to competing comparison shopping services

(462)    The Conduct has led to a decrease in generic search traffic from Google's general search results pages on a lasting basis to almost all competing comparison shopping services in each of the thirteen EEA countries in which the Conduct takes place. This is confirmed by the following.

(463)    First, competing comparison shopping services have provided specific examples of the impact of the Conduct on generic search traffic from Google’s general search results pages to their websites.

(464)    In the first place, as indicated by […], the Panda algorithm led to a [significant] decrease in traffic from generic search results on Google's general search results pages to the website of [Comparison Shopping Site #1] in the United Kingdom in February 2011.553 The website of [Comparison Shopping Site #2] also experienced decreases in generic search traffic from Google's general search results pages in 2010. In […] 2010, those decreases in generic search traffic from Google's general search results pages, as expressed in year-to-year terms relative to […] 2009 […] were [significant] in [Member State X], [Member State Y] and [Member State Z].554 As indicated by […], a further decrease of traffic to [Comparison Shopping Site #2's] website occurred in February 2011. Generic search traffic from Google’s general search results pages to [Comparison Shopping Sites #1 and #2] in the EEA […] has followed  a  declining  trend  since  2011,  marked  with  significant  and recurrent decreases. [The reply of Comparison Shopping Site #2 indicates] that the decrease corresponds to the implementation of different versions of the Panda algorithm.555

(465)    In the second place, as indicated by […], it experienced eight instances of a decrease in generic search traffic from Google’s general search results pages to its comparison shopping services between March 2007 and August 2008. The decreases in traffic in these instances ranged from 69.9% to 89.9%.556 […] also claims that its local website in the United Kingdom experienced a decrease of more than 85% in generic search traffic from Google’s general search results pages on 27 November 2010: “96% of the search phrases on which [[…]] used to rank in #1 were dropped off from rank

#1, 67% of the search phrases on which [[…]] used to rank in the first 5 positions were dropped off from the first 5 positions and 50% of all the search phrases on which [[…]] used to rank in page number #1 (first 10 results) dropped off from page

#1”.557 Generic search traffic from Google’s general search results pages to […] website in the United Kingdom then increased again on 20-21 January 2011,  reaching 30% of the previous generic search traffic from Google’s general search results pages to the website, before going down on 31 January 2011 to approximately 85% of the prior week’s levels, reflecting approximately a 90% decrease relative to the original period. Generic search traffic from Google’s general search results pages then  went  back  up  again  on  16  February  2011  and  stabilised  at  a  level  of approximately 70% of pre-November 2010 traffic, before decreasing again on 26 February 2011.

(466)    In the following period, […] reported further instances of decreases in generic search traffic from Google’s general search results pages to its websites in the EEA. As indicated by […], these decreases took two forms. First, there were instances where the volume of generic search traffic from Google’s general search results pages decreased rapidly in a matter of days. This includes, for instance, an 85% decrease in generic search traffic from Google’s general search results pages to […] as from 6 July 2011, and for 6 months, which was followed by a partial recovery to 50% of the previous volume. Second, there were instances of significant but somewhat more gradual decreases in generic search traffic from Google's general search results pages over several months. As indicated by the figures provided by […], generic search traffic from Google’s general search results pages to […] websites in France, Germany and the United Kingdom, for instance, gradually decreased to materially insignificant numbers over the period ranging from 2012 to the end of 2014. […] reports that all these instances of decreases are closely associated with equivalent reductions in the visibility of […] websites in Google's general search results pages as measured in particular by the Searchmetrics SEO Visibility Index.558

(467)    In the third place, as indicated by […], generic search traffic from Google’s general search results pages to its French website suddenly decreased by 80.4% on 28 October 2010.559 Generic search traffic from Google’s general search results pages to […] French website remained at this low level until at least the end of 2010.560 […] also referred to a number of sudden decreases in generic search traffic from Google's general search results pages to its comparison shopping services (in particular […] and […]) between 2011 and 2014, at the same time as changes to the Panda  algorithm were introduced.561 […] also stated that the release of update 2.4 of Panda on 12 August 2011 led to a progressive decrease in generic search traffic from Google’s general search results pages to its Italian website, […], with the only temporary exception being the time period between updates 4.0 and 4.1 of Panda (19 May 2014 to 23 September 2014). The same considerations apply to […].562

(468)    In the fourth place, as indicated by […], generic search traffic from Google’s general search results pages to its website in mid-2007 that it claimed cannot be explained by a change to the websites it publishes: “[g]enerally speaking, [[…]] listings dropped from a mixture of first and second page rankings to the last page of results (tenth page or lower), for a period of approximately three months”.563 […] also reported a sudden decrease of 30% in generic search traffic from Google’s  general  search results pages around 24 February 2011.

(469)    In the fifth place, as indicated by […], generic search traffic from Google’s general search results pages to its German website suddenly decreased in April 2010 and that decrease cannot be explained by any changes to the websites it publishes.564 […] also submitted that in April 2011, when the first version of Panda was extended to all English language queries, generic search traffic from Google’s general search results pages to […] website in the United Kingdom suddenly fell by […].565 As indicated  by […], between the first quarter of 2011 and the last quarter of 2014, generic search traffic from Google’s general search results pages to […] websites decreased by  more than […] overall, with a stronger impact in key EU countries: […] in France, […] in Germany, and […] in the United Kingdom.566

(470)    In the sixth place, as indicated by […], generic search traffic from Google’s general search results pages to its main comparison shopping website […] suddenly decreased by about 31.4% in November 2012. The decrease corresponds to a reduction in the visibility of the website in Google's general search results pages as reflected by the Sistrix Visibility Index. As indicated by […], the visibility of its website, as well as generic search traffic from Google’s general search results pages, has continued to decrease; by May 2014, that traffic had decreased by 72%. As from 18 May 2014, the visibility of the website in Google's general search results pages progressively  returned  to  its  previous  level,  as  did  generic  search  traffic   from Google’s general search results pages.567

(471)    In the seventh place, as indicated by […], several national […] websites in the EEA suffered sudden decreases in generic search traffic from Google’s general search results pages in the summer of 2011 and in October 2014. […] also reported that its main website, […], suffered a sudden decrease in generic search traffic from Google’s general search results pages of 38% in October 2014.568

(472)    Beyond these two instances, […] also indicated long-term trends of decreases in generic search traffic from Google’s general search results pages to its websites (including the […] websites, the […] websites and the […] websites), which it attributes to a corresponding reduction in the visibility of these websites in Google's general search results pages, as measured for instance by the Sistrix Visibility Index. In aggregate, generic search traffic from Google’s general search results pages to  […] European websites decreased by about 50% between 2011 and 2014.569

(473)    In the eighth place, as indicated by […], generic search traffic from Google’s general search results pages to its websites across a number of EEA countries significantly decreased between 2011 and 2014. These decreases started with an initial sudden decrease in 2011, followed by a continued decrease at a significant rate over the months and years that followed. The rate of the overall decrease in generic search traffic from Google’s general search results pages varied depending on the countries. In larger Member States, such as France, Germany or the United Kingdom, generic search traffic from Google’s general search results pages to […] websites decreased by about 90% between 2011 and 2014. As indicated by […], the decrease in generic search traffic from Google's general search results pages coincided with a significant drop in ranking of its websites in Google’s general search results pages.570

(474)    In the ninth place, as indicated by […], there have been 39 individual instances in which its websites ([…]) experienced significant decreases in generic search traffic from Google’s general search results pages.571 The decreases ranged from  7% to 92%. In some instances, generic search traffic from Google’s general search results pages recovered partially, but in many instances, there was no recovery. […] outlines that  these  decreases  coincided  with  sudden  reductions  in  visibility  in Google's general search results pages (as measured by the Searchmetrics SEO Visibility Index or the Sistrix Visibility Index).

(475)    Second, the Commission has compared during certain reference periods between 2008 and 2014 the evolution of: (i) generic search traffic from Google's general search results pages572 to the websites of three competing comparison shopping services in the United Kingdom (ciao.co.uk, kelkoo.co.uk, dealtime.co.uk), Germany (idealo.de, preisroboter.de, guenstiger.de) and France (ciao.fr, touslesprix.com, twenga.fr);573 with (ii) the Sistrix Visibility Index for these domains.574

(476)    The comparison indicates that the evolution of the Sistrix Visibility Index for these specific domains of competing comparison shopping services follows a similar trend to the evolution of the generic search traffic from Google's general search results pages to these domains. […].575 The Commission's analysis is also illustrated by Graphs 18 to 26.

Graph 18: United Kingdom - evolution of generic search traffic from Google's general search results pages to ciao.co.uk compared to the Sistrix Visibility Index for that domain

[…]

Graph 19: United Kingdom - evolution of generic search traffic from Google's general search results pages to kelkoo.co.uk compared to the Sistrix Visibility Index for that domain

[…]

Graph 20: United Kingdom - evolution of generic search traffic from Google's general search results pages to dealtime.co.uk compared with the Sistrix Visibility Index for that domain

[…]

Graph 21: Germany - evolution of generic search traffic from Google's general search results pages to idealo.de compared with the Sistrix Visibility Index for that domain

[…]

Graph 22: Germany - evolution of generic search traffic from Google's general search results pages to preisroboter.de compared with the Sistrix Visibility Index for that domain

[…]

Graph 23: Germany - evolution of generic search traffic from Google's general search results pages to guenstiger.de compared with the Sistrix Visibility Index for that domain

[…]

Graph 24: France - evolution of generic search traffic from Google's general search results pages to ciao.fr compared with the Sistrix Visibility Index for that domain

[…]

Graph 25: France - evolution of generic search traffic from Google's general search results pages to touslesprix.com compared with the Sistrix Visibility Index for that domain

[…]

Graph 26: France - evolution of generic search traffic from Google's general search results pages to twenga.fr compared with the Sistrix Visibility Index for that domain

[…]

(477)    Contrary to Google's claims,576 the comparison of the evolution of generic search traffic from Google's general search results pages to a website with the Sistrix Visibility Index for that website provides reliable evidence of the impact of the ranking of a given website on its generic search traffic from Google's general search results pages. Indeed, the Sistrix Visibility Index is calculated based solely on the visibility of a website in generic search results on Google's general search results pages (including the website's rank and the layout of the relevant Google's general search results page), and not on the specific content of that website (“domain”) (see recital (361), footnote 398, for more details on the methodology for the Sistrix Visibility Index).

(478)    Third, while the evolution of traffic to a website in a given country can be influenced by several factors, the Conduct has, in most instances led to a lasting decrease in traffic from Google's general search results pages to almost all competing  comparison shopping services in each of the thirteen EEA countries in which the Conduct takes place. This is confirmed by the following.

(479)    In the first place, Google analysed the evolution in the United Kingdom, Germany, France  and the  Netherlands of  generic search traffic  from Google's  general search results pages to the sample of the top ten competing comparison shopping services from those four countries considered in the “Difference-in-Differences” Analysis annexed to the SO Response.577 That analysis indicates a lasting decline of the  generic search traffic from Google's general search results pages to competing comparison shopping services, notably after the introduction of the Panda algorithm, but also over the longer run. This lasting decline applies to all the competing comparison shopping services considered in the analysis in the United Kingdom and France, all but one in Germany, and all but two in the  Netherlands.578

(480)    Contrary to what Google claims, the analysis in recital (479) is representative  because it is based on a sample of the top ten competing comparison shopping services in each country, in terms of generic search traffic from Google's general search results pages, calculated in the three-month period preceding the date of beginning of the Conduct in the country in question as defined by the  Commission.579

(481)    In the second place, the Commission analysed the evolution of generic search traffic from Google's general search results pages to the 361 SO Response Aggregators during the reference period January 2004 to December 2016 in the United Kingdom, France,  Germany,  Spain,  the  Netherlands,   Italy,  Denmark  and   Poland.580      The

analysis indicates that traffic to almost all competing comparison shopping services has declined on a lasting basis or remained broadly stable.

(1)   In the United Kingdom, France, Germany, and Spain, generic search traffic from Google's general search results pages to the 361 SO Response Aggregators has declined on a lasting basis. This is also illustrated by Graphs 27 to 30.

(2)   In the Netherlands and Italy, generic search traffic from Google's  general search results pages to the 361 SO Response Aggregators remained broadly stable. This is also illustrated by Graphs 31 and 32.

(3)   In Denmark and Poland, generic search traffic from Google's general search results pages to the 361 SO Response Aggregators, apart from […] in Denmark and […] in Poland, declined on a lasting basis. This is also illustrated by  Graphs 33 to 36.

(482)    In the third place, the Commission analysed the long term evolution of total traffic to comparison shopping services in the United Kingdom, Germany, the Netherlands  and France during the reference period January 2011-December 2016.581  582  The four Member States are illustrative of the overall long term impact of the way that comparison shopping services are positioned and displayed in Google’s general search results pages because the average “triggering rate”583 in 2016 of Google's Shopping Unit was particularly high ([…]) in each of those four Member States.584

The United Kingdom

(483)    The United Kingdom was the first national market in the EEA in which Google launched the Product Universal in January 2008585 and one of the first national markets in which it launched the Shopping Unit in February 2013. The triggering  rate of the Shopping Unit in the United Kingdom has also always been higher than in other EEA countries.586

(484)    During the period 2011-2016, total traffic to the 12 competing comparison shopping services in the Commission's sample587 decreased significantly and on a lasting basis. The Commission's analysis is illustrated by Graph 53 (see recital (498)).

Germany

(485)    During the period 2011-2016, total traffic to the 9 competing comparison shopping services in the Commission's sample588 decreased significantly and on a lasting basis. The Commission's analysis is illustrated by Graph 54 (see recital (499)).

The Netherlands

(486)    During the period 2011-2016, total traffic to the 6 competing comparison shopping services in the Commission's sample589 increased until 2014. Since 2014, however, total traffic to the comparison shopping services in the sample has been decreasing on a lasting basis. The Commission's analysis is illustrated by Graph 55 (see recital (500)).

France

(487)    During the period 2011-2016, total traffic to the 8 competing comparison shopping services in the Commission's sample590 decreased significantly and on a lasting basis. The Commission's analysis is illustrated by Graph 56 (see recital (501)).

(488)    Contrary to what Google claims, the Commission's analyses above (recitals (481)  and (482) to (487)) do not “focu[s] on a handful of losers and ignor[e] winners”,591 but are based on objective and representative samples. The analysis in recital (481) includes the competing comparison shopping services from the 361 SO Response Aggregators list identified by Google in the EEA. The analysis in recitals (482) to

(487) covers a representative sample of the most important comparison shopping services in the United Kingdom, Germany, the Netherlands and France, which represent […]% of traffic from Google's general search results pages to the 361 SO Response Aggregators in the United Kingdom, […]% in Germany, […]% in the Netherlands and […]% in France.592

7.2.3.3. Impact of the Conduct on traffic to Google's own comparison shopping service

(489)    The Conduct has led to an increase in traffic to Google's comparison  shopping service on a lasting basis in each of the thirteen EEA countries in which the Conduct takes place. This is confirmed by the following.

(490)    First, before the start of the Conduct, in October 2007, Google's comparison  shopping service had been unsuccessful in gaining traffic. By the beginning of 2007, it was losing traffic at a pace of 21% year on year while Google's general search service was gaining 23% of traffic year on year.593 As a Google executive wrote on  13 March 2006: “In my opinion, Froogle isn't really a serious contender today”.594 On 27 February 2007, another Google executive wrote: “Froogle simply doesn't work”.595 A Google memo dated 27 April 2009 mentions “the generally bad reputation of Froogle”, and “[t]he distance we [Google] had to come starting with Froogle”.596

(491)    Contrary to what Google claims597, the above internal Google documents indicate that Froogle's pages were unlikely to be ranked highly in generic search results on Google's general search results pages. Google was aware of this, as confirmed by an internal Google email exchange of 4-5 April 2006 with Google's Engineering Director responsible for Froogle, which stated that “[…] Onebox result items often stink” and warned that “(1) [t]he [Froogle] pages may not get crawled without special treatment; without enough pagerank or other quality signals, the content may not get crawled. (2) If it gets crawled, the same reasons are likely to keep it from being indexed; (3) If it gets indexed, the same reasons are likely to keep it from showing up (high) in search results. […] We'd probably have to provide a lot of special treatment to this content in order to have it be crawled, indexed, and rank well”.598

(492)    Second, in November 2007, the first month following the launch of the Product Universal in the United States, traffic to Google’s comparison shopping service at least doubled.599 An internal Google note of 24 October 2008, which discussed, among other things, the impact of Product Search on google.com ad revenue, acknowledges the increase in traffic to Google's own comparison shopping service, stating that: “Traffic-wise, Product Search realizes 10MM pageviews per day, up 150 percent from 4MM in Sept. 2007. Most of this growth is from improved google.com integration”.600

(493)    Third, Google indicates on its website that paid search results in the Shopping Unit attract “[m]ore traffic and leads”, and that “[m]any businesses experience significantly higher clickthrough rates (CTR) with Product Listing Ads compared to text ads shown in the same location for Google Shopping searches”. Google goes on to  state  that:  “[i]n  some  cases,  advertisers  have  experienced  double or triple standard clickthrough rates”.601

(494)    Fourth, the Commission's comparison during certain reference periods between 2008 and 2014 of the trigger rates602 of Product Universals and Shopping Units in Google's general search results pages in the United Kingdom, Germany, France and the Netherlands with traffic from Google's general search results pages to Google Product Search and Google Shopping603 indicates that higher trigger rates of  Google's own comparison shopping service in its general search results pages are associated with increased traffic to that service. The Commission's analysis is also illustrated by Graphs 37 to 44.

(495)    Fifth, while the evolution of traffic to a website in a given country can be influenced by several factors, the Conduct is a factor that has led to a lasting increase in traffic  to Google's comparison shopping service in each of the thirteen EEA countries in which the Conduct takes place. This is confirmed by two analyses undertaken by the Commission.

(496)    In the first place, the Commission has analysed during certain reference periods between January 2008 and December 2016, the evolution of traffic from Google's general search results pages to Google's own comparison shopping service604 and to the 361 SO Response Aggregators in the United Kingdom, Germany, France,  Spain, Italy, the Netherlands, Denmark and Poland.605 That analysis indicates that traffic to Google's comparison shopping service from Google's general search results pages  has increased significantly whereas traffic from Google's general search results pages to the 361 SO Response Aggregators has decreased (in the United Kingdom, Germany, France and Spain), remained broadly stable (in Italy and the Netherlands) or increased to a more limited degree (in Denmark and Poland):

(1)   In the United Kingdom, Germany, France and Spain, traffic from Google's general search results pages to Google's own comparison shopping service increased significantly whereas traffic from Google's general search results pages to the 361 SO Response Aggregators decreased. This is also illustrated by Graphs 45 to 48.

(2)   In Italy and the Netherlands, where the Conduct has been in place for a shorter period (since May 2011) than in the United Kingdom (since January 2008), Germany (since January 2008) and France (since October 2010), traffic from Google's general search results pages to Google's own comparison shopping service increased significantly whereas traffic from Google's general search results pages to the 361 SO Response Aggregators remained broadly stable. This is also illustrated by Graphs 49 to 50.

(3)   In Denmark and Poland, where the Conduct has been in place for an even shorter period (since November 2013) than in the United Kingdom (since January 2008), Germany (since January 2008), France (since October 2010), Spain (since May 2011), Italy (since May 2011) and the Netherlands (since May 2011), traffic from Google's general search results pages to Google's own comparison shopping service increased significantly whereas traffic to the 361 SO Response Aggregators increased but to a more limited degree. This is also illustrated by Graphs 51 to 52.

(497)    In the second place, the Commission has analysed, during the reference period 2011- 2016, the evolution of traffic from Google's general search results pages to Google's own comparison shopping service606 and total traffic to representative samples of competing comparison shopping services covered by the analysis set out in recitals (482) to (487) in the United Kingdom, Germany, the Netherlands and   France.607  608 That analysis indicates that in each country, traffic to Google's comparison shopping service from Google's general search results pages increased significantly and on a lasting basis during the reference period:

(498)    In the United Kingdom, traffic to Google's comparison shopping service from Google's general search results pages increased significantly and on a lasting basis. By the end of 2016, traffic to Google's comparison shopping service from Google's general search results pages was nearly 14 times higher than the total traffic to the entire sample of the 12 competing comparison shopping services in the United Kingdom.609  This is also illustrated by Graph 53.

(499)    In Germany, traffic to Google's comparison shopping service from Google's general search results pages increased significantly and on a lasting basis. By the end of 2016, traffic to Google's comparison shopping service from Google's general search results pages was more than 2 times higher than the total traffic to the entire sample of the 9 competing comparison shopping services in Germany.610 This is also illustrated by Graph 54.

(500)    In the Netherlands, traffic to Google's comparison shopping service from Google's general search results pages increased significantly and on a lasting basis. By the end of 2016, traffic to Google's comparison shopping service from Google's general search results pages was more than 2.7 times higher than the total traffic to the entire sample of the 6 competing comparison shopping services in the Netherlands.611   This is also illustrated by Graph 55.

(501)    In France, traffic to Google's comparison shopping service from Google's general search results pages increased significantly and on a lasting basis. By the end of 2016, traffic to Google's comparison shopping service from Google's general search results pages was more than 4.7 times higher than the total traffic to the entire sample of the 8 competing comparison shopping services in France.612 This is also illustrated by Graph 56.

7.2.3.4. Google's arguments and the Commission's response

(502)    Google essentially relies on five arguments to contest the fact that the Conduct decreases traffic from Google's general search results pages to  competing comparison shopping services and increases traffic from Google's general search results pages to Google's own comparison shopping service.

(503)    First, Google claims that the Commission has defined the Conduct as being “limited to the way that Google positions and displays Product Universals and Shopping Units”613 in its general search results pages and not including the fact that Google's own comparison shopping service is unaffected by the […] and Panda algorithms that are prone to demoting competing comparison shopping services.614

(504)    Second, Google claims that the […] algorithm has not led to any decrease in generic search traffic from Google's general search results pages to competing comparison shopping services. In fact, generic search traffic from Google's general search results pages to competing comparison shopping services has increased since […].615

(505)    Third, Google claims that the Panda algorithm has not led to any decrease in generic search traffic from Google's general search results pages to competing comparison shopping services because traffic to these services has not decreased in a consistent manner in each of the thirteen EEA countries in which the Conduct takes place.616 Rather, the business model of the competing comparison shopping services and the presence of merchant platforms, in particular Amazon, in a given country are  “more plausible” causes of the decrease in generic search traffic from Google's general search results pages to competing comparison shopping services.617 This is because:

(1)   “merchant platforms' direct purchase functionality makes them stronger, not weaker, competitors, because it provides users with added value”,618 whereas the lack of this functionality “renders aggregators less, not more, attractive for users”;619

(2)   unlike comparison shopping services, merchant platforms have attracted a significant volume of direct traffic (or navigational queries),620 through their apps621 and invested “heavily” in marketing efforts;622 and

(3)   merchant platforms “offer after sale support”.623

(506)    Fourth, Google claims that the Product Universal and the Shopping Unit are unlikely to have led to any decrease in generic search traffic from Google's general search results pages to competing comparison shopping services, or that, in any event, such  a decrease would be, at most, “marginal”.624  In support of this claim, Google relies on two submissions which purport to analyse the impact of the positioning and display of the Product Universal and the Shopping Unit on generic search traffic  from Google's general search results pages to competing comparison shopping services: (i) a “Difference-in-Differences” Analysis (see recitals (479) and (519)- (521)); and (ii) an Ablation Experiment (see recitals (522)-(532)).

(507)    As mentioned in footnote 577, the “Difference-in-Differences” Analysis is a cross- country comparison analysis that compares visually the evolution of generic search traffic from Google's general search results pages to the national domains of competing comparison shopping services in pairs of countries with overlapping languages, and in which the Product Universals/Shopping Units were launched at different times.

(508)    In the Ablation Experiment, Google removed (or “ablated”) the Shopping Unit from its general search results pages for a number of users (“the ablation group”) and compared the number of clicks leading to competing comparison shopping services625  from users in the ablation group with the number of clicks from users to whom Google displayed an unchanged general search results page (“the control group”).

(509)    According to Google, the results of the Ablation Experiment indicate that:

(1)   when the Shopping Unit is removed, between […] of the Shopping Unit clicks go to AdWords results and other Google elements, whereas […] go to comparison shopping services ([…] in the United Kingdom, […] in   Germany, […] in France, and […] in the Netherlands, and similar results in the […] additional countries covered by the SSO);626 and

(2)   when the Shopping Unit is removed, the impact on generic search traffic from Google's general search results pages represents […] of all traffic of comparison shopping services (from […] in Norway to […] in Germany).627

(510)    Fifth, Google claims that the positioning and display of the Product Universal and  the Shopping Unit have not led to any increase in traffic from Google's general search results pages to its own comparison shopping service. Rather, traffic has increased to its own comparison shopping service because (i) Google applies specific relevance criteria when determining whether to display results from its comparison shopping service in its general search results pages;628 and (ii) the rich features of the Product Universal and the Shopping Unit allow users to better assess their relevance.629

(511)    For the reasons set out below, each of these claims is unfounded.

(512)    First, the Conduct consists not only of the positioning and display of the Product Universal and the Shopping Unit but also includes that, unlike competing  comparison shopping services, Google's own comparison shopping service is not prone to being demoted by the […] and Panda algorithms. This latter aspect was set out in both the SO (paragraphs 312-314) and in the SSO (paragraphs 68-69).

(513)    Second, the […] algorithm has contributed to a decrease in generic search traffic  from Google's general search results pages to competing comparison shopping services (see section 7.2.1.1.1). By contrast, Google's comparison shopping service has never been demoted by the […] algorithm.

(514)    Third, the Panda algorithm has contributed to a decrease in generic search traffic  from Google's general search results pages to competing comparison shopping services (see section 7.2.1.1.1). By contrast, Google's comparison shopping service has never been demoted by the Panda algorithm.

(515)    Fourth, there are a number of reasons why the business model of the comparison shopping services and the presence of merchant platforms are not “more plausible” causes of the decrease in generic search traffic from Google's general search results pages to competing comparison shopping services.

(516)    In the first place, merchant platforms, and in particular Amazon, existed before the introduction of Panda, at a time when traffic to comparison shopping services was increasing.630

(517)    In the second place, if the presence of merchant platforms were a “more plausible” cause of the decrease in generic search traffic from Google's general search results pages to competing comparison shopping services, such a decrease should also have affected Google's own comparison shopping service. Traffic from Google's general search results pages to Google's own comparison shopping service has, however, increased on a lasting basis in each of the thirteen EEA countries in which the Conduct takes place (see section 7.2.3.3), notwithstanding the fact that:

(1)   Google Shopping does not offer a direct purchase functionality;

(2)   Google Shopping receives little direct traffic (see recital (581));

(3)   there is no Google Shopping app, and Google “does not promote or market [Google Shopping] to users”;631 and

(4)   Google Shopping offers no after sale services.

(518)    In the third place, the three statements […]632 […]:

(1)   the probative value of these statements is limited because they are unsworn,633 made by […] that have an interest in providing information denying the existence of any infringement634 and are unclear as to the legal consequences that any false information contained in those statements might have for […];635

(2)   the content of the statements is contradicted by the body of evidence of high probative value, including traffic data […], user surveys, contemporaneous internal Google documents and responses to requests of information under Articles 18(2) and 18(3) of Regulation (EC) No 1/2003. That body of evidence demonstrates that in each of the thirteen EEA countries in which the Conduct takes place, generic search traffic from Google's general search results pages has decreased on a lasting basis to almost all competing comparison shopping services whereas traffic to Google's own comparison shopping service has increased on a lasting basis (see recital (496) and section 7.2.3.3).

(519)    Fifth, the “Difference-in-Differences” Analysis does not indicate that the Conduct was unlikely to have affected generic search traffic from Google's general search results pages to competing comparison shopping services.

(520)    In the first place, the “Difference-in-Differences” Analysis examines the impact of the Product Universal and the Shopping Unit on generic search traffic from Google's general search results pages to competing comparison shopping services; it does not take account of the application of the […] and Panda algorithms. The “Difference-in- Differences” Analysis therefore fails to analyse one aspect of the Conduct.

(521)    In the second place, in order for Google's “Difference-in-Differences” Analysis to be able to compare properly the evolution of generic search traffic from  Google's general search results pages in pairs of countries with overlapping languages, traffic in the pair countries would have had to have been evolving in the same way prior to the introduction of the Product Universal and the Shopping Unit. In other words, visually,  the  lines  depicting the  evolution  of generic  search  traffic  from Google's general search results pages in the pair countries would have to be parallel. However, this is not the case in Google's “Difference-in-Differences” Analysis. For example, in the period prior to the introduction of the Product Universal in Germany, generic search traffic from Google's general search results pages to […] in Germany and Austria was evolving differently. Similarly, in the period prior to the introduction of the Product Universal in France, generic search traffic from Google's general  search results pages to […] and […] was evolving differently in France and Belgium.636

(522)    Sixth, the Ablation Experiment does not indicate that the Conduct has led, at most, to a marginal decrease in generic search traffic from Google's general search results pages to competing comparison shopping services.

(523)    In the first place, like Google's “Difference-in-Differences” Analysis, the Ablation Experiment examines the impact of the Product Universal and the Shopping Unit on generic search traffic from Google's general search results to competing comparison shopping services; it does not take account of the application of the […] and Panda algorithms. The Ablation Experiment therefore fails to analyse one aspect of the Conduct.

(524)    In the second place, both sets of results reported in Google's Ablation Experiment (see recitals (509)(1) and (509)(2)) are based on the examination of an inflated number of queries.

(525)    The first set of results reported by Google (see recital (509)(1)) purports to examine where Shopping Unit clicks (in the control group) go when the Shopping Unit is removed (in the ablation group).

Table 20: Percentage of Shopping Unit clicks going to competing comparison shopping services when the Shopping Unit is removed

[…]637

(526)    Those percentage figures, however, ignore the fact that, due to the application of the […] and Panda algorithms, competing comparison shopping services are rarely ranked highly in generic search results in Google's general search results pages when the Shopping Unit is removed (see Table 21). As Google puts it, “[i]t is trite to observe that a result is more likely to be clicked on when it is present than when it  is not”.638 The Commission uses as a proxy for the visible appearance of a generic search result in Google's general search results pages whether it appears within the first four generic search results on the first Google general search results page. This  is because, for most product queries, there are four or fewer generic search results displayed above the fold of the first Google general search results page, given that AdWords text ads are also displayed.

Table 21: Percentage of queries where a competing comparison shopping service appears in one of the first four generic search results on the first Google general search results page when the Shopping Unit is removed

[…]639

(527)    Consequently, when the Shopping Unit is removed, Google should have benchmarked the percentage of Shopping Unit clicks going to competing comparison shopping services only against the percentage of queries for which at least one competing comparison shopping service appears in one of the first four generic search results on the first Google general search result page.

(528)    When the first set of results reported in Google's Ablation Experiment is calculated based on the correct subset of queries, the results indicate that, when the Shopping Unit is removed, users often click on a competing comparison shopping service when it appears in the first four generic search results on the first Google general search result page and when the Shopping Unit is removed.

Table 22: Percentage of users that, when the Shopping Unit is removed, click on a competing comparison shopping service when it appears in the first four generic search results on the first Google general search result page

[…]640

(529)    The second set of results reported by Google (see recital (509)(2)) purports to examine the change in clicks on competing comparison shopping services in the ablation and the control group, as a percentage of clicks on competing comparison shopping services in the ablation group. Google calculated these percentages based on all queries, including non-product-related queries and product-related queries in which the Shopping Unit was not displayed (in the control group) or would not have been displayed (in the ablation group).

(530)    This, however, ignores the fact that the Shopping Unit is displayed only in response to a limited subset of all queries. According to information provided by Google relating to the Product Universal, product queries (i.e. those queries in which it is more likely that the Shopping Unit is triggered) accounted for around […] of total queries that Google received. Furthermore, out of the addressable universe of product queries, Google showed Product Universals for a maximum of […] of product queries.641

(531)    Consequently, contrary to what Google claims,642 it should have benchmarked the change in clicks on competing comparison shopping services when the Shopping Unit is removed only against the queries for which the Shopping Unit is displayed (in the control group) or would have been displayed (in the ablation group).

(532)    When the second set of results reported in Google's Ablation Experiment is calculated based on the correct subset of queries, the results indicate that the positioning and display of the Product Universal and the Shopping Unit has led to a non-negligible decrease in generic search traffic from Google's general search results pages to competing comparison shopping services. This is illustrated by Table 23.

Table 23: Impact of the display of the Shopping Unit on generic search traffic from Google's general search results pages to competing comparison shopping services

[…]643

(533)    Seventh, and in any event, whereas the Conduct both decreases traffic from Google's general results pages to competing comparison shopping services and  increases traffic from Google's general search results pages to Google's own comparison shopping service (see section 7.2), the “Difference-in-Differences” Analysis and the Ablation Experiment assess only the impact of the positioning and display of the Product Universal and the Shopping Unit on generic search traffic from Google's general search results pages to competing comparison shopping services. Consequently, even if the “Difference-in-Differences” Analysis or the Ablation Experiment did indicate that the positioning and display of the Product Universal and the Shopping Unit has not led to a decrease in generic search traffic from Google's general search results pages to competing comparison shopping services, this would not alter the fact that the Conduct has led to an increase in traffic from Google's general search results pages to Google's comparison shopping service.

(534)    Eighth, contrary to what Google claims, it is not more likely that “users click on Product Universals and Shopping Units because users accurately assess that these results are what they are looking for”644 and that “clicks reflect user preferences”.645

(535)    In the first place, users tend to consider that search results that are ranked highly in generic search results on Google’s general search results pages are the most relevant for their queries and click on them irrespective of whether other results would be more relevant for their queries (see section 7.2.3.1). For example, prior to Google beginning the Conduct in October 2007, Google's comparison shopping service was losing traffic at a pace of 21% year on year while Google’s general search service was gaining 23% in the same time frame.646 Google started the Conduct even though as a result, it did not always show to users the most relevant results (as ranked by its algorithms) at least for certain queries.

(536)    In the second place, Google did not inform users that the Product Universal was positioned in its general search results pages using different underlying mechanisms than those used to rank generic search results. As for the Shopping Unit, while the “Sponsored” label may suggest that different positioning mechanisms are used, that information is likely to be understandable only by the most knowledgeable users.

(537)    Ninth, it is irrelevant whether traffic has increased to Google's own comparison shopping service because Google applies specific relevance criteria when determining whether to display results from its comparison shopping service in its general search results pages. This is because the Commission does not object to Google applying specific criteria per se but to the fact that Google prominently positions and displays results only from its own comparison shopping service and not from competing comparison shopping services.

(538)    Tenth, it is irrelevant whether traffic has increased to Google's own comparison shopping service because the rich features of the Product Universal and the Shopping Unit allow users to better assess their relevance. This is because the Commission does not object to Google applying rich features to certain results but to the fact that Google applies such rich features only to its own comparison shopping service and not to competing comparison shopping services.

7.2.4. The traffic diverted by the Conduct accounts for a large proportion of traffic to competing comparison shopping services and cannot be effectively replaced by other sources currently available to comparison shopping services

(539)    The Commission concludes that the traffic diverted by the Conduct, i.e. generic search traffic from Google's general search results pages, accounts for a large proportion of traffic to competing comparison shopping services (section 7.2.4.1) and cannot be effectively replaced by other sources of traffic currently available to comparison shopping services including AdWords, mobile applications, direct  traffic, referrals from partner websites, social network sites and other general search engines (section 7.2.4.2).

(546) In the first place, certain users value the perceived editorial integrity of  Google's  generic search results services and look for them as a matter of priority in Google general search results pages.661 Conversely, in the case of AdWords results, the highest bidders in the auction are presented and these bidders for advertisement  space control the message associated with the results, which may appeal to other users.

(547) In the second place, AdWords results are positioned separately from generic search results on Google’s general search results pages. Over time, Google has used  different means to position and display AdWords results, such as coloured shading around advertisements and a small “Ad” label next to paid links. Many users are aware that part of the space on Google's general search results pages is allocated to AdWords results. If a user is routinely disappointed by what they find when they  click on AdWords results, they may change their future behaviour and look for generic search results instead.662

(548) In the third place, heat-map experiments and click-through-rate  (“CTR”)  studies indicate that users look more frequently at, or click on, generic search results than AdWords results, even if the AdWords results appear above the generic search results.663  Rather than considering the top left-hand result, users have learned to look on Google’s general search results pages for generic search results to avoid clicking on AdWords results. Graph 57 and Graph 58 indicate the number of times, since the introduction of AdWords results in July 2007, an AdWords result was presented on Google’s general search results pages and the CTR on that result, depending on whether it was presented on the right or at the top of the generic search results in the EEA. […]664 This is consistent with users learning over time that AdWords results are advertisements and not generic search results.

Graph 57: CTR for RHS ads in the EEA

[…]

Source: Google's response to […] the Commission's request for information […].

Graph 58: CTR for top ads in the EEA

[…]

Source: Google's response to […] the Commission's request for information […].

(549)          […],665 […].

Graph 59: Austria – CTR by result type and impression position

[…]

Graph 60: Belgium – CTR by result type and impression position

[…]

Graph 61: Czech Republic – CTR by result type and impression position

[…]

Graph 62: Germany – CTR by result type and impression position

[…]

Graph 63: Denmark – CTR by result type and impression position

[…]

Graph 64: Spain – CTR by result type and impression position

[…]

Graph 65: France – CTR by result type and impression position

[…]

Graph 66: Italy – CTR by result type and impression position

[…]

Graph 67: Netherlands – CTR by result type and impression position

[…]

Graph 68: Norway – CTR by result type and impression position

[…]

Graph 69: Poland – CTR by result type and impression position

[…]

Graph 70: Sweden – CTR by result type and impression position

[…]

Graph 71: United Kingdom – CTR by result type and impression position

[…]

(550)    Third, competing comparison shopping services that responded to the Commission's requests for information indicated that generic search traffic from Google’s general search results pages and traffic from AdWords are imperfect substitutes. Out of the nine comparison shopping services that responded,666 six answered that they do not regard generic search traffic from Google’s general search results pages and traffic from AdWords as substitutes. Some of these respondents even specified that they

consider generic search traffic from Google’s general search results pages and traffic from AdWords as two complementary sources of traffic. Another respondent explained that generic search traffic from Google’s general search results pages and traffic from AdWords are only interchangeable to a limited extent because different users are addressed.667 That respondent also stated that it would not stop advertising on AdWords if its site was ranked as the top generic search result as it aims to address different user types through online search advertising.

(551)    […] argues that it is impossible for comparison shopping websites to compensate decreases in generic search traffic from Google’s general search results pages with traffic from AdWords. It views them as “separate distribution channels” and argues that traffic from AdWords cannot constitute a substitute to generic search traffic  from Google’s general search results pages. “Comparison shopping sites use ads to gain a larger audience which in turn makes their platform more attractive for merchants. However, commercial ads are only seen as a complementary tool to attract users. They cannot substitute (lost) traffic coming via natural search”.668

(552)    Fourth, Google's claim that […]669 […] is misleading for three reasons.

(553)    In the first place, only 5 of these […] had more than 1 million unique visitors based on ComScore data from October 2014, while roughly three quarters of them had less than 100 000 visitors or did not appear at all on the same ComScore list.670  Moreover, roughly a quarter of those […] were identified by […] as not working or having such a residual audience that they could be considered as no longer functional.671

(554)    In the second place, almost all of those […] are small. In addition, […] generally represents only a limited share of comparison shopping services' traffic. This is confirmed by the Commission's analysis of the Ablation Experiment672 […]:

Graph 72: Share of clicks by result type

[…]

Graph 73: Share of clicks by result type

[…]

Graph 74: Share of clicks by result type

[…]

Graph 75: Share of clicks by result type

[…]

(555)    Fifth, the impact of increased spending and reliance on traffic from AdWords is, in any event, limited because the Conduct also reduces the effectiveness of AdWords results. This is supported by the Commission's analysis of the Ablation Experiment, […].673 […]

Table 26: Impact of the display of the Shopping Unit on AdWords traffic to comparison shopping services

[…]674

(556)    Sixth, comparison shopping services have not started to spend significantly more on attracting traffic from AdWords. Graph 76 illustrates the yearly evolution between 2011 and 2014 of total traffic, generic search traffic from Google’s general search results pages, traffic from AdWords and the corresponding yearly spend on traffic from AdWords for five comparison shopping services ([…]).

(557)    Of the five competing comparison shopping services depicted in Graph 76, four of them ([…]) did not make any significant attempt to replace generic search traffic from Google's general search results pages with traffic from AdWords, as illustrated by the fact that spending on Google and the resulting paid traffic was stable or declining. This means that these four companies: (i) did not consider traffic from AdWords to be a substitute; (ii) were already spending optimally on AdWords;677 or (iii) were financially constrained. In all three situations, traffic from AdWords could not effectively replace generic search traffic from Google's general search results pages.

(558)    Only […] has sought to replace generic search traffic from Google's general search results pages with increased spending on traffic from AdWords. That increased spending has, however, not compensated for the decreased generic search traffic from Google's general search results pages. As indicated in Graph 76, between 2011 and 2014, […] experienced a decrease in generic search traffic from  Google’s general search results pages of approximately 87 million visits whereas the increase in traffic from AdWords equalled approximately 34 million visits. In four years, […] therefore managed to compensate only 40% of the decreased generic search traffic from Google's general search results pages. Finally, in October 2015, […] stopped purchasing AdWords traffic altogether,  in consideration of the fact  that AdWords had gradually become an unviable option from an economic point of view.678

(559)    Seventh, and in any event, even if competing comparison shopping services could recover generic search traffic from Google's general search results pages with traffic from AdWords, it would not be an economically viable solution for the following reasons.

(560)    In the first place, the costs faced by comparison shopping services are significantly higher for generating traffic from AdWords than generic search traffic from Google’s general search results pages. Comparison shopping services paid between EUR 0.03 and EUR 0.28 per visit for traffic from AdWords in 2014679 whereas the cost of generic search traffic from Google’s general search results pages (the spending on Search Engine Optimisation) was less than EUR 0.015 per visit in 2014.680 The costs of AdWords traffic are therefore, even in the most conservative scenario, double the costs for generic traffic.

(561)    The large cost difference cannot be explained by traffic coming from AdWords being of higher quality than generic search traffic from Google’s general search results pages. In fact, as indicated in Table 27, for some comparison shopping services (e.g. […]), the conversion rate of traffic from AdWords is even lower than the conversion

the cost of AdWords, as they bid for the same keywords that comparison shopping services would seek to increase their own traffic.

(563) The low overall profitability for traffic from AdWords compared to generic search traffic from Google's general search results pages is further confirmed by comparison shopping services.

(1)   […] indicated that: “First, the profitability of AdWords traffic (and paid traffic in general) is not guaranteed. […] Second, even in instances where tight SEM optimization leads AdWords traffic to be marginally profitable to a comparison shopping website, its profitability would by essence be much reduced  compared to the margin of SEO traffic (which has no acquisition cost). […] Third, the marginal profitability (if any) of AdWords traffic for advertising websites tends to decline as the volumes of such traffic increase”.684

(2)   […] indicated that “Due to the fact that Google progressively replaced traditional text ads [AdWords results] with “Product Listing Ads (PLA)” for the most economically attractive search queries (in order to cut out the “middlemen” like us), and price comparison services are not allowed to participate in the PLA program, it was no longer possible for us to profitably buy  traffic on  Google. As  a  result  of this,  the use  of  Google’s  paid search [AdWords] service had to be stopped for: […]”.685

(3)   […] indicated that “due to the increasingly dominant position of Google Shopping results instead of regular AdWords ads, the effect of the investment has been nullified. Instead of increasing site-traffic, […] now faces fast and significantly decreasing traffic through the Google AdWords campaign”.686

(564) In the second place, whereas Google Shopping does not have to bear any costs associated with acquiring traffic from generic search results or AdWords results on Google's general search results pages, competing comparison shopping services not only have to bear such costs, but also, in the case of AdWords traffic, have to pay Google to acquire such traffic:

(1)   […] indicated that, while “Google […] argues that concerned sites can undo any harm from the tie by buying advertising, namely AdWords; […] such an option entails payment to Google thus further exacerbating the asymmetry between Google and its competitors”.687

(2)   […] indicated: “Buying traffic via AdWords is not mainly building-up brand awareness (especially in […]) and buying traffic is simply buying traffic, i.e. you need vast financial resources for building-up a consistent, long-term  online business model based on pushing your traffic via AdWords. Furthermore, Google […] proposes AdWords (i.e. Google AdWords) without mentioning that the major benefiter would be Google itself, charging monopolistic AdWords prices without any costs on their side. Using this as a

AdWords (relative to generic search traffic from Google’s general search results pages) achieve only a limited profit margin on that channel. For instance, out of the total revenue they generated from traffic from AdWords purchased between 2011 and 2014, the price paid to Google by […] was more than 40% of that revenue. For […], it was between [75% and 110%] of that revenue.692

(567)    In the fifth place, Google's claim that the higher costs of purchasing AdWords traffic do not make AdWords traffic an unviable alternative to generic search traffic from Google's general search results pages, as all businesses would prefer a free promotion opportunity,693  ignores that:

(1)                            by using the AdWords channel, comparison shopping services would pay higher costs for traffic that is of lower commercial value than traffic from the Shopping Unit;

(2)                            Google itself lists “comparison shopping sites” among the types of websites that are likely to merit low landing page quality scores and that “may be difficult to advertise affordably” in AdWords.694

7.2.4.2.2. Mobile apps

(568)    Mobile applications (apps) are not a viable alternative for replacing generic search traffic from Google's general search results pages for the following reasons.

(569)    First, while developing a mobile app may be a less expensive alternative, as Google claims,695 the download of a given app already presumes strong brand awareness, which comparison shopping services would have to build before deploying such application, […].696

(570)    Second, contrary to what Google claims, the mobile channel is a minor source of traffic for major comparison shopping services.

(571)    In the first place, as indicated in Table 24, between 2011 and 2016, direct traffic together with traffic from mobile apps represented, in most cases, less than 20% of all traffic to comparison shopping services. A significant increase in the volume of traffic from mobile apps would therefore be necessary to replace effectively generic search traffic from Google’s general search results pages.

(572)    In the second place, as can be seen […],697 […].

(573)    In the third place, the volume of mobile app traffic represents only a small proportion of the total traffic to comparison shopping services during the period 2014-2016, as indicated in Table 28.

(577)    In the first place, comparison shopping services and merchant platforms are not in the same market (see section 5.2.2.4).

(578)    In the second place, […]705 […].

(579)    Sixth, even achieving some increase in traffic from Google's mobile general search results pages seems to be difficult, because of the limited space on the screen of mobile devices, which means that comparison shopping services will almost never appear above the fold, especially in those instances when Google's Shopping Unit is displayed. This is confirmed by […], the owner of […], who indicated that  “Google’s layout of its mobile Search Engine Results Page (SERP) has left less and less space above the fold for natural search results. Simultaneously, Google has displayed in an anti-competitive manner its own comparison shopping service (Google Shopping) prominently. This has deprived competing services […] from a portion of natural mobile traffic growth”.706

7.2.4.2.3. Direct traffic

(580)    Direct traffic is not a viable alternative for replacing generic search traffic from Google's general search results pages, for the following reasons:

(581)    First, for most comparison shopping services, direct traffic represents a small percentage of their total traffic as can be seen from Table 24 and further detailed in recital (571). This also holds for Google Shopping as Google estimates that direct traffic to Google Shopping amounts to […].707 Thus, while it is true, as Google claims,708 that marketing and promotions can help increase traffic to comparison shopping services, they are only a complementary source of traffic and cannot effectively replace generic search traffic from Google's general search results pages.

(582)    Second, even those comparison shopping services that sought to increase  direct traffic through offline and non-search advertising were not successful.

(1)  […] launched significant press initiatives between 2008 and 2010 to promote  its brand, but considered that such efforts based on offline marketing are not a feasible alternative to generic search traffic from Google's general search results.709

(2) Despite […] efforts to increase traffic through a brand-awareness campaign, this initiative was rather a strategy used to mitigate the decrease in traffic: “Without the brand awareness campaign, the decrease in site traffic would probably have been even more severe”.710

(583)    Third, in order to increase direct traffic, a comparison shopping service would need a strong brand. However, as confirmed by comparison shopping services, building a strong brand entails the investment of significant financial resources over a long period of time, without any guarantees of success. For example, as indicated by […], building brand awareness would cost “generally tens of millions of euros over a long time period. For small and mid-sized companies, it is virtually impossible to conduct an expensive and time-consuming brand campaign while also focusing extensively on providing an excellent quality for users. These companies simply do not have the

resources for both of these undertakings”.711 Therefore, in the case of comparison shopping services, marketing through offline advertising is a low value-for-money alternative.

7.2.4.2.4. Other sources

(584)    None of the other sources of traffic reported by competing comparison shopping services could effectively replace generic search traffic from Google's general search results pages.

(585)    First, referrals of traffic to competing comparison shopping services from partner websites (which generate traffic on an affiliate basis) and newsletters would also increase traffic acquisition costs to a level that would prevent them from viably competing against Google Shopping. For instance, […] argues that “SEO traffic from Google (i.e. free traffic from natural search results) has been one of the highest quality traffic sources for our business. Hence, with the loss of such large quantities of SEO traffic in our business, our overall traffic mix has suffered significantly. We have attempted to replace some of this high quality traffic loss from Google with other sources of traffic, but the available sources of traffic typically have a much lower ‘quality’ score. As this traffic mix has changed over time (more publisher traffic, less SEO and SEM traffic) this has led to our overall RPL (revenue per lead) weakening over time. This has lowered our overall revenue and profit as a business”.712

(586)    Second, while traffic to competing comparison shopping services from social network sites and other general search services would not lead to significantly higher traffic acquisition costs, neither source of traffic could effectively replace generic search traffic from Google's general search results pages.

(587)    In the first place, traffic from social network sites is limited to situations where, for instance, on Facebook, a user suggests a link to a comparison shopping site to a friend.

(588)    In the second place, traffic from other general search services (such as Bing or Yahoo) is insignificant and unlikely to increase due to the barriers to entry to the national markets for general search services (see section 6.2.2).

7.3. The Conduct has potential anti-competitive effects on several markets

(589)    The Commission concludes that the Conduct is capable of having, or is likely to  have,  anti-competitive effects in the national markets for comparison shopping services (section 7.3.1) and in the national markets for general search services (section 7.3.3).

(590)    Moreover, even if the alternative product market definition proposed by Google, comprising both comparison shopping services and merchant platforms, were to be followed, the Conduct would be capable of having, or likely to have, anti- competitive effects in at least the comparison shopping services segments of possible national markets comprising both comparison shopping services and merchant platforms (section 7.3.2).

7.3.1.    Potential anti-competitive effects in the national markets for comparison shopping services

(591)    In section 7.2.3 the Commission concluded that the Conduct decreases traffic from Google's general results pages to competing comparison shopping services and increases traffic from Google's general search results pages to Google's own comparison shopping service. In section 7.2.4 the Commission concluded that generic search traffic from Google's general search results pages accounts for a large proportion of traffic to competing comparison shopping services and cannot be effectively replaced by other sources currently available to comparison shopping services.

(592)    In light of the above (recital (591), and also for the reasons outlined in this section (recitals (593)-(607)), the Conduct is capable of having, or likely to have, anti- competitive effects, in the national markets for comparison shopping services.

(593)    First, the Conduct has the potential to foreclose competing comparison shopping services, which may lead to higher fees for merchants, higher prices for consumers, and less innovation for the reasons explained below.

(594)    In the first place, the Conduct is capable of leading competing comparison shopping services to cease providing their services. This would allow Google to impose and maintain higher fees on merchants for participation in its own comparison shopping service. These higher costs for merchants are capable of leading to higher product prices for consumers.

(595)    In the second place, the Conduct is likely to reduce the incentives of competing comparison shopping services to innovate. Competing comparison shopping services will have an incentive to invest in developing innovative services, improving the relevance of their existing services and creating new types of services, only if they can reasonably expect that their services will be able to attract a sufficient volume of user traffic to compete with Google's comparison shopping service. Moreover, even if competing comparison shopping services may try to compensate to some extent  the decrease in traffic by relying more on paid sources of traffic, this will also reduce the revenue available to invest in developing innovative services, improving the relevance of their existing services and creating new types of services.713

(596)    In the third place, the Conduct is likely to reduce the incentives of Google to improve the quality of its comparison shopping service as it does not currently need  to compete on the merits with competing comparison shopping services.

(597)    Second, the Conduct is likely to reduce the ability of consumers to access the most relevant comparison shopping services. This is for two reasons.

(598)    In the first place, as explained in section 7.2.3.1, users tend to consider that search results that are ranked highly in generic search results on Google’s general search results pages are the most relevant for their queries and click on them irrespective of whether other results would be more relevant for their queries. For example, prior to Google beginning the Conduct in October 2007, Google's comparison shopping service was losing traffic at a pace of 21% year on year while Google’s general search service was gaining 23% in the same time frame714 (see also recital (535)). Google started the Conduct even though as a result, it did not always show to users the most relevant results (as ranked by its generic search algorithms) at least for certain queries.

(599)    In the second place, Google did not inform users that the Product Universal was positioned and displayed in its general search results pages using different underlying mechanisms than those used to rank generic search results. As for the Shopping Unit, while the “Sponsored” label may suggest that different positioning and display mechanisms are used, that information is likely to be understandable only by the  most knowledgeable users (see also recital (536)).

(600)    The Conduct therefore risks undermining the competitive structure of the national markets for comparison shopping services. The prospects of commercial success of Google’s comparison shopping service are enhanced not because of the merits of that service, but because Google applies different underlying mechanisms on the basis of the advantages provided to it by its dominant position in the national markets for general search services.715

(601)    Third, the Commission's conclusions are not called into question by Google's claims that (i) despite the duration of the Conduct, the Commission has not identified any competing comparison shopping service that has  ceased to offer its service;716  and (ii) it has identified “hundreds of aggregators” from the list of the 361 SO Response Aggregators that remain active.717

(602)    In the first place, the Commission is not required to prove that the Conduct has the actual effect of leading certain competing comparison shopping services to cease offering their services. Rather, it is sufficient for the Commission to demonstrate that the Conduct is capable of having, or likely to have, such an effect.718

(603)    In the second place, the absolute number of comparison shopping services  that remain active is irrelevant. In the absence of the Conduct, the number of comparison shopping services that would remain active and their ability to compete might be  even greater.719

(604)    In the third place, a large percentage of the competing comparison services included in Google's list of 361 SO Response Aggregators have never offered, or no longer offer, comparison shopping services:

(1)                            As indicated by […], the majority of the 361 SO Response Aggregators do not provide comparison shopping services but instead arbitrate traffic720 between comparison shopping services and paid search traffic, and approximately 38% of the 361 SO Response Aggregators are no longer active.721

(2)                            As indicated by […], certain of the 361 SO Response Aggregators (e.g. 2good.info, checkoutspy.com, smartdeal.co.uk) do not provide comparison shopping services but instead redistribute traffic from other established comparison shopping services or merchant platforms.722

(3)                            As indicated by […], approximately 21% of the 361 SO Response Aggregators are inactive, and approximately 37% receive traffic that is too low to be sustainable.723

(4)                            As indicated by […], it provides only product reviews and no comparison shopping services.724

(5)                            As indicated by […], it is not a comparison shopping service itself as it relies on data from […] comparison shopping service and […] is included in the list of 361 SO Response Aggregators.725

(605)    Fourth, the Commission's conclusion is not called into question by Google's claim that the Commission has failed to demonstrate, as part of its analysis of anti- competitive effects, a causal link between, on the one hand, the Conduct and the decrease in traffic to competing comparison shopping services and, on the other hand, the Conduct and the increase in traffic to Google's comparison shopping service.726

(606)    In the first place, the Commission is not required to prove that the Conduct has the actual effect of decreasing traffic to competing comparison shopping services and increasing traffic to Google's comparison shopping service. Rather, it is sufficient for the Commission to demonstrate that the Conduct is capable of having, or likely to have, such effects.727

(607)    In the second place, and in any event, the Commission has demonstrated by “tangible evidence”728 that the Conduct decreases traffic to competing comparison shopping services (section 7.2.3.2) and increases traffic to Google's own comparison shopping service (section 7.2.3.3).

7.3.2.    Potential anti-competitive effects of the Conduct in possible national markets for comparison shopping services and merchant platforms

(608)    For the reasons set out in sections 5.2.2 and 7.3.1 above, the Commission concludes that comparison shopping services constitute a distinct relevant product market and that the Conduct is capable of having, or likely to have, anti-competitive effects in  the national markets for comparison shopping services.

(609)    Moreover, even if the alternative product market definition proposed by Google, comprising both comparison shopping services and merchant platforms, were to be followed, the Commission concludes that the Conduct would be capable of having,  or likely to have, anti-competitive effects in at least the comparison shopping services segments of the possible national markets comprising both comparison shopping services and merchant platforms.729

(610)    First, contrary to what Google claims,730 competing comparison shopping services, and not merchant platforms, would be the closest competitors to Google's own comparison shopping service on the possible national markets for comparison shopping services and merchant platforms. This is because, as indicated in section 5.2.2.4:

(a)              Google Shopping is a comparison shopping service;

(b)             Google distinguishes comparison shopping services (including its own comparison shopping service) from merchant platforms;

(c)              Comparison shopping services and merchant platforms are (and consider themselves as) business partners in a vertical relationship;

(d)             Comparison shopping services and merchant platforms serve different purposes from the perspective of users; and

(e)              Comparison shopping services and merchant platforms provide different services to online retailers.

(611)    Second, the Commission's conclusion is not called into question by Google's claim731 that, during the relevant period, certain merchant platforms have gained traffic from Google's general search results pages both in the absolute and relative to comparison shopping services. It would be unsurprising if merchant platforms have gained traffic as a result of the Conduct because:

(a)   merchant platforms are not prone to being demoted by the […] and Panda algorithms;732 and

(b)   the majority of merchant platforms are eligible to be displayed in the Product Universals/Shopping Units and are in fact among Google Shopping's top customers (see section 5.2.2.4). Therefore, to the extent that their offers have been displayed in the Product Universals and/or the Shopping Units, traffic to merchant platforms have benefited from the Conduct.

(612)    Third, in reaching the conclusions set out in recitals (609)-(611), the Commission has undertaken two analyses relating to the United Kingdom, Germany, France, the Netherlands, Austria, Belgium, the Czech Republic, Denmark, Italy, Norway, Poland, Spain and Sweden, as well as to these thirteen countries in which  the Conduct takes place in the aggregate, based on different sets of data and methodologies covering the period 2011-2016 (the “First Analysis” and the “Second Analysis”). The purpose of these two analyses is twofold: (i) to estimate the size of the comparison shopping services segments (including Google's own comparison shopping service and the competing comparison shopping services) of the possible national markets comprising both comparison shopping services and merchant platforms (i.e. the segments affected by the Conduct); and (ii) to analyse the evolution of traffic to Google's own comparison shopping service, the 361 SO Response Aggregators and a representative sample of merchant platforms.

(613)    Both analyses take into account the approximately 380 services that Google has identified as competing with Google Shopping, namely:

(a)              the following comparison shopping services:

(1)   the sample of the most important comparison shopping services in the EEA included in the analysis in Section 7.2.3.2. That sample includes: […];

(2)   the other “aggregators” listed by Google in Annex 3 to the SO Response (hereinafter, the Commission will refer to all SO Response Aggregators other than the comparison shopping services listed at point (1) as “the Other SO Response Aggregators”;

(3)   the product listing units displayed on the general search results pages of […];733

(4)   the product listing units displayed on third party websites by […], when these units are displayed in reply to a query;734 and

(b)             the following samples of merchant platforms: […];735 […];736 […].737

Data and methodologies governing the analyses

(614)    In both the analyses, traffic to Google's own comparison shopping service is based on:

(a)              For Google Product Search, the sum of the clicks on links that led the user to the standalone Google Product Search website, including the Shopping menu link displayed at the top of Google's general search results pages;738

(b)             For Google Shopping, the sum of the clicks on links that lead the user to the standalone Google Shopping website, including the Shopping menu link displayed at the top of Google's general search results pages, and the clicks on links that lead the user directly to a webpage of a merchant (see recital (421)).739

(615)    Traffic to competing comparison shopping services and merchant platforms is based on different sets of data and methodologies, depending on the analysis:

(a)   For the First Analysis, on data on total traffic provided by comparison  shopping services740 and merchant platforms.741

The First Analysis is relevant for the assessment of whether the Conduct would be capable of having, or likely to have, anti-competitive effects in the possible national markets for comparison shopping services and merchant platforms because it assesses the impact of the Conduct taking into account all the possible sources of traffic available to the services active in these possible markets.

(b)   For the Second Analysis, data provided by Google concerning only traffic from Google's general search results pages to all services taken into account in the First Analysis.742

The Second Analysis is relevant for the assessment of whether the Conduct would be capable of having, or likely to have, anti-competitive effects in the possible national markets for comparison shopping services and merchant platforms because it focusses on users that reach these websites from Google's general search results pages.

(616)    In the absence of existing comprehensive databases and/or surveys, both analyses are also based on the Commission's own compilation of data from the approximately 380 services identified by Google as competing with Google Shopping.

(617)    Moreover, regarding the First Analysis, as not all data was provided homogeneously or in the format and with the granularity requested, the Commission has had to make certain methodological decisions concerning, in particular: (i) the geographic allocation of traffic for certain services;743  and (ii) certain services for which  figures on total traffic were unavailable (“Missing Figures”). In particular, for the latter, the Commission has used as a proxy for total traffic the data provided by Google regarding traffic from its general search results pages (“Commission's Treatment of Missing Figures”). This concerns the following services: (i) the Other SO Response Aggregators;744 (ii) the comparison shopping service […];745 (iii) the merchant platforms […]746; and (iv) certain other comparison shopping services and merchant platforms that did not provide data throughout the period 2011-2016.

(618)    Moreover, in both analyses, the Commission has also made certain adjustments to   the    data    regarding traffic to  merchant platforms(“Merchant Platforms Adjustments”). These adjustments reflect the fact that substitutability between merchant platforms and comparison shopping services, if any, is limited to the webpages of these platforms that users visit with a view to comparing offers for new products from several professional online retailers, as an alternative to the services provided by comparison shopping services (see section 5.2.2.4). As a proxy for such limited substitutability, the Commission has therefore taken into account only traffic to the webpages of merchant platforms that display offers from: (i) different sellers for a specific item or product category, e.g. in response to a query for a “iPhone 7” or “running shoes” (“multiple offers webpages”);747 and (ii) professional online retailers for new products only.748 By contrast the Commission has not taken into account traffic to webpages of merchant platforms that display offers from: (i) a merchant platform's own inventory or from a single seller (“single offer webpages”);749 and (ii) private individuals selling goods or services on a non-professional basis.750

Google's arguments relating to the analyses and the Commission's response

(619)    Google essentially relies on three arguments to contest the analyses, the first two relating to the First Analysis, the third relating to the Second Analysis.

(620)    First, Google claims that the First Analysis is vitiated by certain calculation errors concerning data from a number of services.751

(621)    Second, Google claims that the First Analysis suffers from four methodological flaws.

(622)    In the first place, clicks on links within the Shopping Units that lead the user directly to a webpage of a merchant should not be counted as visits to Google Shopping because they are ads and clicks on them would not benefit Google Shopping.

(623)    In the second place, each individual click on a link within the Shopping Units cannot be counted as a separate visit to Google Shopping because they may be “multiple clicks” for a single query, such as whenever a user, having clicked on a link within the Shopping Unit, goes back to Google's general search results pages and clicks on another link within the Shopping Unit.752

(624)    In the third place, the Merchant Platforms Adjustments applied by the Commission are inappropriate and inconsistent with the explanations or data provided by the relevant merchant platforms.753

(625)    In the fourth place, the Commission's Treatment of Missing Figures understates the visits to the relevant services for which total traffic data is unavailable because it ignores traffic that these services obtain from other possible sources.754 The figures on traffic from Google's general search results pages should therefore be adjusted further, based on a number of different assumptions (“Google's Adjustments for Missing Figures”).755

(626)    Third, Google claims that the Second Analysis is irrelevant for the assessment of whether the Conduct would be capable of having, or likely to have, anti-competitive effects in the possible national markets for comparison shopping services and merchant platforms because it ignores traffic from sources other than Google's general search results pages.756

(627)    Each of these claims is unfounded.

(628)    First, while the First Analysis as set out in the LoF contained clerical errors relating to data from fifteen services757 (due, inter alia, to omitting or double counting data for certain websites/months due to inadvertent code errors), those errors did not have any material impact on the results of that analysis and the Commission has corrected those errors in the Decision (see Annex 1) by undertaking the First Analysis based on the dataset compiled by Google during the data room held after the LoF.758

(629)    Second, the First Analysis does not suffer from any of the four methodological flaws claimed by Google.

(630)    In the first place, as noted above (see recital (421)), clicks on links within the Shopping Units that lead the user directly to a webpage of a merchant should be counted as visits to Google Shopping because Google's comparison shopping service benefits economically from clicks on those links in the same manner as if the user had taken the intermediary step of going through the standalone Google Shopping website before clicking on the product of that merchant partner.

(631)    In the second place, each individual click on a link within the Shopping Unit should be counted as a separate visit to Google Shopping. This is because any subsequent clicks on another link within the Shopping Unit after a user has clicked on a link within the Shopping Unit and gone back to Google's general search results pages is influenced by the Conduct.

(632)    In the third place, as explained in recital (618), the Merchant Platforms Adjustments applied by the Commission are appropriate and consistent with the explanations and data provided by merchant platforms that their primary purpose is to sell the products of their merchants, not to provide a comparison among them.759 Moreover, if anything, even after the application of the Merchant Platforms Adjustments, the relevant traffic to these merchant platforms is still overestimated. This is because the Merchant Platforms Adjustments do not exclude traffic from users which, despite visiting merchant platforms with the aim of buying rather than comparing products, are shown webpages with alternative offers from third-party sellers along the retail offers of merchant platforms.

(633)    In any event, and for the sake of completeness, the Commission provides below (as was already the case in the LoF), the results of both the First Analysis and the  Second Analysis without the Merchant Platforms Adjustments. Those results are consistent with those of the corresponding analyses with the Merchant Platforms Adjustments.

(634)    In the fourth place, the Commission's Treatment of Missing Figures underestimates traffic that the services for which total traffic data is unavailable obtain from other possible sources whereas Google's Adjustments for Missing Figures  overestimate that traffic.

(1) The Commission's Treatment of Missing Figures underestimates the traffic that the services for which total traffic data is unavailable obtain from other  possible sources because it does not include traffic that these services obtain from other possible sources.

(2) Google's Adjustments for Missing Figures overestimate traffic that the services for which total traffic data is unavailable obtain from other possible sources. This is because when estimating total traffic to those services (e.g. Other SO Response Aggregators), Google applied a fixed adjustment ratio calculated by reference to direct traffic versus traffic from Google's general search results pages to services belonging to the same category for which data on total traffic is available (e.g. comparison shopping services in the case of the Other SO Response Aggregators). However, such a fixed ratio fails to reflect the fact that the services for which total traffic data is unavailable are likely to attract a lesser volume of direct traffic than services for which total traffic data is available, the latter being the most representatives ones (see recital (488)) and in general have (or had) a good brand awareness.

(635)    In any event, and for the sake of completeness, the Commission provides below the results  of  the  First  Analysis  without  Merchant  Platforms  Adjustments  and  with Google's Adjustments for Missing Figures, with the exception of the adjustments concerning the calculation of traffic to Google Shopping (see recitals (630)-(631)).

(636)    Third, the focus of the Second Analysis on users that reach comparison shopping services and merchant platforms from Google's general search results pages is precisely what makes that analysis relevant for the assessment of whether the Conduct would be capable of having, or likely to have, anti-competitive effects in the possible national markets for comparison shopping services and merchant platforms. This is because users that reach comparison shopping services and merchant platforms from Google's general search results pages are more likely to be users that have not yet made a definitive choice as to which site is the most relevant for their purpose, unlike users that go directly to the website of a  comparison  shopping service or merchant platform.

Results of the analyses

(637)    The Commission sets out below separately for the United Kingdom, Germany, France, the Netherlands, Austria, Belgium, the Czech Republic, Denmark, Italy, Norway, Poland, Spain and Sweden, as well for the thirteen countries in which the Conduct takes place the aggregate, the results of the following analyses:

(a)              First Analysis with Merchant Platforms Adjustments;

(b)             First Analysis without Merchant Platforms Adjustments;

(c)              First Analysis without Merchant Platforms Adjustments and with Google's Adjustments for Missing Figures;

(d)             Second Analysis with Merchant Platforms Adjustments; and

(e)              Second Analysis without Merchant Platforms Adjustments.760

(638)    First, the results of the analyses indicate that the Conduct would have been capable  of having, or likely to have, anti-competitive effects on a significant part of the possible market for comparison shopping services and merchant platforms (i.e. the segment including Google's comparison shopping service and the competing comparison  shopping  services)  that,  during  the  period  2011-2016,  accounted for approximately:761

●                 in the United Kingdom:

(a)              12%-24%, based on the First Analysis with Merchant Platforms Adjustments;

(b)             9%-18%,   based    on    the    First   Analysis without      Merchant     Platforms Adjustments;

(c)              11%-19%,   based   on    the   First   Analysis without        Merchant     Platforms Adjustments and with Google's Adjustments for Missing Figures;

(d)             29%-57%,   based    on    the   Second    Analysis with             Merchant     Platforms Adjustments;

(e)              20%-48%,    based    on    Second    Analysis without         Merchant      Platforms Adjustments.

●                 in Germany:

(a)              22%-28%, based on the First Analysis with Merchant Platforms Adjustments;

(b)             14%-19%,   based   on    the   First   Analysis without        Merchant     Platforms Adjustments;

(c)              15%-18%,   based   on    the   First   Analysis without        Merchant     Platforms Adjustments and with Google's Adjustments for Missing Figures;

(d)             42%-54%,   based    on    the    Second   Analysis with             Merchant     Platforms Adjustments;

(e)              30%-44%,    based    on    Second    Analysis without         Merchant      Platforms Adjustments.

●                 in France:

(a)              36%-45%, based on the First Analysis with Merchant Platforms Adjustments;

(b)             24%-30%,   based   on    the   First   Analysis without        Merchant     Platforms Adjustments;

(c)              25%-31%,   based   on    the   First   Analysis without        Merchant     Platforms Adjustments and with Google's Adjustments for Missing Figures;

(d)             45%-56%,   based    on    the   Second    Analysis with             Merchant     Platforms Adjustments;

(e)              31%-40%,    based    on    Second    Analysis without         Merchant      Platforms Adjustments.

●                 in the Netherlands:

(a)              57%-65%, based on the First Analysis with Merchant Platforms Adjustments;

(b)             54%-61%,   based   on    the   First   Analysis without        Merchant     Platforms Adjustments;

(c)              45%-53%,   based   on    the   First   Analysis without        Merchant     Platforms Adjustments and with Google's Adjustments for Missing Figures;

(d)             63%-69%,   based    on    the   Second    Analysis with             Merchant     Platforms Adjustments;

(e)              60%-69%,    based    on    Second    Analysis without         Merchant      Platforms Adjustments.

●                 in Italy:

(a)              31%-41%, based on the First Analysis with Merchant Platforms Adjustments;

(b)             24%-33%,   based   on    the   First   Analysis without        Merchant     Platforms Adjustments;

(c)              23%-33%,   based   on    the   First   Analysis without        Merchant     Platforms Adjustments and with Google's Adjustments for Missing Figures;

(d)             39%-57%,   based    on    the   Second    Analysis with             Merchant     Platforms Adjustments;

(e)              31%-46%,    based    on    Second    Analysis without         Merchant      Platforms Adjustments.

●                 in Spain:

(a)              26%-65%, based on the First Analysis with Merchant Platforms Adjustments;

(b)             20%-50%,   based   on    the   First   Analysis without        Merchant     Platforms Adjustments;

(c)              21%-49%,   based   on    the   First   Analysis without        Merchant     Platforms Adjustments and with Google's Adjustments for Missing Figures;

(d)             43%-68%,   based    on    the   Second    Analysis with             Merchant     Platforms Adjustments;

(e)              32%-52%,    based    on    Second    Analysis without         Merchant      Platforms Adjustments.

●                 in Austria:

(a)              22%-37%, based on the First Analysis with Merchant Platforms Adjustments;

(b)             18%-24%,   based   on    the   First   Analysis without        Merchant     Platforms Adjustments;

(c)              16%-28%,   based   on    the   First   Analysis without        Merchant     Platforms Adjustments and with Google's Adjustments for Missing Figures;

(d)             26%-48%,   based    on    the   Second    Analysis                    with      Merchant Platforms Adjustments;

(e)              24%-39%,    based    on    Second    Analysis                     without Merchant        Platforms Adjustments.

●                 in Belgium:

(a)              26%-44%, based on the First Analysis with Merchant Platforms Adjustments;

(b)             22%-29%,   based   on    the   First   Analysis                    without Merchant Platforms Adjustments;

(c)              21%-26%,   based   on    the   First   Analysis                    without Merchant Platforms Adjustments and with Google's Adjustments for Missing Figures;

(d)             35%-54%,   based    on    the   Second    Analysis                    with      Merchant Platforms Adjustments;

(e)              31%-41%,    based    on    Second    Analysis                     without Merchant        Platforms Adjustments.

●                 in Czech Republic:762

(a)              79%-87%, based on the First Analysis with Merchant Platforms Adjustments;

(b)             79%-87%,   based   on    the   First   Analysis without        Merchant     Platforms Adjustments;

(c)              47%-61%,   based   on    the   First   Analysis without        Merchant     Platforms Adjustments and with Google's Adjustments for Missing Figures;

(d)             73%-85%,   based    on    the   Second    Analysis with             Merchant     Platforms Adjustments;

(e)              73%-85%,    based    on    Second    Analysis without         Merchant      Platforms Adjustments.

●                 in Denmark:

(a)              64%-74%, based on the First Analysis with Merchant Platforms Adjustments;

(b)             59%-70%,   based    on    the   First   Analysis without        Merchant     Platforms Adjustments;

(c)              39%-51%,   based   on    the   First   Analysis without        Merchant     Platforms Adjustments and with Google's Adjustments for Missing Figures;

(d)             65%-74%,   based    on    the   Second    Analysis with             Merchant     Platforms Adjustments;

(e)              60%-71%,    based    on    Second    Analysis without         Merchant      Platforms Adjustments.

●                 in Norway:

(a)              51%-64%, based on the First Analysis with Merchant Platforms Adjustments;

(b)             42%-58%,   based   on    the   First   Analysis without        Merchant     Platforms Adjustments;

(c)              18%-34%,   based   on    the   First   Analysis without        Merchant     Platforms Adjustments and with Google's Adjustments for Missing Figures;

(d)             42%-62%,   based    on    the   Second    Analysis                    with      Merchant Platforms Adjustments;

(e)              39%-60%,    based    on    Second    Analysis                     without Merchant        Platforms Adjustments.

●                 in Poland:

(a)              20%-42%, based on the First Analysis with Merchant Platforms Adjustments;

(b)             17%-41%,   based   on    the   First   Analysis                    without Merchant Platforms Adjustments;

(c)              17%-23%,   based   on    the   First   Analysis without        Merchant     Platforms Adjustments and with Google's Adjustments for Missing Figures;

(d)             40%-46%,   based    on    the   Second    Analysis with             Merchant     Platforms Adjustments;

(e)              37%-41%,    based    on    Second    Analysis without         Merchant      Platforms Adjustments.

●                 in Sweden:

(a)              64%-85%, based on the First Analysis with Merchant Platforms Adjustments;

(b)             57%-66%,   based   on    the   First   Analysis without        Merchant     Platforms Adjustments;

(c)              41%-49%,   based   on    the   First   Analysis without        Merchant     Platforms Adjustments and with Google's Adjustments for Missing Figures;

(d)             64%-77%,   based    on    the    Second   Analysis with             Merchant     Platforms Adjustments;

(e)              59%-69%,    based    on    Second    Analysis without         Merchant      Platforms Adjustments.

(639)    Second, the results of the analyses indicate that, during the period 2011-2016, total traffic based on the First Analysis and traffic from Google's general search results pages based on the Second Analysis in relative terms: (i) increased to Google's comparison shopping service; (ii) decreased to competing comparison shopping services; and (iii) remained relatively stable or increased at a lower rate (compared to Google's comparison shopping service) to merchant platforms.

(640)     […]

7.3.3.    Potential anti-competitive effects in the national markets for general search services

(641)              The Commission concludes that the Conduct is also capable of having, or likely to have, anti-competitive effects in the national markets for general search services.

(642)              By positioning and displaying more favourably, in its general search results pages, its own comparison shopping service compared to competing comparison shopping services, Google protects the part of the revenue that it generates on its general  search results pages and which serves to finance its general search service. Indeed that revenue could be channelled directly to competing comparison  shopping  services (therefore bypassing Google's general search service). The Conduct may therefore make it more difficult for competing comparison shopping services to reach a critical mass of users that would allow them to compete against Google.

(643)              This is confirmed by several internal Google documents:

(1)     […]763

(2)     […]764 […]765

(3)     […]766

(4)     […]767

7.4. Google's arguments with regard to the applicable legal test

(644)              Google claims that the Conduct cannot amount to an abuse of a dominant position on the facts of the case.

(645)              First, Google claims that the Conduct could be considered abusive only if the criteria established in the Bronner case768 are fulfilled. By failing to apply those criteria, the Commission is imposing on Google a duty to promote competition by allowing competing comparison shopping services to have access to a significant proportion of its general search results pages, despite access to those pages not being indispensable in order to compete.769

(646)              Second, Google claims that there is no precedent for characterising the Conduct as an abuse: even if “Article 102 TFEU does not establish a closed list of abusive  conducts, new abuse categories must be consistent with the legal framework of Article 102 TFEU”, and the “rules must be knowable in advance”.770

(647)              Third, Google claims that the Product Universal and the Shopping Unit constitute product design improvements, which are a form of competition on the merits and “therefore, can be found abusive only in exceptional circumstances”.771

(648)              Google's arguments are unfounded.

(649)              First, it is not novel to find that conduct consisting in the use of a dominant position on one market to extend that dominant position to one or more adjacent markets can constitute an abuse (see recital (334)). Such a form of conduct constitutes a well- established, independent, form of abuse falling outside the scope of competition on the merits.

(650)              Second, the Conduct does not concern a passive refusal by Google to give competing comparison shopping services access to a proportion of its general search results pages, but active behaviour relating to the more favourable positioning and display by Google, in its general search results pages, of its own comparison shopping service compared to comparison shopping services. As indicated in recital (379): (i) Google's own comparison shopping service is not subject to the same ranking mechanisms as its competitors, including adjustment algorithms such as […] and Panda; and (ii) when triggered, Google positions results from its own comparison shopping service on its first general results page in a highly visible place (i.e. either above all generic search results or, in the majority of cases, within or at the level of the first generic search results).

(651)              Third, the Bronner criteria are irrelevant in a situation, such as that of the present case, where bringing to an end the infringement does not involve imposing a duty on the dominant undertaking to “transfer an asset or enter into agreements with persons with whom it has not chosen to contract”.772 While the Decision requires Google to cease the Conduct, it does not require it either to transfer any asset or to enter into agreements with one or more competing comparison shopping services.

(652)              Fourth, there is no indication in the case law that alleged improvements in product designs should be assessed under a different legal standard to that developed to  assess the use of a dominant position on one market to extend that dominant position to one or more adjacent markets (see recital (334)).

7.5. Objective justification and efficiency claims

(653)              The Commission concludes that notwithstanding its arguments, Google has not demonstrated that the Conduct is either objectively necessary, or that  the exclusionary effect produced may be counterbalanced, outweighed even, by advantages in terms of efficiency gains that also benefit consumers.

(654)              Google essentially provides five justifications for the Conduct.

(655)              First, Google claims that it must be entitled to apply adjustment mechanisms in order to preserve the usefulness of its generic search results: “[…] an inability to demote low-quality sites would not serve competition or consumers […] as it would expose Google to a flood of low-quality results […] to the ultimate detriment of users”.773

(656)              Second, Google claims that the positioning and display of Product Universals and Shopping Units is justified because they improve the quality of Google’s search service for users and advertisers. According to Google, the technologies underlying the Product Universals and the Shopping Units, their specialised ranking signals, the organisation of the structured data and their formats improve the quality of Google's general search service for consumers by “provid[ing] users with the most relevant and useful search results possible”.774

(657)              Third, Google claims that if it were required to position and display competing comparison shopping in its general search results pages in the same way as its own comparison shopping service, this would reduce, rather than increase, competition because: (i) search services compete by showing their results, not results from other services and users do not expect search services to provide results from other services;  and (ii)  Google would be unable to monetise space on its general search results pages.775

(658)              Fourth, Google claims that a requirement to show results from competing comparison shopping services would unduly affect its rights and freedoms under the Charter of Fundamental Rights of the European Union (CFR) to impart   information (Article 11), to protect its property (Article 17), and to conduct its business (Article 16).776

(659)              Fifth, Google claims that technically, it “cannot rank results from aggregators alongside its own in a coherent way”.777 Moreover, ranking offers from inventories  of competing comparison shopping services would turn their results into Google Shopping results.778

(660)              The Commission concludes that none of Google's five claims constitutes an objective justification for the Conduct.

(661)              First, the Commission is not preventing Google from applying adjustment mechanisms. The abuse established by this Decision concerns simply the fact that Google does not apply these mechanisms in the same way to Google's comparison shopping service and competing comparison shopping services.

(662)              Second, the Commission is not preventing Google from displaying categories of specialised search results, such as shopping results, in its general search results pages when it determines that they are likely to be relevant or useful to a query. The abuse established by this Decision concerns simply the fact that Google does not position and display in the same way results from Google's comparison shopping service and from competing comparison shopping services.

(663)              Third, Google has provided no evidence to demonstrate that users do not expect search services to provide results from others. On the contrary, as indicated in recital (599), Google did not inform users that the Product Universal was positioned and displayed using different underlying mechanisms than those used to rank generic search results. Similarly, in the case of the Shopping Unit, while the “Sponsored” label may suggest that different positioning and display mechanisms are used, that information is likely to be understandable only by the most knowledgeable users.

(664)              Fourth, a requirement on Google to treat competing comparison shopping services no less favourably than its own comparison shopping service within its general search services does not generally prevent it from monetising its general search results pages. Google can choose the specific measures through which it intends to comply with this Decision and the possible measures Google might take do not preclude the monetisation of its general search results pages when making this choice.

(665)              Fifth, any restriction on Google's right to impart information, right to protect its property and freedom to conduct its business brought about by the Decision (i) is provided for by law; (ii) corresponds to objectives of general interest recognised by the Union; (iii) respects the essence of those rights and freedoms; and (iv) is necessary to protect the rights and freedoms of others.

(666)              In the first place, the Decision is based on Article 102 of the Treaty, a provision of primary Union law that is sufficiently precise to meet the requirement that any restriction on Google’s fundamental rights  and freedoms resulting from a decision under Article 7 of Regulation (EC) No 1/2003 finding an infringement of Article 102 of the Treaty is “provided by law”.779

(667)              In the second place, any restriction on Google's rights and freedoms corresponds to the objective of the Union to establish an internal market, which in accordance with Protocol No 27 on the internal market and competition, annexed to the Treaty of Lisbon, includes a system ensuring that free competition is not distorted to the detriment of the public interest, individual undertakings and consumers.780

(668)              In the third place, any restriction on Google's rights and freedoms respects  the essence of those rights and freedoms.781 In particular, any restriction does not interfere with how and what information Google can provide to users in response to a given query, including in terms of grouping together product information.

(669)              In the fourth place, any restriction on Google's rights and freedoms is necessary to protect the freedom of competing comparison shopping services and other economic operators to conduct a business consisting in returning product offers from merchant websites, enabling users to compare them. This freedom, in turn, furthers the above- mentioned objective of general interest that competition is not distorted to the detriment of the public interest and consumers.

(670)              Any restriction on Google's rights and freedoms is also necessary to protect the right of users to receive information from competing comparison shopping services. As indicated in recitals (437) and (599), users are not all aware of the fact that Product Universals or the Shopping Units are subject to different underlying processes and mechanisms than competing comparison shopping services for being triggered and ranked in Google's general search results pages in response to a product query.

(671)              Sixth, Google has failed to demonstrate that it cannot use the same underlying processes and methods in deciding the positioning and display of the results of its own comparison shopping service and for those of competing comparison shopping services. Rather, the scenarios proposed and considered by Google during the commitments  discussions  […],782   […]783   suggests  that  the  implementation  of an

equal treatment remedy is technically feasible.

8. JURISDICTION

8.1. Principles

(672)              In order to justify the Commission’s jurisdiction, it is sufficient that a conduct is either  implemented  in  the  EEA  or  is  liable  to  have  immediate,  substantial  and foreseeable effects in the EEA.784 Those two approaches for establishing the Commission’s jurisdiction are alternative.785

(673)              The criterion of implementation is satisfied by making sales within the EEA irrespective of the location of sources of supply or of production plants.786 It follows that direct sales of the products covered by the conduct to customers in the EEA are not the only means of implementation of a conduct.787

(674)              The criterion of immediate, substantial and foreseeable effects of conduct in the EEA is satisfied when the conduct in question is capable of having such an effect, there being no need to show actual effects.788 A relevant factor in conducting this assessment is whether the conduct was intended to produce effects within the internal market.789

8.2. Application to this case

(675)              The Commission has jurisdiction to apply both Article 102 of the Treaty and Article 54 of the EEA Agreement to the Conduct because the Conduct is implemented in the territories of the Contracting Parties to the EEA Agreement and it is also in those territories that the Conduct is likely to produce substantial, immediate and foreseeable effects.790

(676)              Google does not contest that the Commission has jurisdiction over the Conduct.

9. EFFECT ON TRADE BETWEEN MEMBER STATES

9.1. Principles

(677)              Article 102 of the Treaty prohibits as incompatible with the internal market an abuse of a dominant position “in so far as it may affect trade between Member States”. Article 54 of the EEA Agreement contains a similar prohibition with respect to trade between Contracting Parties to the EEA Agreement.

(678)              The effect on trade criterion consists of three elements.

(679)              First, “trade” must be potentially affected. The concept of trade is not limited to traditional exchanges of goods and services across borders, but covers all cross- border economic activity. It also encompasses practices affecting the competitive structure of the internal market by eliminating or threatening to eliminate a competitor operating within the territory of the Union.791

(680)              Second, the practice does not necessarily need to reduce trade;792 it is sufficient to show that the abuse “may affect trade between Member States”. In other words, it must be foreseeable with a sufficient degree of probability on the basis of a set of objective factors of law or fact that the practice in question has an influence, direct or indirect, actual or potential, on the pattern of trade between Member States.793 Where a dominant undertaking engages in exclusionary conduct in more than one Member State, such conduct is presumed, by its very nature, to be capable of affecting trade between Member States.794

(681)              Third, the effect on trade between Member States must be “appreciable”. This element requires that effect on trade between Member States must not be insignificant and is assessed primarily with reference to the position of the undertaking(s) on the relevant product market(s).795 The stronger the position of an undertaking, the more likely it is that the effect of a practice on trade between Member States will be appreciable.796

9.2. Application to this case

(682)              The Conduct has an appreciable effect on trade between Member States (and Contracting Parties to the EEA Agreement) for the following reasons.

(683)              First, the Conduct is, by its very nature, cross-border in scope. In addition, the Conduct affects the competitive structure of the internal market by eliminating or threatening to eliminate competitors operating within the territory of the Union (and the EEA). It therefore affects trade between Member States (and Contracting Parties to the EEA Agreement).

(684)              Second, Google and its competitors are active in several Member States and the Conduct has taken place in the territories of 13 Contracting Parties to the EEA Agreement. It is therefore capable of having an effect on trade between Member States (and Contracting Parties to the EEA Agreement ).

(685)              Third, Google has a dominant position in the relevant product and geographic markets defined by this Decision. In 2016, Google's shares in the national markets  for general search services were above 90% in almost all of the EEA Countries (see section 6.2.1). The abuse is therefore capable of having at least an appreciable effect on trade between Member States (and Contracting Parties to the EEA Agreement).

10.  DURATION

(686)              The Commission concludes that the infringement started in each of the thirteen national markets for general search services from the moment Google launched the Product Universal in that national market, or, if the Product Universal was never launched in that national market, from the moment Google launched the Shopping Unit in that market, namely:

–             since January 2008 in Germany and the United Kingdom;

–             since October 2010 in France;

–             since May 2011 in Italy, the Netherlands, and Spain;

–             since February 2013 in the Czech Republic; and

–             since November 2013 in Austria, Belgium, Denmark, Norway, Poland and Sweden.797

(687)              The infringement is ongoing in each of the thirteen national markets for general search services as at the date of adoption of this Decision.

11. ADDRESSEES

11.1. Principles

(688)              Article 102 of the Treaty and Article 54 of the EEA Agreement are addressed to undertakings. The concept of an undertaking refers to any entity engaged in an economic activity, regardless of its legal status and the way in which it is financed.798 The term “undertaking” must also be understood as designating an economic unit even if in law that economic unit consists of several persons, natural or legal.799

(689)              When such an economic entity infringes the competition rules of the Treaty (and of the EEA Agreement), it falls, according to the principle of personal responsibility, to that entity to answer for that infringement.800 However, the infringement of competition law must be imputed unequivocally to a legal person on whom fines  may be imposed and the statement of objections must be addressed to that person. It is also necessary that the statement of objections indicates in which capacity a legal person is called on to answer the allegations.801

(690)              The conduct of a subsidiary may be imputed to the parent company even if the parent company does not participate directly in the infringement when the parent company and the subsidiary form a 'single economic entity', that is to say a single 'undertaking' within the meaning of Articles 101 and 102 of the Treaty, because, in such a case, the parent  company  exercises  a  decisive  influence  over  the  subsidiary  which has participated in it.802 A parent company that owns 100% (or almost 100%) of a subsidiary has the ability to exercise decisive influence over that subsidiary. In such  a case, there exists a rebuttable presumption that the parent company also in fact exercises  that  influence  without  the  need  for  the  Commission  to  adduce further evidence on the actual exercise of influence (the parental liability presumption).803 In those circumstances, it is sufficient for the Commission to prove that the subsidiary  is wholly owned by the parent company in order to assume that the parent company exercises a decisive influence over the commercial policy of the subsidiary. The parent company can then be held jointly and severally liable for the payment of the fine imposed on its subsidiary, unless the parent company, which has the burden of rebutting that presumption, adduces sufficient evidence to show that its subsidiary acts independently on the market.804 The same principles hold true for the purposes  of the application of Article 54 of the EEA Agreement.

11.2. Application to this case

(691)              The infringement in this case should be imputed to Google because it directly engaged in the Conduct that resulted in the infringement (see section 7).

(692)              The infringement should also be imputed to Alphabet with respect to the Conduct during the period from 2 October 2015 because Google has been wholly owned by Alphabet since that date805 and Alphabet has not provided any evidence to rebut the presumption that it has exercised decisive influence over Google since that date.

12. REMEDIES

12.1. Principles

(693)              Article 7(1) of Regulation (EC) No 1/2003 provides that where the Commission  finds that there is an infringement of Article 102 of the Treaty and Article 54 of the EEA Agreement it may by decision require the undertaking concerned to bring such infringement to an end. For this purpose, it may also impose on the undertaking concerned any behavioural or structural remedies which are proportionate to the infringement committed and necessary to bring the infringement effectively to an  end.

(694)              It follows that a decision pursuant to Article 7(1) of Regulation No (EC) 1/2003 may include an order to “do certain acts or provide certain advantages which have been wrongfully withheld as well as prohibiting the continuation of certain action, practices or situations which are contrary to the Treaty”.806  The Commission may require the undertaking concerned to submit to it proposals with a view to bringing the situation into conformity with the requirements of the Treaty.807

(695)              The requirement that a remedy has to be effective808 empowers the Commission to enjoin a dominant undertaking to refrain from adopting any measures having the same or an equivalent object or effect as the conduct identified as abusive.809 Any remedy must also apply in relation to the infringement that has been established810 and be proportionate to the infringement identified.811

(696)              Where there is more than one way of bringing an infringement effectively to an end in conformity with the Treaty, it is for the addressee of a decision to choose between those various ways.812

12.2. Application to this case

(697)              Google and Alphabet should be required to bring the infringement established by this Decision effectively to an end and henceforth refrain from any measure that has the same or an equivalent object or effect.

(698)              As there is more than one way in conformity with the Treaty of bringing that infringement effectively to an end, it is for Google and Alphabet to choose between those various ways.

(699)              Any measure chosen by Google and Alphabet should, however, ensure that Google treats competing comparison shopping services no less favourably than its own comparison shopping service within its general search results pages. The principles mentioned in recital (700) should apply irrespective of whether Google chooses to display a Shopping Unit or another equivalent form of grouping of links to or search results from comparison shopping services.

(700)              In particular, any measure chosen by Google and Alphabet:

(a)   should apply to all devices, irrespective of the type of device on which the search is performed;

(b)   should apply to all users of Google situated in the thirteen EEA countries in which the Conduct takes place, irrespective of the Google domain that they use (including Google.com);

(c)   should subject Google's own comparison shopping service to the same underlying processes and methods for the positioning and display in Google's general search results pages as those used for competing comparison shopping services. Such processes and methods should include all elements that have an impact on the visibility, triggering, ranking or graphical format of a search result in Google's general search results pages, including:

●                                         processes and methods or relevance standards determining the triggering of comparison shopping services on the general search results pages in response to a query;

●                                         processes and methods determining the positioning and display of comparison shopping services in response to a query, including relevance standards, ranking algorithms, adjustment or demotion mechanisms and their respective conditions, parameters and signals;

●                                         type of landing pages for clicks on comparison shopping services;

●                                         visual appearance on comparison shopping services and branding possibilities;

●                                         type and granularity of information on the results of comparison shopping services available to users; and

●                                         the possibility of interaction with users.

(d)   should not lead to competing comparison shopping services being charged a  fee or another form of consideration that has the same or an equivalent object  or effect as the infringement established by this Decision.

(701)              Google and Alphabet should be given 90 days from the date of the notification of this Decision to implement measures that bring the infringement effectively to an end. A 90 days period is sufficient to implement such measures, considering in particular that, at the time of the discussions on commitments, Google itself offered such a deadline for the implementation of those commitments.

(702)              Google and Alphabet should be required to notify the Commission, within 60 days from the date of notification of this Decision, of the measures by means of which they intend to bring the infringement effectively to an end. That communication should be sufficiently reasoned and detailed to enable the Commission to make a preliminary assessment as to whether those measures will ensure that  the infringement is brought to an end effectively and in accordance with the principles  set out in recital (700). Any statements by the Commission to Google and Alphabet or silence on the part of the Commission between the 60 day deadline and 90 day deadline should not be interpreted as an indication that the intended measures communicated by Google and Alphabet will ensure that the infringement is brought to an end effectively.

(703)              The Commission is entitled to monitor the implementation by Google and Alphabet of the remedies ordered in the Decision. For those purposes, it is entitled to use the powers of investigation provided for in Regulation No (EC) 1/2003.813

(704)              Considering the variety and complexity of the measures that Google and Alphabet may take to bring the infringement effectively to an end, Google and Alphabet  should provide the Commission with periodic reports on the way they comply with this Decision. The first of those reports should be sent on the day when Google and Alphabet bring the infringement effectively to an end. The next reports should be  sent every four months, for a period of five years from that day.

(705)              For the same reasons set out in recital (704), the Commission may also decide to use the services of one or several external technical expert(s).814

13. PERIODIC PENALTY PAYMENTS

13.1. Principles

(706)              Pursuant to Article 24(1)(a) of Regulation (EC) No 1/2003 and Article 5 of Council Regulation (EC) No 2894/94815, the Commission may, by decision, impose on undertakings or associations of undertakings periodic penalty payments  not exceeding 5% of the average daily turnover in the preceding business year per day and calculated from the day appointed by the decision, in order to compel them to

put an end to the infringement, in accordance with a decision taken pursuant to Article 7 of Regulation (EC) No 1/2003.

13.2. Application to this case

(707)              The Commission concludes that it is necessary to impose periodic penalty payments pursuant to Article 24(1)(a) of Regulation No 1/2003 and Article 5 of Council Regulation (EC) No 2894/94 if Google and Alphabet were to fail to: (i) implement measures that bring the infringement effectively to an end within 90 days from the date of notification of this Decision; (ii) notify the Commission within 60 days from the date of notification of this Decision of the specific measures by means of which they intend to bring the infringement effectively to an end; and (iii) provide the Commission with periodic reports every four months, for a period of five years, on the way they comply with this Decision.

(708)              In setting the level of the periodic penalty payments, the Commission has taken into account the need to impose periodic penalty payments sufficient to ensure  compliance by Google and Alphabet with this Decision. The Commission has also taken into account the need to set periodic penalty payments that are sufficient to ensure compliance by other undertakings of a similar size and with similar financial resources.

(709)              Consequently, if Google and Alphabet were to fail to comply with any of the requirements set out in recital (707), the Commission hereby imposes a daily  periodic penalty payment of 5% of Alphabet's average daily turnover in the business year preceding such a failure to comply.

14. FINES

14.1. Principles

(710)              Pursuant to Article 23(2)(a) of Regulation (EC) No 1/2003 and Article 5 of Council Regulation (EC) No 2894/94, the Commission may by decision impose fines on undertakings, where, either intentionally or negligently, they infringe Article 102 of the Treaty or Article 54 of the EEA Agreement.

(711)              An infringement of Article 102 of the Treaty or Article 54 of the EEA Agreement is committed intentionally or negligently where the undertaking concerned cannot be unaware of the anticompetitive nature of its conduct, whether or not it was aware that it was infringing the competition rules of the Treaty.816 Regarding specifically an undertaking in a dominant position, the undertaking is aware of the anti-competitive nature of its conduct where it is aware of the essential facts justifying both the finding of a dominant position on the relevant market and the finding by the Commission of an abuse of that position.817

(712)              Pursuant to Article 23(3) of Regulation (EC) No 1/2003, in fixing the amount of the fines, the Commission must have regard to all relevant circumstances and  particularly to the gravity and to the duration of the infringement. In doing so, the Commission will set the fines at a level sufficient to ensure deterrence. The Commission will reflect any aggravating or mitigating circumstances in the fines imposed.

(713)              In setting the fines to be imposed, the Commission refers to the principles laid down in its Guidelines on the method of setting fines imposed pursuant to Article 23(2)(a) of Regulation (EC) No 1/2003 (“the Guidelines on fines”).818

(714)              First, the Commission defines the basic amount of the fine.819 That amount is to be set by reference to the value of sales, namely the value of the undertaking’s sales of goods or services to which the infringement directly or indirectly relates in the relevant geographic area in the EEA.820 The value of sales will be assessed before VAT and other taxes directly related to the sales.821

(715)              The Commission will normally take into account the sales made by the undertaking during the last full business year of the occurrence of the infringement.822

(716)              The amount of the value of sales taken into account will correspond to a percentage which is set at a level of up to 30% of the value of sales.823 The choice of a given percentage will depend on the degree of gravity of the infringement. In assessing the gravity of the infringement, the Commission has regard to a number of factors, such as the nature of the infringement, the market shares of the undertaking concerned, the geographic scope of the infringement and whether or not the infringement has been implemented.824

(717)              The proportion of the value of sales resulting from that percentage will then be multiplied by the duration of the infringement.825

(718)              The Commission may also include in the basic amount an additional amount of up to 25% of the value of sales, irrespective of the duration.826

(719)              Second where applicable, the Commission adjusts the basic amount upwards or downwards to take into account aggravating or mitigating circumstances.827 Those circumstances are listed non-exhaustively in points 28 and 29 of the Guidelines on fines.

(720)              Third, the Commission will pay particular attention to the need to ensure that fines have a sufficiently deterrent effect. To that end, the Commission may increase the fine to be imposed on an undertaking which has a particularly large turnover beyond the sales of goods or services to which the infringement relates.828

(721)              Finally, pursuant to Article 23(2) of Regulation (EC) No 1/2003, the fine for an infringement must not exceed 10% of the undertaking’s total turnover in the preceding business year.

14.2. Intention or negligence

(722)              The Commission concludes that, contrary to what Google claims,829 Google and Alphabet committed the infringement described in this Decision intentionally or negligently.

(723)              First, Google and Alphabet could not have been unaware of the fact that Google held a dominant position in the national markets for general search services (see section 6.2).

(724)              In the first place, Google and Alphabet ought to have been familiar with the principles governing market definition in competition cases and, where necessary, taken appropriate legal advice regarding the definition of the markets for general search services and for comparison shopping services.830

(725)              In the second place, Google and Alphabet ought to have been familiar with the significance of Google's strong and stable market shares in the national markets for general search services (see section 6.2.1) and taken them into consideration in relation to the Conduct.831

(726)              Second, Google and Alphabet could not have been unaware of the fact that the Conduct constitutes an abuse of Google's dominant position on each of the thirteen national markets for general search services in which the Conduct takes place.

(727)              In the first place, conduct consisting in the use of a dominant position on one market to extend that dominant position to one or more adjacent markets constitutes a well- established form of abuse falling outside the scope of competition on the merits (see recital (649)).

(728)              In the second place, the Commission formally set out its competition concerns regarding the Conduct in the Preliminary Assessment adopted on 13 March 2013 in which it explained why it considered that the Conduct was contrary to Article 102 of the Treaty.832

(729)              In the third place, even if the Conduct may have certain features that have not been examined in past decisions, this does not prevent the imposition of a fine.833

14.3. Imposition of a fine notwithstanding commitment discussions

(730)              The Commission concludes that, contrary to what Google claims834, it can, and indeed should, impose a fine, notwithstanding the fact that it had considered adopting a decision under Article 9 of Regulation (EC) No 1/2003 and irrespective of whether Google offered three sets of commitments in good faith.835

(731)              First, the Commission has a margin of discretion in the choice between adopting a decision under Article 7 and adopting a decision under Article 9 of that  Regulation.836

(732)              Second, there were several reasons why, in this case, the Commission decided to revert to the procedure under Article 7 of that Regulation (see recitals (123)-(137)).

(733)              Third, having reverted to the procedure under Article 7 of that Regulation, the Commission is entitled to make use of the full range of its powers under that Regulation, including the power to order that an infringement be brought to an end and the power to impose fines for that infringement.837

(734)              Fourth, it is only in strictly exceptional situations, such as where an undertaking’s cooperation has been decisive in establishing an infringement, that a fine may not be imposed.838 In this case, however, Google's offer of commitments in no way assisted the Commission in establishing the infringement.

14.4. Calculation of the fine

14.4.1. Joint and several liability

(735)              The Commission has concluded that Alphabet is jointly and severally liable for the infringement as of 2 October 2015 (see section 11).

(736)              The Commission therefore concludes that Google and Alphabet should be  held jointly and severally liable to pay the fine insofar as it relates to the period from that date.

14.4.2.         Basic amount of the fine

14.4.2.1.The value of sales

(737)              The Commission concludes that the infringement directly or indirectly relates to the revenues generated by Google's comparison shopping service because the infringement is capable of having, or likely to have, anti-competitive effects in the national markets for comparison shopping services (see section 7.2).

(738)              For the purpose of the value of sales, the Commission therefore uses revenue generated by Google's comparison shopping service in each of the thirteen national markets in which the Conduct takes place (Austria, Belgium, the Czech Republic, Denmark, France, Germany, Italy, the Netherlands, Norway, Poland, Spain, Sweden and the United Kingdom). This includes revenues from the paid product results displayed in the Shopping Unit, revenues from the paid product results displayed on the standalone Google Shopping website and revenues obtained from bottom text ads displayed on the standalone Google Shopping website.

(739)     […]839

14.4.2.2.The last business year

(740)  In this case, there are no exceptional reasons to deviate from the basic principle that  the fine should be based on the last full business year's revenues.

(741) The Commission therefore concludes that the value of sales should be based on the revenues generated by Google's comparison shopping service during the last full business year of the infringement, namely 2016.

14.4.2.3.Gravity

(742) The Commission concludes that the proportion of the value of sales to be used to establish the basic amount of the fine should be […]%.

(743)     In reaching this conclusion, the Commission takes into account the following factors:

(a)   the national markets for comparison shopping services and general search services are of significant economic importance. This means that any anti- competitive behaviour on these markets is likely to have a considerable impact; and

14.4.2.5. Additional amount

(749)              The Commission concludes that the basic amount should include an additional amount of […]% of the relevant value of sales.840

(750)              In reaching this conclusion, the Commission takes into account the factors set out in recital (743) and the need to ensure that the fine imposed has a sufficient deterrent effect on undertakings of a similar size and with similar resources.841

14.4.3. Adjustments to the basic amount

14.4.3.1. Aggravating and mitigating circumstances

(751)              The Commission concludes that there are no aggravating or mitigating circumstances that should result in an increase or decrease in the basic amount of the fine.

14.4.3.2.Specific increase for deterrence

(752)              The Commission concludes that the basic amount of the fine imposed should be multiplied by 1.3.

(753)              In reaching this conclusion, the Commission takes into account: (i) the  need  to ensure that the fine has a sufficiently deterrent effect not only on Google and Alphabet, but also on undertakings of a similar size and with similar resources,842  and (ii) the fact that Alphabet had a particularly large turnover in  2016 (approximately EUR  81 597  million)  beyond  the  revenues  generated  by Google's

comparison shopping service in each of the thirteen national markets in which the Conduct takes place.

14.4.4.  Final amount of the fine

(754)              The Commission concludes that the final amount of the fine to be  imposed  on Google should be EUR 2 424 495 000, of which EUR 523 518 000 jointly and severally with Alphabet.

(755)              Alphabet's turnover in the business year ending 31 December 2016 was approximately EUR 81 597 million. As the final amount of the fine set is below 10% of that figure, no adaptation is necessary.

HAS ADOPTED THIS DECISION:

Article 1

By positioning and displaying more favourably, in Google Inc.'s general search results pages, Google Inc.'s own comparison shopping service compared to competing comparison shopping services, the undertaking consisting of Google Inc. and also, since 2 October 2015, of Alphabet Inc. has infringed Article 102 of the Treaty and Article 54 of the Agreement on the European Economic Area. The infringement has been taking place in the following countries since the following dates:

(a)              as regards Google Inc.:

–             since January 2008 in Germany and the United Kingdom;

–             since October 2010 in France;

–             since May 2011 in Italy, the Netherlands, and Spain;

–             since February 2013 in the Czech Republic; and

–             since November 2013 in Austria, Belgium, Denmark, Norway, Poland and Sweden.

(b)             as regards, additionally, Alphabet Inc., since 2 October 2015, in each of the countries listed in point (a).

The infringement is continuing in each of those countries as at the date of adoption of this Decision.

Article 2

For the infringement referred to in Article 1, the following fine is imposed:

Google Inc.: EUR 2 424 495 000, of which EUR 523 518 000 jointly and severally with Alphabet Inc.

The fine shall be credited in euros, within a period of three months from the date of notification of this Decision, to the following bank account held in the name of the European Commission:

BANQUE ET CAISSE D'EPARGNE DE L'ETAT

1-2, Place de Metz

L-1930 Luxembourg

IBAN: LU02 0019 3155 9887 1000 BIC: BCEELULL

Ref.: European Commission – BUFI/AT.39740

After the expiry of that period, interest shall automatically be payable at the interest rate applied by the European Central Bank to its main refinancing operations on the first day of the month in which this Decision is adopted, plus 3.5 percentage points.

Where an addressee of this Decision brings an action for annulment against the Decision, it shall cover the fine by the due date, either by providing an acceptable financial guarantee or by making a provisional payment of the fine in accordance with Article 90 of Commission Delegated Regulation (EU) No 1268/2012.843

Article 3

The undertaking referred to in Article 1 shall, within 90 days from the date of notification of this Decision, bring effectively to an end the infringement referred to in that Article, in so far as it has not already done so.

The undertaking referred to in Article 1 shall refrain from repeating any act or conduct  described in Article 1, and from any act or conduct having the same or an equivalent object or effect.

Article 4

The undertaking referred to in Article 1 shall notify the Commission, within 60 days from the date of notification of this Decision, of the specific measures through which it intends to comply with this Decision.

The undertaking referred to in Article 1 shall provide the Commission with periodic reports  on the way it complies with this Decision. The first of those reports shall be sent on the day  on which the undertaking brings the infringement effectively to an end. Subsequent reports shall be sent every four months from that day, for a period of five years from that day.

Article 5

If the undertaking referred to in Article 1 fails to comply with the orders set out in Articles 3 and 4, the Commission hereby imposes a daily periodic penalty payment of 5% of its average daily turnover in the business year preceding such a failure to comply.

Article 6

This Decision is addressed to Google Inc. and Alphabet Inc., both of 1600 Amphitheatre Parkway, Mountain View, CA 94043, United States of America.

This Decision shall be enforceable pursuant to Article 299 of the Treaty and Article 110 of the Agreement on the European Economic Area.

Done at Brussels, 27.6.2017

1                  OJ L 1, 4.1.2003, p. 1. With effect from 1 December 2009, Articles 81 and 82 of the EC Treaty have become Articles 101 and 102, respectively, of the Treaty on the Functioning of the European Union (“the Treaty”). The two sets of provisions are, in substance, identical. For the purposes of this Decision, references to Articles 101 and 102 of the Treaty should be understood as references to Articles 81 and 82, respectively, of the EC Treaty when where appropriate. The Treaty also introduced certain changes in terminology, such as the replacement of “Community” by “Union” and “common market” by “internal market”. The terminology of the Treaty is used throughout this Decision.

2                  OJ L 123, 27.4.2004, p. 18.

3                  Throughout this Decision, whenever the Commission refers to the more favourable positioning and display in Google’s general search results pages of Google’s own comparison shopping service compared to competing comparison shopping services, the Commission means the more favourable positioning and display of: (i) links to Google’s own comparison shopping service (see section 2.2.5, recital (29), and section 7.2.1.3, recitals (408) to (411)); and/or (ii) parts or all of Google’s own comparison shopping service (see section 2.2.5, recital (32), and section 7.2.1.3, recitals (412) to (423)).

4                  Amount converted from USD into EUR using the average annual reference exchange rate published by the     European     Central     Bank     for     the     year     2016,     i.e.     1     USD     =     0.9039  EUR (https://www.ecb.europa.eu/stats/policy and exchange rates/euro reference exchange rates/html/euro fxref-graph-usd.en.html).

5                  Alphabet’s Form 10-K Annual Report for the US fiscal year ending 31 December 2016, available at https://abc.xyz/investor/pdf/20161231 alphabet 10K.pdf, downloaded on 3 March  2017.

6                  Google’s Form 10-K Annual Report for the US fiscal year ending 31 December 2015, available at https://www.sec.gov/Archives/edgar/data/1288776/000165204416000012/goog10-k2015 htm, downloaded on 3 March 2017.

7                  “General search” is also known as “online search” or “horizontal search”. The term “general  search” is used throughout this Decision.

8                  “Generic search results” are also known as “organic search results” or “natural search results”. The term “generic search results” is used throughout this Decision.

9                  “Specialised   search”  is   also   referred   to   as  “vertical   search”  or  “universal   search”. The  terms “specialised search” and “specialised search results” are used throughout this Decision.

10                 Google’s  submission […].

11                 Google's  submission […].

12                 Google,        “Webmaster        tools:        How        Google        Search        Works”,        available        at http://support.google.com/webmasters/bin/answer.py?hl=en&answer=70897#content,  downloaded   on 6 March 2015.

13                 Sergey Brin and Larry Page, “The PageRank Citation Ranking: Bringing Order to the   Web”, available at http://ilpubs.stanford.edu:8090/422/1/1999-66.pdf, downloaded on 6 March 2015.

14                 Until February 2016, Google positioned AdWords results also on the right side of its general search results pages (see https://searchenginewatch.com/2016/02/21/google-is-removing-all-right-hand-side- ads-on-serps-worldwide/, downloaded on 21 March 2017).

15                 A second-price auction is an auction in which the bidder who submitted the highest bid is   awarded the object (or service) being sold and pays a price equal to the amount bid by the second highest bidder.

16                 The predicted click-through rate represents the probability that a search advertisement will receive a click if placed for a particular query (see Google’s reply to […] the Commission’s request for information […]).

17                 Google’s reply to […] the Commission’s request for information […].

18                 Google's submission […].

19                 Google’s submission […].

20                 Traffik,       “Google       Launches       Froogle       UK”,       12       October       2004,     available      at http://www.traffick.com/2004/10/google-launches-froogle-uk.asp, downloaded on 6 March 2015.

21                 Google     blog,     “Have     a     Very     Froogle     Friday”,     25     November     2004,   available    at https://googleblog.blogspot.be/2004/11/have-very-froogle-friday html, downloaded on 28 March 2017.

22                 Google blog, “Back to Basics”, 18 April 2007, available at http://googleblog.blogspot.be/2007/04/back- to-basics.html, downloaded on 6 March 2015.

23                 Google's submission […].

24                 Google   Commerce   “Building   a   better   shopping   experience”,   31   May   2012,       available   at http://googlecommerce.blogspot.be/2012/05/building-better-shopping-experience html, downloaded on 6 March 2015.

25                 Screenshot taken on www.google.be on 21 June 2017.

26                 SO Response, paragraphs 76-77.

27                 Google's reply to […] the Commission's request for information […].

28                 SO Response, paragraphs 76-77.

29                 Google's reply to […] the Commission's request for information […].

30                 OJ C 101, 27.4.2004, p. 43.

31                 OJ L 275, 20.10.2011, p. 29.

32                 The Preliminary Assessment did not take a view as to the relationship between Google’s use of original content from third party websites and intellectual property law.

33                 Commission  letters  to  1PlusV of 12  June 2014,  to  AEDE of 12  June 2014,  to  BDZV  and VDZ of 17 June 2014, to  BEUC  of  17  June  2014,  to  Ciao  of  11  August  2014,  to  Deutsche  Telekom  of 19 June 2014, to Elf of 5 June 2014, to Euro-Cities of 17 June 2014, to Expedia of 17 June 2014, to Foundem of 2 June 2014, to Hot Maps of 17 June 2014, to ICOMP of 20 June 2014, to Microsoft of    27 May 2014, to Nextag and Guenstiger of 11 June 2014, to nntp.it of 12 June 2014, to Odigeo of        12 June  2014,  to  Streetmap  of  18  June  2014,  to  TripAdvisor  of  4  June  2014,  to  Twenga  of   12 June 2014, to VfT of 5 June 2014 and to Visual Meta of 13 June 2014.

34                 1PlusV’s  submission  of  15  July  2014;  AEDE’s  submission  of  25  July  2014;  BDZV and VDZ’s submission  of  1  August  2014;  BEUC’s  submission  of  14  July  2014;  Ciao’s  submission  of       10 September 2014;  Deutsche  Telekom’s  submission   of   22   July   2014;   Elf’s   submission   of  19 June 2014; Euro-Cities’ submission of 4 August 2014; Expedia’s submission of 18 July 2014; Foundem’s submission of 11 July 2014; Hot Maps’ submission of 4 August 2014; ICOMP’s submissions of 28 July 2014 and 5 August 2014; Microsoft’s submission of 7 July 2014; Nextag and Guenstiger’s submission of 22 July 2014; Odigeo’s submission of 14 July 2014; TripAdvisor’s submission of 7 July 2014; Twenga’s submission of 25 July 2014; VfT’s submission of 23 July 2014 and Visual Meta’s submission of 30 July 2014.

35                 Email […] of 4 September 2014. See also Submission of Google of […].

36                 The Commission's decision to revert to the procedure under Article 7 of Regulation (EC) No  1/2003 in relation to part of the first business practice is without prejudice to the compatibility with Articles 101 and 102 of the Treaty of the other business practices by Google identified in the Preliminary Assessment. In that regard, on 14 July 2016, the Commission addressed to Google and Alphabet a Statement of Objections in COMP/C-3/40.411 – Google Search (AdSense) in relation to the third business practice by Google identified in the Preliminary Assessment.

37                 Commission letters to 1PlusV of 19 June 2015, to Acheter moins cher of 18 June 2015, to BDZV and VDZ of 22 June 2015, to BEUC of 26 June 2015, to Ciao of 19 June 2015, to Company AC of             19 June 2015, to Deutsche Telekom of 19 June 2015, to ETTSA of 4 September 2015, to Euro-Cities of 19 June 2015, to Expedia of 26 June 2015, to FairSearch of 19 June 2015, to Foundem of 18 June 2015,

to Getty of 26 June 2015, to HolidayCheck of 19 June 2015, to Hot Maps of 18 June 2015, to ICOMP of 26 June 2015, to Kelkoo of 19 June 2015, to LeGuide of 19 June 2015, to Microsoft of 19 June 2015, to MIH of 10 July 2015, to MoneySupermarket of 18 June 2015, to News Corp of 15 July 2015, to Nextag and Guenstiger of 18 June 2015, to OCU of 19 June 2015, to Odigeo of 18 June 2015, to OIP of 26 June 2015, to TradeComet of 15 July 2015, to TripAdvisor of 18 June 2015, to Trivago of 18 June

2015, to Twenga of 19 June 2015, to VfT of 19 June 2015, to Visual Meta of 22 June 2015, and to Yelp of 26 June 2015. The Commission also sent a non-confidential  version of the  SO to VG Media on       7 October 2016.

38                 Submission of 1PlusV of 17 July 2015; Submission of BDZV and VDZ of 3 August  2015; Submission of BEUC of 24 July 2015; Submission of Ciao and Microsoft of 31 July 2015; Submission of Company AC of 29 July 2015;  Submission of Deutsche  Telekom of  16  July 2015;  Submission of ETTSA of  16 October 2015; Submission of Euro-Cities of 31 July 2015; Submission of Expedia of 31 July 2015; Submission of FairSearch of 10 August 2015; Submission of Foundem of 29 July 2015; Submission of Getty of 7 August 2015; Submission of Hot Maps of 31 July 2015; Submission of ICOMP of 31 July 2015; Submission of Kelkoo of 3 August 2015; Submission of LeGuide of 20 July 2015; Submission of News Corp of 26 August 2015; Submission of Nextag and Guenstiger of 16 July 2015; Submission of OCU of 26 July 2015; Submission of OIP of 3 August 2015; Submission of TradeComet of 12 August 2015; Submission of TripAdvisor of 24 July 2015; Submission of Trivago of 24 July 2015; Submission of Twenga of 29 July 2015; Submission of Visual Meta  of 3  August  2015; Submission of Yelp of       7 August 2015.

39                 Commission letters to 1PlusV of 6 November 2015, to Acheter moins cher of 27 October 2015, to BDZV and VDZ of 11 November 2015, to BEUC of 6 November 2015, to Ciao and  Microsoft of       27 October 2015, to Company AC of 30 October 2015, to Deutsche Telekom of 5 November 2015, to ETTSA of 27 October 2015, to Euro-Cities of 19 November 2015, to Expedia of 30 October 2015, to FairSearch of 5 November 2015, to Foundem of 30 October 2015, to Getty of 27 October 2015, to HolidayCheck of 19 November 2015, to Hot Maps of 30 October 2015, to ICOMP of 27 October 2015, to Kelkoo of 6 November 2015, to LeGuide of 9 November 2015, to MIH of 6 November 2015, to MoneySupermarket of 9 November 2015, to News Corp of 30 October 2015, to Nextag and Guenstiger of 30 October 2015, to Odigeo of 30 October 2015, to OIP of 24 November 2015, to TradeComet of      9 November 2015, to TripAdvisor of 30 October 2015, to Trivago of 30 October 2015, to Twenga of 30 October 2015, to VfT of 19 November 2015, to Visual Meta of 9 November 2015, and to Yelp of        27 October 2015.

40                 Submission of 1PlusV of 3 December 2015; Submission of Acheter moins cher of 18  November 2015; Submission of BDZV and VDZ of 23 December 2015; Submission of BEUC of 18 December 2015; Submission of Company AC of 16 December 2015; Submissions of Expedia of 14 December 2015 and 6 January 2016; Submission of FairSearch of 18 December 2015 updated on 19 January 2016; Submission of Foundem of 29 December 2015; Submission of Getty of 15 December 2015; Submission of Kelkoo of 18 December 2015; Submissions of LeGuide of 7 December 2015 and 29 January 2016; Submission of MIH of 23 December 2015 updated on 20 January 2016; Submission of News Corp of 15 December 2015; Submissions of Nextag and Guenstiger of 12 and 22 December 2015; Submission of OIP of 5 January 2016; Submissions of TripAdvisor of 14 December 2015 and 6 January 2016; Submission of Trivago of 14 December 2015; Submission of Twenga of 16 December 2015; Submission of VfT of 16 December 2015; Submission of Visual Meta of 23 December 2015; Submission of Yelp of 11 December 2015.

41                 Commission letters to 1PlusV of 22 September 2016, to Acheter moins cher of 22 September   2016, to BDZV and VDZ of 10 October 2016, to BEUC of 7 October 2016, to Company AC of 22 September 2016, to Deutsche Telekom of 30 September 2016, to ETTSA of 22 September 2016, to Expedia of     22 September 2016, to FairSearch of 30 September 2016, to Foundem of 22 September 2016, to Getty of 22 September 2016, to HolidayCheck of 22 September 2016, to ICOMP of 13 October 2016, to Kelkoo of 22 September 2016, to MIH of 30 September 2016, to News Corp of 22 September 2016, to Nextag of 22 September 2016, to Odigeo of 22 September 2016, to Promt of 12 October 2016, to TradeComet of 10 October 2016, to TripAdvisor of 22 September 2016, to Trivago of 22 September 2016, to Twenga of 22 September 2016, to VG Media of 22 September 2016, to Visual Meta of           22 September 2016, and to Yelp of 22 September 2016.

42                 Submission  of  BEUC  of  4  November  2016;  Submission  of  Company  AC  of  20   October  2016; Submission of FairSearch of 10 November 2016; Submission of Foundem of 19 October 2016; Submission of Getty of 31 October 2016; Submission of ICOMP of 10 November 2016; Submission of Kelkoo of 4 November 2016; Submission of NewsCorp of 2 November 2016; Submission of Promt of 9 November 2016;  Submission  of  TradeComet  of  7  November  2016;  Submission  of  VG  Media  of 4 November 2016; Submission of Yelp of 5 November 2016.

43                 In this Decision, all references to Google's SSO Response and LoF Response should be taken to refer to the joint responses submitted by Google and Alphabet. Similarly, all references to other submissions filed  by  Google  should  be  taken  to  refer  to  joint  submissions  by  Google  and  Alphabet  since  14 July 2016, the date on which the Commission initiated proceedings against Alphabet pursuant to Article 2(1) of Regulation No 773/2004.

44                 SSO Response, paragraphs 372-374.

45                 SSO Response, paragraph 377.

46                 SSO Response, paragraph 377.

47                 SO Response, paragraphs 233 and 234.

48                 SSO Response, paragraphs 81 to 92.

49                 SSO Response, paragraph 379 and footnotes 402-405.

50                 Joined Cases T-191/98, T-212/98 to T-214/98, Atlantic Container Line AB and    Others v Commission, EU:T:2003:245, paragraphs 349-359, Case T-38/02, Groupe Danone v Commission, EU:T:2005:367, paragraph 66; Case T-286/09, Intel v Commission, EU:T:2014:547, paragraph 619.

51                 Case T-286/09, Intel v Commission, EU:T:2014:547, paragraph 614.

52                 Case T-286/09, Intel v Commission, EU:T:2014:547, paragraphs 614 and 620.

53                 Case T-286/09, Intel v Commission, EU:T:2014:547, paragraphs 621 and 1646.

54                 SO Response, paragraphs 450-455.

55                 SO Response, paragraph 453 and SSO Response, paragraph 384.

56                 Case C-441/07 P, Commission v Alrosa, EU:C:2010:377, paragraphs 92 and 95.

57                 SSO, paragraphs 36-38.

58                 Case T-65/98, Van den Bergh Foods, EU:T:2003:281, paragraph 194.

59                 Case C-441/07 P, Commission v Alrosa, EU:C:2010:377, paragraphs 92 and 95. See also Case T- 491/07, RENV CB v Commission, EU:T:2016:379, paragraphs 461 and 470.

60                 SSO Response, paragraphs 385-387.

61                 Case T-395/94, Atlantic Container Line v Commission, EU:T:2002:49, paragraph 418.

62                 SO, paragraph 487 and SSO, paragraph 177.

63                 SO, paragraph 489.

64                 See by analogy Case T-404/12, Toshiba Corp. v Commission, EU:T:2016:18, paragraphs 42, 47 and 75.

65                 Case T -69/89, Radio Telefis Eireann v Commission, EU:T:1991:39, paragraph 98, upheld on  appeal in Joined Cases C-241/91 P and C-242/91 P, EU:C:1995:98, paragraph 91; Case T-24/90, Automec v Commission, EU:T:1992:97, paragraph 52; Case T-128/98, Aéroports de Paris v Commission, EU:T:2000:290, paragraphs 82-83; Case T-401/08 R, Säveltäjäin Tekijänoikeustoimisto Teosto v Commission, EU:T:2008:501, paragraph 46; Case T-410/08 R, GEMA v Commission, EU:T:2008:502, paragraphs 46 and 49; Case T-411/08 R, Artisjus v Commission, EU:T:2008:503, paragraph 50; Case T- 422/08 R, SACEM v Commission, EU:T:2008:504, paragraphs 41 and 44; Case T-425/08 R, KODA v Commission, EU:T:2008:551, paragraphs 43 and 46; Case T-433/08 R, SIAE v Commission, EU:T:2008:520, paragraph 37; Case T-167/08, Microsoft v Commission, EU:T:2012:323, paragraph 95.

66                 Case T-229/94, Deutsche Bahn v Commission, EU:T:1997:155, paragraph 54; Case    T-219/99, British Airways v Commission, EU:T:2003:343, paragraph 91; Case T-321/05, AstraZeneca v Commission, EU:T:2010:266, paragraph 31.

67                 Case 322/81, NV Nederlandsche Banden Industrie Michelin v Commission,  EU:C:1983:313, paragraph 37; Case T-556/08, Slovenská pošta v Commission, EU:T:2015:189, paragraph 112. See also Commission Notice on the definition of relevant market for the purposes of Community competition  law (“Commission notice on market definition”), OJ C 372, 9.12.1997, p. 5.

68                 Commission Notice on market definition, paragraph 7.

69                 Commission Notice on market definition, paragraph 8.

70                 Commission Notice on market definition, paragraph 13.

71                 Commission Notice on market definition, paragraph 20.

72                 Case T-201/04, Microsoft v Commission, EU:T:2007:289, paragraph 917.

73                 Case T-201/04, Microsoft v Commission, EU:T:2007:289, paragraph 925.

74                 Case T-201/04, Microsoft v Commission, EU:T:2007:289, paragraphs 966-970; Case T-79/12, Cisco Systems, Inc. and Messagenet SpA v Commission, EU:T:2013:635, paragraphs 65-74.

75                 Case  T-342/07,  Ryanair  v  Commission,  EU:T:2010:280,  paragraph  136;  Case T-175/12, Deutsche Börse v Commission, EU:T:2015:148, paragraph 133; Case T-699/14, Topps Europe v Commission, EU:T:2017:2, paragraph 82.

76                 http://www.bing.com. See http://searchengineland.com/microsoft-yahoo-search-deal-simplified-23299, downloaded on 2 April 2015.

77                 http://www.seznam.cz. […]

78                 http://www.yahoo.com.      See                                              http://searchengineland.com/microsoft-yahoo-search-deal-simplified- 23299, downloaded on 2 April 2015.

79                 http://www.aol.com. AOL's search results page contain at the bottom of disclaimer that reads “Powered by Bing™”.

80                 http://www.ask.com.    Ask   operated    its   own   search   technology    until    November   2010.   See http://news.cnet.com/8301-30684 3-20022253-265 html?tag=mncol;4n, downloaded on 6 March 2015. The agreement with Google to provide general search services was extended until 31 March 2020. As stated in the IAC/InterActiveCorporation 8k report filed with the SEC on 26 October 2015: “Pursuant to the Services Agreement, Google will provide the Registrant with, among other things, search advertisements through Google’s AdSense for Search (“AFS”) and AdSense for Content (“AFC”) advertising services, web algorithmic search services through Google’s Websearch Service, and image search services”. The 8k report is available at:http://ir.iac.com/sec.cfm?DocType=&DocTypeExclude=&SortOrder=Date%20Descending&Year=2 015&PageNum=2&FormatFilter=&CIK=&NumberPerPage=10, downloaded on 2 February 2017.

81                 Google's Terms of Service, available at http://www.google.com/intl/en/policies/terms/ see “Privacy and Copyright Protection”, downloaded on 6 March 2015.

82                 Google’s Privacy Policy, available at http://www.google.com/intl/en/policies/privacy/ see “Information we collect” and “How we use information we collect”, downloaded on 6 March 2015.

83                 Microsoft's Privacy and Cookies statement, available    at http://www.microsoft.com/PrivacyStatement, downloaded on 6 March 2015, and Blekko’s Terms of Service and Privacy Policy, available at http://blekko.com/legal, downloaded on 6 March 2015.

84                 Google’s Privacy Policy, available at http://www.google.com/intl/en/policies/privacy/ see “Information we collect” and “How we use information we collect”, downloaded on 6 March 2015.

85                 […] submission […]

86                 Google Supplemental Response to the Preliminary Assessment – AFS – […]

87                 Google, http://www.google.com/about/company/tech html, see “Speed”, downloaded on 27 June 2012.

88                 http://www.webpronews.com/google-launches-new-image-search-design-2010-07,                                                                                                  downloaded  on 27 June 2012.

89                 Google, http://www.google.com/about/company/tech.html, see “Comprehensiveness”,   downloaded on 27 June 2012.

90                 http://www.google.com/about/company/philosophy/, downloaded on 6 March 2015.

91                 http://en.wikipedia.org and http://www.imdb.com.

92                 http://www.nytimes.com and http://www.nature.com/nature/index html.

93                 SO Response, paragraphs 124-127.

94                 Google’s reply to […] the Commission’s request for information […].

95                 By  way of example,  a  search on  https://www.google.co.uk  (the  website  of Google’s general search service in the United Kingdom) for “Martin Luther King” on 24 February 2015 returned 148 000 000 results, with the links on the first Google general search results page relating primarily to either historical, biographical or news-related information sources relating to Martin Luther King. By contrast, a search for “Martin Luther King” at virtually the same moment on https://www.google.co.uk/shopping (the website of Google’s comparison shopping service Google Shopping in the United Kingdom), returned 33 pages of results (i.e. just several hundred results in total), with nearly all the links leading to commercial offerings relating to or linked with Martin Luther King (e.g. books, clothing, audio works etc.). Pages downloaded on 6 March 2015.

96                 Sergey Brin and Larry Page, “The PageRank Citation Ranking: Bringing Order to the   Web”, available at http://ilpubs.stanford.edu:8090/422/1/1999-66.pdf, downloaded on 6 March 2015.

97                 Reply of […] to […] the Commission’s search related request for information […]

98                 Reply of […] to […] the Commission’s search related request for information […]

99                 www.shopzilla.com,      www.leguide.com,      www.idealo.de,      www.beslist nl,     www.kelko.co.uk, www.twenga fr, www.tripadvisor.com, www.yelp.com, www kayak.com, www.easyvoyage.com, www.moneysupermaket.com and www.confused.com. Each of these services may have several different websites.

100               https://www.google.co.uk/intl/en/about/products/, downloaded on 6 March 2015.

101               https://support.google.com/faqs/answer/2987537?hl=en, downloaded on 6 March 2015.

102               Ibid.

103               http://www.comscore.com/Insights/Rankings/comScore-Releases-February-2016-US-Desktop-Search- Engine-Rankings downloaded on 6 February 2017. See also Google's reply to […] the Commission's request for information […]

104               SO Response, paragraphs 421, 422 and 430.

105               The German Monopolies Commission reached a similar conclusion in its June 2015 report entitled “The challenge of digital markets” in which it stated that “General and specialized search engines are partly substitutable. While there are of course certain types of search queries that are typically only entered into general search engines, there are other types of information (such as products, hotels and restaurants) that can be found through general as well as specialized search engines”. See http://www.monopolkommission.de/index.php/en/home/84-pressemitteilungen/285-competition-policy- the-challenge-of-digital-markets.

106               See Table 24.

107               ComScore’s definition of Europe does not coincide with the EU or the EEA. In particular, it includes the Russian Federation and Switzerland. However, the Commission considers that this difference in scope is not substantial enough to significantly alter the meaning of the statistics.

108               http://www.clickz.asia/3592/search is dead  long live  search, downloaded   on 12 January 2012.

109               […] reply to […] the Commission's request for information […].

110               […] reply to […] the Commission's request for information […].

111               SO Response, paragraph 128.

112               SO Response, paragraph 128.

113               Deposition of […].

114               See  for  instance  the   global  average   monthly  market  share   percentages    from  NetMarketShare: http://marketshare.hitslink.com/search-engine-market-share.aspx?qprid=4&qpcustomd=1, downloaded on 7 February 2017. The comparison of the figures published in March 2015 and February 2017 indicates that the six largest providers of general search on mobile devices during that timeframe have consistently been Google, Yahoo, Bing, Baidu, Ask and AOL.

115               A merchant platform (or online marketplace) is an online platform which allows users to buy online items from different sellers without leaving the platform. The Alternative Dispute Resolutions Directive 2013/11/EU defines “online marketplace” as “a service provider […], which allows consumers and traders to conclude online sales and service contracts on the online marketplace’s website”. This definition is consistent with Google's own definition of merchant platforms: “Marketplaces are commerce sites that host items and/or websites of multiple individual sellers on the same domain [and] must provide a way for users to purchase the product online through either a payment service or directly from [the marketplaces'] website”; Google's policy on Marketplaces, Marketplace Sellers and Aggregators        as                                      of                  29                 February                                             2016,            available http://webcache.googleusercontent.com/search?q=cache:ocYClixZj54J:https://support.google.com/merc hants/answer/2796531%3Fhl%3Den&num=1&hl=en&gl=be&strip=0&vwsrc=0, downloaded on 2 March 2016, or at https://web.archive.org/web/20150908070524/https://support.google.com/merchants/answer/2796531?h l=en&ref  topic=3404779, downloaded on 30 June 2016.

116               Reply of […] to […] the Commission's request for information […].

117               Reply  of  […]  to  […]  the  Commission’s  request  for  information  […];  reply  of  […]  to   […]  the Commission’s request for information […]; and submission of Company X […].

118               Reply of […] to […] the Commission’s request for information […].

119               Reply of […] to […] the Commission’s request for information […].

120               Reply of […] to […] the Commission’s request for information […].

121               Google’s submission […].

122               Replies of […] to […] the Commission’s request for information […]; replies of […] to […] and […] to […] the Commission’s request for information […]; reply of […] to […] the Commission’s request for information […].

123               Reply of […] to […] the Commission’s request for information […].

124               Replies of […] to […] the Commission’s request for information […].

125               In its “AdWords policies”  (https://support.google.com/adwordspolicy/answer/6008942?hl=en), Google states that “AdWords enables businesses of all sizes, from around the world, to promote a wide variety of products, services, applications, and websites on Google and across our network”, whereas on its webpage   “About             Shopping              campaigns                             and                    Shopping              ads” (https://support.google.com/adwords/answer/2454022?co=ADWORDS.IsAWNCustomer%3Dfalse&hl =en), Google explains: “If  you're  a  retailer,  you  can  use  Shopping  campaigns”,  downloaded  on 29 May 2017.

126               See   the   requirements   listed   in   Google's   websites   “Requirements   for     Shopping   campaigns” (https://support.google.com/adwords/answer/6275312?hl=en&ref topic=6275320) and “Google Merchant Center Help”, in particular the webpages concerning “Feed specification” (https://support.google.com/merchants/answer/3164023?hl=en&ref topic=3163841), “Feeds overview” (https://support.google.com/merchants/answer/188478?hl=en&ref topic=3163841), “Supported file formats”      (https://support.google.com/merchants/answer/160567?hl=en&ref  topic=3163841),     and “Register a feed” (https://support.google.com/merchants/answer/188475?hl=en&ref topic=3163841), downloaded on 29 May 2017.

127               On          its         webpage         “About          Shopping          campaigns           and          Shopping ads” (https://support.google.com/adwords/answer/2454022?co=ADWORDS.IsAWNCustomer%3Dfalse&hl

=en), Google explains that the results from its Google Shopping campaigns are “more than a text ad-- they show users a photo of your product, plus a title, price, store name, and more”, downloaded on      29 May 2017.

128               Google’s submission […].

129               On          its         webpage         “About          Shopping          campaigns           and          Shopping ads” (https://support.google.com/adwords/answer/2454022?co=ADWORDS.IsAWNCustomer%3Dfalse&hl =en), Google explains that Google Shopping “use your existing Merchant Center product data -- not keywords -- to decide how and where to show your ads”, downloaded on 29 May 2017.

130               Google’s reply to […] the Commission’s request for information […].

131               Replies  of  […]  to  […]  the  Commission’s  request  for  information  […];  reply  of  […]  to […] the Commission's request for information […].

132               Replies of […] to […] the Commission’s request for information […]; replies of […] to […] the Commission’s request for information […].

133               Reply of […] to […] the Commission’s request for information […].

134               SO  Response,  paragraphs  90  and  124-127.  Google  also  claims  that  social  networking sites offer “product search alternatives” (see recital (183)).

135               SO Response, paragraph 124.

136               SO Response, paragraphs 98-122; SSO Response, paragraphs 171-219; LoF Response, paragraphs 181- 214.

137               Reply  of  […]  to  […]  the  Commission’s  request  for  information  […];  replies  of  […]  to […] the Commission's request for information […].

138               Reply  of  […]  to  […]  the  Commission’s  request  for  information  […];  reply  of  […]  to   […]  the Commission’s request for information […]; replies of […] to […] the Commission's request for information […]; reply of […] to […] the Commission's request for information […]; reply of […] to […] the Commission's request for information […]; submission of […]; submission of […]; submission of […].

139               Submission of […]; submission of […]; submissions of […]; reply of […] to […] the Commission's request for information […]; reply of […] to […] the Commission's request for information […].

140               Replies  of  […]  to  […]  the  Commission’s  request  for  information  […];  reply  of  […]  to […] the Commission's request for information […]; replies of […] to […] the Commission's request for information […]; replies of […] to […] the Commission's request for information […].

141               As  noted,  this  definition  is  consistent  with  both the  definition  adopted  in  the Alternative Dispute Resolutions Directive 2013/11/EU and the one publicly used by Google in its policies (see footnote 115). See also reply of […] to […] the Commission’s request for information […]. See also replies of […] to […] the Commission's request for information […]; and reply of […] to […] the Commission's request for information […].

142               Submission of […]; reply of […] to […] the Commission's request for information […]; reply of […] to […] the Commission's request for information […]: as indicated by […], the primary purpose of its merchant platform is to sell the products of its merchants, not to provide a comparison among them: “Our first aim is to promote the offers from our merchants, not to put them in comparison” (translated from the original French version: “Notre but premier est la mise en avant des catalogues de nos marchands mais pas de les mettre en comparaison”). Google also acknowledges that certain merchant platforms, such as Amazon and Fnac, are online retailers that have also decided to include and sell on

their websites third party products: see SO Response (paragraph 118) and Annex 6 to the SO Response (paragraphs 49 and 59).

143               Replies  of  […]  to  […]  the  Commission’s  request  for  information  […];  reply  of  […]  to […] the Commission’s request for information […]; replies of […] to […] the Commission's request for information […]; replies of […] to […] the Commission's request for information […]; reply of […] to […] the Commission's request for information […]; reply of […] to […] the Commission's request for information […]; submission of […]; submission of […]; submission of […]; reply of […] to […] the Commission’s request for information […]; reply of […] to […] the Commission’s request for information […].

144               Replies  of  […]  to  […]  the  Commission's  request  for  information  […];  reply  of  […]  to […] the Commission's request for information […]; submission of […]; submission of […]; submissions of  […].

145               See submissions of […] to […] the Commission's request for information […] and submissions  of […] to […] the Commission's request for information […].

146               Available  at  https://support.google.com/faqs/answer/2987537?hl=en#FAQ,  downloaded  on  30  June 2016.

147               Google's  policy  on  Marketplaces,  Marketplace  Sellers  and  Aggregators  as  of  29  February  2016, available http://webcache.googleusercontent.com/search?q=cache:ocYClixZj54J:https://support.google.com/merc hants/answer/2796531%3Fhl%3Den&num=1&hl=en&gl=be&strip=0&vwsrc=0,downloaded   on 2 March 2016, or https://web.archive.org/web/20150908070524/https://support.google.com/merchants/answer/2796531?h l=en&ref  topic=3404779, downloaded on 30 June 2016.

148               Reply of Google to […] the Commission's request for information […]. See also replies of […]   to […] the Commission's request for information […]; reply of […] to […] the Commission's request for information […]; reply of […] to […] the Commission's request for information […]; Submission of […]; submissions of […]; reply of […] to the Commission’s request for information […].

149               […]

150               […]

151               […]

152               On […] and […], the Commission addressed requests for information to a representative sample of comparison shopping services (see recital (488)). In particular, out of the nineteen relevant respondents:

-            eleven indicated that their comparison shopping services do not compete with merchant platforms, which rather are among their main business partners (replies of […] to […] the Commission’s request for information […]; replies of […] to […] the Commission's request for information […]; reply of […] to […] the Commission's request for information […]); See also submission of […]; submission of […]; submission of […]; submission of […]; reply of […] to […] the Commission’s request for information […];

-            two respondents indicated that their comparison shopping services compete with merchant platforms (reply of […] to […] the Commission’s request for information […]; reply of […] to […] the Commission's request for information […]);

-            the remaining seven respondents provided mixed answers. In particular:

o    […] indicated that “At best, merchant platform (“marketplace”) sites might be considered to compete indirectly with our comparison shopping service, while Google Shopping is clearly a direct competitor. On the contrary, merchant platform sites are potential partners and sources of revenue” (replies of […] to […] the Commission’s request for information […] and […] the Commission's request for information […], emphasis added);

o    […] indicated “we consider merchant platform sites as competitors only by a merchant acquisition perspective” (reply to […] the Commission’s request for information of […], emphasis added);

o    […] indicated that merchant platforms compete with its comparison shopping service “but to a lesser extent”. It also explained that “We typically list those marketplaces (Amazon, ebay, Rakuten, and a few other marketplaces) on our comparison shopping service. However, in addition to the products they sell directly, they typically concentrate on smaller merchants since larger retailers don’t always want their products listed on those marketplaces. Retailers don’t generally consider comparison shopping services as the same since those merchant platforms generally process the order and therefore compete more closely with the retailer. The shopping services launched by the general search services (e.g. Google Shopping) act more like direct comparison to comparison shopping services. In addition, a marketplace such as amazon would prefer to direct the user to purchase a product that they sell themselves and often rely on their marketplace as backfill. Comparison shopping services and Google Shopping are merchant agnostic since we do not do fulfilment of the order” (reply of […] to […] the Commission’s request for information […], emphasis added);

o    […] (reply of […] to […] the Commission’s request for information of […]);

o    […] indicated that its comparison shopping service competes with merchants platforms, although it also noted that the comparison shopping functionality offered by merchant platforms is “rudimentary” (reply of […] to […] the Commission’s request for information […]);

o    […] indicated that it competes with merchant platforms (and online retailers), but only in the context of AdWords bids (reply of […] to […] the Commission’s request for information  […]);

o    […] indicated that “merchant platforms are competition to our shopping services”; however it also explained that […] “work[s] with them extensively as they are attractive partners for us” and “merchant platforms are still limited in the catalogue size that they are able to offer – they can only have products from retailers who have opted to work with them, which means the other 99% of retailers aren't on their platform. For us, it means that when we can aggregate their catalogues with other merchant platforms and merchant feeds, we can offer a wider choice of products to our users” (reply of […] to […] the Commission’s request for information […]).

On […], the Commission also sent a request for information to […], but the reply of […] is not discussed in this Decision because […] indicated that its reply is confidential and has not provided a non-confidential version.

153               On […], the Commission addressed a request for information to a representative sample of merchant platforms, i.e. the “leading merchant platforms” listed by Google in the SO Response (paragraphs 117- 122 and Annex 6, paragraphs 48-65). In particular, out of the nine relevant respondents to the question “do you consider that your merchant platform services compete with comparison shopping services?”:

-            five indicated that their merchant platforms do not compete with comparison shopping services, which they rather consider as business partners and an important source of traffic (replies of […] to […] the Commission's request for information […]; reply of […] to […] the Commission's request for information […]; reply of […] to […] the Commission's request for information of […]);

-            […] responded “[…] competes with comparison shopping services to the extent that both services try to promote third party products […] However, […] rather competes with other online marketplaces offering similar services” (reply of […] to […] the Commission's request for information […], emphasis added);

-            […] responded “Google Shopping – yes” (reply of […] to […] the Commission's request for information […]);

-            […] indicated that its marketplaces compete with comparison shopping services, but it also indicated that its marketplaces have no features in common with comparison shopping services (reply of […] to […] the Commission's request for information […]);

-            […] indicated that it is not yet in a position to take a view regarding the level of competition between merchant platforms and comparison shopping services in the EEA (reply of […] to […] the Commission's request for information […]).

See also reply of […] to […] the Commission's request for information […], where […], on the one hand, indicates that it operates a marketplace that is “comparable to eBay or Amazon” and, on the other hand, clarifies that it “neither operates horizontal (non-sector or domain specific) online search  services nor operates vertical (sector or domain specific) online search services and in particular comparison shopping services”.

154               Replies of […] to […] the Commission’s request for information […]; replies of […] to […] the Commission's request for information […]; reply of […] to […] the Commission's request for information […]; reply of […] to […] the Commission's request for information […]; replies of […] to […] the Commission's request for information […]; reply of […] to […] the Commission's request for information […]; reply of […] to […] the Commission's request for information […]; reply of […] to […] the Commission’s request for information […]; submission of […].

155               Reply of […] to […] the Commission's request for information […]; submission of […];  submission of […].

156               […]

157               […]

158               […]

159               […]

160               […]

161               […]

162               […]

163               […]

164               […]

165               […]

166               […]

167               […]

168               See the final report “Study on the coverage, functioning and consumer use of comparison tools and third-party                      verification             schemes             for            such            tools”,             available at http://ec.europa.eu/consumers/consumer evidence/market studies/docs/final report study  on  compari son  tools.pdf, downloaded on 6 July 2016.

169               Ibid., section 5.4.2.

170               SSO Response, Annex 6 (Bloomreach's presentation “State of Amazon 2016”), page 5. See also replies of […] to […] the Commission’s request for information […]; reply of […] to […] the Commission's request for information […]; reply of […] to […] the Commission's request for information […]; submission of […]; submission of […]; replies of […] to […] the Commission’s request for information […]; reply of […] to […] the Commission's request for information […]; replies of […] to […] the Commission's request for information […]; replies of […] to […] the Commission's request for information […].

171               Replies  of  […]  to  […]  the  Commission's  request  for  information  […];  reply  of  […]  to […] the Commission's request for information […]; submission of […]; submission of […]; submissions of  […].

172               On  the  difference  between  the  retailers  listed  by,  respectively,  comparison  shopping services and merchant platforms, see also replies of […] to […] the Commission’s request for information […];  reply of […] to […] the Commission's request for information […]; reply of […] to […] the Commission's request for information […]; submission of […] ; submission of […]; submission of […]; submission of […]; submission of […].

173               […]

174               […]

175               […]

176               […]

177               […]

178               […]

179               […]

180               […]

181               […]

182               […]

183               Submission of […].

184               Submission of […].

185               Submission of […].

186               Reply of […] to […] the Commission’s request for information […].

187               Replies of […] to […] the Commission's request for information […]. See also reply of […] to  […] the Commission’s request for information […].

188               Replies  of  […]  to  […]  the  Commission’s  request  for  information  […];  reply  of  […]  to […] the Commission's request for information […]; replies of […] to […] the Commission's request for information […]; reply of […] […] the Commission's request for information […]; reply of […] to […] the Commission's request for information of […]; replies of […] to […] the Commission's request for information […].

189               SO Response, paragraphs 98-122; SSO Response, paragraphs 171-219; LoF Response, paragraphs 181- 214.

190               SO Response, paragraphs 98-122; SSO Response, paragraphs 171-219; LoF Response, paragraphs 181- 214.

191               SO Response, paragraphs 101-102; SSO Response, paragraphs 211-212 and Annex 10.

192               SO Response, paragraphs 103-104; SSO Response, paragraph 179 and Annexes 6-7; LoF Response, paragraphs 190-193 and Annex 12.

193               SO Response, paragraphs 105-106; SSO Response, paragraph 179.

194               […]

195               SSO Response, paragraphs 183-190.

196               SO Response, paragraphs 109-112.

197               SSO Response, paragraph 215.

198               See reference at recital (223) to the replies of […].

199               As indicated by […] “merchant platforms are still limited in the catalogue size that they are able to offer – they can only have products from retailers who have opted to work with them, which means the other 99% of retailers aren't on their platform. For us, it means that when we can aggregate their

catalogues with other merchant platforms and merchant feeds, we can offer a wider choice of products to our users” (reply of […] to […] the Commission’s request for information […]).

200               SSO Response, Annex 10.1.

201               SSO Response, Annex 10.2.

202               These findings are consistent with the findings of the studies conducted by […] referred to at recital (223)).

203               […]

204               […]

205               […]

206               SSO Response, Annex 6.

207               LoF Response, paragraph 193 and Annex 21.19.

208               SO Response, paragraph 139; SSO Response, paragraph 179.

209               Moreover, the paper states that merchant platforms “differ from traditional comparison   shopping sites in that price comparisons are not the central focus of these sites”.

210               SO Response, paragraph 104; SSO Response, paragraph 179.

211               LoF Response, paragraph 191 and Annex 12.

212               SO Response, paragraph 105.

213               Reply of […] to […] the Commission's request for information […].

214               […] SO Response, Annex 18, Part Two, document No. 3.

215               […] SO Response, Annex 18, Part Two, document No. 3.

216               Reply of […] to […] the Commission’s request for information […].

217               SSO Response, footnote 195.

218               Reply of […] to […] the Commission's request for information […]. Moreover, in response to […] confirmed that it lists its offers on Google Shopping and in other comparison shopping services. […] also indicates that “the main difference between […]'s market place and comparison shopping services is that comparison shopping services display a large number of offers for one and the same product by several third-party sellers. Those third-party sellers usually provide their own shops and own websites. On […]'s marketplace customers searching for products always stay within the framework of the […] marketplace. As […] is a marketplace for unique and handcrafted products customers may find a large variety of products within one category. E.g. a customer searching for “a scarve” will find a large number of products but all products will differ”.

219               SSO Response, footnote 196.

220               Reply of […] to […] the Commission's request for information […].

221               SSO Response, footnote 197.

222               Reply of […] to […] the Commission’s request for information […].

223               SSO Response, footnote 199.

224               Reply of […] to […] the Commission’s request for information […].

225               Reply of […] to […] the Commission’s request for information […].

226               SSO Response, footnote 201.

227               Reply of […] to […] the Commission’s request for information of […].

228               […]

229               […]

230               […]

231               OFT decision in case OFT1514dec, Hotel online booking.

232               OFT decision in case ME/5882-12, Anticipated acquisition of Priceline.com Incorporated of Kayak Software Corporation.

233               SSO Response, paragraph 186.

234               […]

235               Replies of […] to […] the Commission's request for information […]; replies of […] to […] the Commission's request for information […]; submission of […]; submission of […]; submission of […]; submission of […]; submission of […].

236               Article  14  of  Regulation  (EU)  No  524/2013  of  the  European  Parliament  and  of  the   Council  of 21 May 2013 on online dispute resolution for consumer disputes and amending Regulation (EC) No 2006/2004 and Directive 2009/22/EC (Regulation on consumer ODR), OJ L 165, 18.6.2013, p. 1–12.

237               […]

238               […]

239               […]

240               SO  Response,  Annex  3.  For  the  purposes  of  this  Decision,  all  references  to  361   SO  Response Aggregators should be understood as excluding […] and […] since these websites are not comparison shopping services. […] is the website of a consumer organisation. […] focuses on product reviews. See replies of […] to […] the Commission's request for information […].

241               Reply  of  […]  to  […]  the  Commission's  request  for  information  […];  reply  of  […]  to   […]  the Commission's request for information of […] ; submission of […]; submission of […].

242               Replies  of  […]  to  […]  the  Commission's  request  for  information  […];  reply  of  […]  to […] the Commission's request for information […], […] to […] the Commission's request for information […]; submission of […] ; submission of […].

243               Submission of […]; submission of […].

244               Case T-699/14, Topps Europe v Commission, EU:T:2017:2, paragraph 82.

245               Case  T-342/07,  Ryanair  v  Commission,  EU:T:2010:280,  paragraph  136;  Case T-175/12, Deutsche Börse v Commission, EU:T:2015:148, paragraph 133

246               Case T-699/14, Topps Europe Ltd v Commission, EU:T:2017:2, paragraph 82.

247               SO Response, paragraphs 98-122; SSO Response, paragraphs 171-219; LoF Response, paragraphs 181- 214.

248               Reply of […] to […] the Commission’s request for information […].

249               Replies of […] to […] the Commission’s request for information […].

250               […]

251               […]  reply to  […]  the  Commission’s  request  for  information  […].  See  also  […]  reply to […] the Commission’s request for information […].

252               […] reply to […] the Commission’s request for information […].

253               […] reply to […] the Commission’s request for information […].

254               Case C-27/76, United Brands and United Brands Continental v Commission, EU:C:1978:22, paragraph 65; Case 85/76, Hoffmann-La Roche v Commission, EU:C:1979:36, paragraph 38; Case T-201/04, Microsoft v Commission, EU:T:2007:289, paragraph 229.

255               Case 27/76, United Brands and United Brands Continental v Commission, EU:C:1978:22, paragraph 66.

256               Case 85/76, Hoffmann-La Roche v Commission, EU:C:1979:36, paragraph 41; Case T-65/98,   Van den Bergh Foods v Commission, EU:T:2003:281, paragraph 154.

257               Case 85/76, Hoffmann-La Roche v Commission, EU:C:1979:36, paragraph 41; Case T-139/98, AAMS v Commission, EU:T:2001:272, paragraph 51; Case T-65/98, Van den Bergh Foods v Commission, EU:T:2003:281, paragraph 154; Case T-336/07, Telefónica SA v Commission, EU:T:2012:172, paragraph 149.

258               Case C-62/86, Akzo v Commission, EU:C:1991:286, paragraph 60; Case T-340/03, France  Télécom SA v Commission, EU:T:2007:22, paragraph 100; Case T-336/07, Telefónica SA v Commission, EU:T:2012:172, paragraph 150.

259               Case T-30/89, Hilti v Commission, EU:T:1991:70, paragraph 92; Joined Cases T-191/98,    T-212/98 to T-214/98, Atlantic Container Line and Others v Commission, EU:T:2003:245, paragraph 907; Case T- 66/01, Imperial Chemical Industries v Commission, EU:T:2010:255, paragraph 257; Case T-336/07, Telefónica SA v Commission, EU:T:2012:172, paragraph 150.

260               Case T-219/99, British Airways v Commission, EU:T:2003:343, paragraph 210.

261               Case T-79/12, Cisco Systems, Inc. and Messagenet SpA v Commission, EU:T:2013:635, paragraph 69.

262               Case T-340/03, France Telecom v Commission, EU:T:2007:22, paragraphs 107-108.

263               Case T-79/12, Cisco Systems, Inc. and Messagenet SpA v Commission, EU:T:2013:635, paragraph 73.

264               Case 27/76, United Brands and United Brands Continental  v Commission,    EU:C:1978:22, paragraph 122; Case 85/76, Hoffmann-La Roche v Commission, EU:C:1979:36, paragraph 48.

265               Case 27/76, United Brands and United Brands Continental v Commission,   EU:C:1978:22, paragraphs 91 and 122.

266               Google's submission […].

267               Google's submission […].

268               SO Response, paragraph 438.

269               […] Report for DG COMP […].

270               […]

271               Including per number of queries, users, page views or per number of sessions.

272               Google’s reply to […] the Commission’s request for information […]. The Nielsen data includes data until August 2010.

273               AT  Internet  search  engine  barometer,  available  at  http://www.atinternet.com/en/documents/search- engine-barometer-november-2014/, downloaded on 6 March 2015.

274               The Commission has not therefore relied on outdated data, contrary to Google's claim (SO Response, paragraph 432 and SSO Response, paragraph 374).

275  StatCounter data for 2016, downloaded on 7 February 2017, http:/gs.statcounter.com/.

276 [...]

277 http://www.stateofsearch.com/an-analysis-of-the-battle-between-google-and-seznam-a-close-call/, downloaded on 6 March 2015. Other reports suggest that Google may have overtaken Seznam even earlier, at the end of 2009. See http://gs.statcounter.com/#search engine-CZ-monthly-200901-201110, downloaded on 6 March 2015.

278  Google's submission [...].

279  Google's submission [...].

280  Yearly market share data is unavailable for all EEA countries in 2007.

281  StatCounter data for 2008-2015 as presented in Table 2 of the Supplementary Statement of Objections of 15 April 2015, updated with information for 2016 downloaded on 7 February 2017, http://gs.statcounter.com/

282               StatCounter data for 2011-2014, http://gs.statcounter.com/, downloaded on 3 April 2015.

283               […] reply to […] the Commission's request for information […].

284               […] reply to […] the Commission’s request for information […].

285               […] reply to […] the Commission’s search related request for information […].

286               […] reply to […] the Commission’s search related request for information […].

287               Google’s reply to […] the Commission’s request for information […].

288               […] reply to […] the Commission’s request for information […]; […] reply to […]   the Commission’s request for information […]; and […] reply to […] the Commission’s request for information […].

289               […]  complaint, and […] submission of […].

290               […]  reply  to  […]  the  Commission’s  request  for  information  […];  and  […]  reply  to       […]  the Commission’s request for information […].

291               Google's submission […].

292               Google's submission […].

293               See  http://www.sics.se/~aho/tor/Volker  Tresp  ToR-101125.pdf,  downloaded  on  6  March 2015. As indicated in the document, machine learning was only introduced by general search engines in April 2003, p. 6.

294               Google's form 10-K filing for the year 2015, p. 23, (available on https://abc.xyz/investor/) states that Google's capital expenditures comprise research and development (R&D) investments in areas of strategic focus for Google, as well as investments in their systems, data centres, real estate and facilities, and information technology infrastructure.

295               […] Yahoo's capital expenditures are generally comprised    of purchases of computer hardware, server equipment, furniture and fixtures, real estate, and capitalised software and labour.

296  Sources: [...] official Y ahoo's form 10-K filings available on https://investor.yahoo.net/sec.cfm?DocType=Annual&Year=&FormatFilters downloaded 2 February 2017 For Google, the data are based on Google's form 10-K filings as included in [...] submission, as well as form 10-K filings available on on https://abc.xyz/investor/ and httpsabs and 3 February 2017. https://www.sec.gov/edgar/searchedgar/companysearch.html, downloaded on 2 and 3 February 2017.

297  Of which USD 1.8 billion relating to the purchase of a building.

298  [...]

299  [...]

300               […]

301               Google's Form 10-K Annual Report for the US fiscal year ending 31 December 2013, available at http://www.sec.gov/Archives/edgar/data/1288776/000128877614000020/goog2013123110-k htm, downloaded on 6 March 2015; and Google's Form 10-K Annual Report for the US fiscal year ending  31 December 2016, available at https://abc.xyz/investor/, downloaded on 6 March 2017.

302               Google's submission […].

303               http://news.cnet.com/8301-30684  3-20022253-265 html?tag=mncol;4n, downloaded on 6 March 2015. The     agreement    with    Google      was     extended    until    31           March            2020,   as              stated  in                 the IAC/InterActiveCorporation 8-k report filed with the SEC on 26 October 2015, available here: http://ir.iac.com/sec.cfm?DocType=&DocTypeExclude=&SortOrder=Date%20Descending&Year=201 5&PageNum=2&FormatFilter=&CIK=&NumberPerPage=10, downloaded on 2 February 2017.

304               […] reply to […] the Commission's request for information […]. […] share of the […] general search market decreased from 3.6% in 2005 to 0.3% in December 2014. […].

305               The  latest  version  of  Microsoft’s  general  search  engine  launched  in  June  2009  is called “Bing”. Previous versions were called Live Search, Windows Live Search, and MSN Search.

306               Nielsen rating’s figures as quoted in Google’s reply to […] the    Commission’s request for information […], and StatCounter data, available at http://gs.statcounter.com.

307               […]

308               http://www.crunchbase.com/company/kosmix, downloaded on 6 March 2015.

309               http://searchengineland.com/walmart-buys-former-search-engine-kosmix-to-power-social-and-mobile- shopping-73599, downloaded on 6 March 2015.

310               http://www.msnbc msn.com/id/25884709, downloaded on 6 March 2015.

311               http://searchengineland.com/cuil-launches-can-this-search-start-up-really-best-google-14459, downloaded on 6 March 2015.

312               http://blog.statcounter.com/2008/08/cuil-latest-stats, downloaded on 6 March 2015.

313               http://searchenginewatch.com/article/2050586/Cuil-is-Stone-Cold-Another-Google-Killer-Bites-the- Dust, downloaded on 6 March 2015.

314               https://duckduckgo.com/traffic.html, downloaded on 6 March 2015.

315               https://duck.co/help/results/sources, downloaded on 6 March 2015.

316               http://searchengineland.com/blekko-lays-off-eight-raises-6-million-more-162122,                                                                                                  downloaded                                                                                                                  on 6 March 2015.

317               http://www.comscore.com/Insights/Press-Releases/2013/7/comScore-Releases-June-2013-US-Search- Engine-Rankings, downloaded on 6 March 2015.

318               Google's reply to […] the Commission's request for information […].

319               http://searchengineland.com/goodbye-blekko-search-engine-joins-ibms-watson-team-217633, downloaded on 2 April 2015.

320               SO, paragraphs 233-239; SO Response, paragraph 442.

321               There are a number of reasons why users have the technical ability to switch between   different general search services. First, in order to access a given general search service and to perform a search, a user need only either type the address of a general search service in a browser’s address bar or use a bookmark. Second, the main general search services run efficiently on most web browsers and have similar user interfaces, with the result that once a user has performed searches on one of them, it can perform searches on competing services without incurring substantial learning costs. Third, the speed at which general search services perform their task makes it technically possible to perform the same search on several competing general search services within a short period of time.

322               […] submission […].

323  [...] submission [...]

324  [...]

325  SO Response, paragraph 442.

326  The online survey of more than 4800 adults by Forrester Research carried out in July 2008 indicated that 35% of the users did not multi-home during a particular week (Forrester Research, "Search Loyalty Is Still Hard To Find, Advertising Options Exist Beyond Just Google", 3 February 2009); Google's submission [...]

327  The survey by Performics and ROI Research performed in June 2010 on a panel of 500 users representative of the population of the US suggested that 76% of the respondents that use Google as their primary general search service engine use a competing general search service at most occasionally (Performics and ROI Research, "Search Engine Results Page (SERP) Insights Study", September 2010). Google's submission [...].

328               […] reply to the Commission's request for information […].

329               eBay  blog,   “Shopping.com  Rebrands  to  eBay  Commerce  Network”,  4 April 2013,      available  at https://www.ebayinc.com/stories/news/shoppingcom-rebrands-ebay-commerce-network/.

330               The   verb   “to   google”   exists   in   several   languages.   Examples:   English:   http://www.merriam- webster.com/dictionary/google;                              Dutch:          http://www.vandale nl/gratis- woordenboek/betekenis/nederlands/googelen#.VqEAMEaf8rU; Polish: http://sjp.pl/googlowa%C4%87.

331               In a 2011 study of the most valuable global brands, WPP company Millward Brown Optimor found Google to be the second most powerful brand in the world. In the preceding four years, Google was considered             the               most               powerful              brand               in               the world. http://www.wpp.com/wpp/press/press/default htm?guid={c7fa34c1-d97b-4833-a356-bee9ac16e161}, downloaded on 6 March 2015.

In the 2016 edition of WPP's BrandZ top 100 report, Google has risen to the first position globally. See http://www.wpp.com/wpp/press/2016/jun/08/2016-brandz-top-100-most-valuable-global-brands/ downloaded on 13 February 2017.

In   2016   Google    was   considered   the    second    most    valuable                      global    brand   by   Forbes: https://www.forbes.com/powerful-brands/list/ downloaded on 13 February 2017.

Google was considered the most valuable brand globally for the year 2017 by BrandFinance: http://brandirectory.com/league tables/table/global-500-2017 downloaded on 13 February 2017.

332               Google's reply to […] the Commission's request for information […].

333               First, an eye tracking experiment published in 2007 indicated that participants trusted   Google’s ability to position results by their relevance to the query. When the participants selected a link to follow from Google’s general search result pages, they favoured links in a higher position, even if the abstracts themselves were less relevant. See Pan, Bet al, “In Google we trust: Users’ decisions on rank, position, and relevance”, Journal of Computer-Mediated Communication, 2007, p. 12. Second, a 2013 study by SurveyMonkey indicated that participants preferred general search results labelled with the “Google” name, even if they were in fact results of another general search engine. See https://www.surveymonkey.com/blog/2013/06/26/google-bing-seo-surveys/  and https://docs.google.com/spreadsheets/d/113TWN2OyGSS3nrDPZ9ObCrssqCj  FufJ2mi5BY sNs4/edit #gid=0 (downloaded on 28 March 2017).

334               SO Response, paragraph 437.

335               Case T-210/01, General Electric Company v Commission, EU:T:2005:456, paragraph 297.

336               Case T-228/97, Irish Sugar plc v Commission, EU:T:1999:246, paragraph 112; Joined  Cases C-395/96 P and C-396/96 P, Compagnie Maritime Belge Transports and Others v Commission, EU:C:2000:132, paragraph 37; Case T-201/04, Microsoft v Commission, EU:T:2007:289, paragraph 229.

337               SO Response, paragraph 435.

338               Case T-79/12, Cisco Systems Inc. and Messagenet SpA v Commission, EU:T:2013:635, paragraph 73.

339               NetMarketShare                           figures,                                                    http://marketshare.hitslink.com/mobile-market- share?qprid=61&qptimeframe=M&qpsp=153&qpnp=11, downloaded on 6 March 2015.

340             […]

341  Source: StatCounter data for 2008-2015 as presented in Table 3 of the Supplementary Statement of Objections of 15 April 2015, updated with information for 2016 downloaded on 7 February 2017, http://gs.statcounter.com/

342  With the exception of Croatia, for which the period of reference covers only the period following the accession of Croatia to the EU on 1 July 2013.

343               Case 322/81, NV Nederlandsche Banden Industrie Michelin v Commission,  EU:C:1983:313, paragraph 57; Case C-209/10, Post Danmark, EU:C:2012:172, paragraph 23; Case C-457/10 P, AstraZeneca v Commission, EU:C:2012:770, paragraph 134; Case T-286/09, Intel v Commission, EU:T:2014:547, paragraph 205.

344               Case C-280/08 P, Deutsche Telecom v Commission, EU:C:2010:603, paragraph 230; Case T-336/07, Telefónica SA v Commission, EU:T:2012:172, paragraph 204.

345               Joined  Cases  C-395/96  P  and  C-396/96  P,  Compagnie  Maritime  Belge  Transports  and  Others  v Commission, EU:C:2000:132, paragraph 114; Case C-52/09, Konkurrensverket v TeliaSonera Sverige AB, EU:C:2011:83, paragraph 84.

346               Case C-286/13 P, Dole Food and Dole Fresh Fruit Europe v Commission,   EU:C:2015:184, paragraph 125, where the Court of Justice recently stated that “Article 81 EC, like the other competition rules of the Treaty, is designed to protect not only the immediate interests of individual competitors or consumers but also to protect the structure of the market and thus competition as such”. See in the same vein Case C-202/07 P, France Télécom v Commission, EU:C:2009:214, paragraph 105; Joined Cases C- 501/06 P, C-513/06 P, C-515/06 P and C-519/06 P, GlaxoSmithKline Services and Others v  Commission and Others, EU:C:2009:610, paragraph 63; Case C-52/09, Konkurrensverket v TeliaSonera Sverige AB, EU:C:2011:83, paragraph 24.

347               Case  85/76,  Hoffmann-La  Roche  v  Commission,  EU:C:1979:36,  paragraph  91;  Case   322/81,  NV Nederlandsche Banden Industrie Michelin v Commission, EU:C:1983:313, paragraphs 57 and 70; Case C-62/86, Akzo v Commission, EU:C:1991:286, paragraph 69; Case T-219/99, British Airways v Commission, EU:T:2003:343, paragraph 241, confirmed on appeal in Case C-95/04 P, British Airways v Commission, EU:C:2007:166, paragraph 66; Case C-202/07 P, France Télécom v Commission, EU:C:2009:214, paragraph 104; Case C-280/08 P, Deutsche Telecom v Commission, EU:C:2010:603, paragraph 173; Case C-52/09, Konkurrensverket v TeliaSonera Sverige AB, EU:C:2011:83, paragraph 27 and the case-law cited therein.

348               Case   6/72,   Europemballage   Corporation   and   Continental   Can   Company   Inc. v  Commission, EU:C:1973:22, paragraph 26; Case 85/76, Hoffmann-La Roche & Co. AG v Commission, EU:C:1979:36, paragraph 91; Case 322/81, NV Nederlandsche Banden Industrie Michelin v Commission, EU:C:1983:313, paragraph 57.

349               Case   311/84,   Centre   belge   d'études   de   marché   -   Télémarketing   (CBEM)   v  SA  Compagnie luxembourgeoise  de  télédiffusion  (CLT)  and  Information  publicité  Benelux  (IPB), EU:C:1985:394,

paragraph 27; Case C-333/94 P, Tetra Pak v Commission, EU:C:1996:436, paragraph 25; Case T- 228/97, Irish Sugar plc v Commission, EU:T:1999:246, paragraph 166; Case T-201/04, Microsoft v Commission, EU:T:2007:289, paragraph 1344.

350               Case   C-333/94   P,   Tetra   Pak   v   Commission,   EU:C:1996:436,   paragraph   25;   Case   C-52/09, Konkurrensverket v TeliaSonera Sverige AB, EU:C:2011:83, paragraph 85.

351               Case 6/72, Europemballage and Continental Can v Commission, EU:C:1973:22, paragraph 26; Case C- 280/08 P, Deutsche Telekom v Commission, EU:C:2010:603, paragraph 173; Case C-52/09, Konkurrensverket v TeliaSonera Sverige AB, EU:C:2011:83, paragraph 26.

352               Case T-286/09, Intel v Commission, EU:T:2014:547, paragraph 216.

353               Case T-286/09, Intel v Commission, EU:T:2014:547, paragraph 219.

354               Case C-52/09 Konkurrensverket v TeliaSonera Sverige AB, EU:C:2011:83, paragraph 55.

355               Case C-280/08 P, Deutsche Telekom v Commission, EU:C:2010:603, paragraph 175 and case-law cited there.

356               Case 322/81, NV Nederlandsche Banden Industrie Michelin v Commission,  EU:C:1983:313, paragraph 57, Case T-111/96, ITT Promedia v Commission, EU:T:1998:183, paragraph 139.

357               Case 6/72, Europemballage and Continental Can v Commission, EU:C:1973:22, paragraphs 27 and 29; Case T-128/98, Aéroports de Paris v Commission, EU:T:2000:290, paragraph 170.

358               Case C-549/10 P, Tomra Systems and Others v Commission, EU:C:2012:221, paragraphs 19-22.

359               Case C-549/10 P, Tomra Systems and Others v Commission, EU:C:2012:221, paragraph 68.

360               Case T-321/05, AstraZeneca, EU:T:2010:266, paragraph 347, confirmed on appeal in Case C-457/10 P, AstraZeneca v Commission, EU:C:2012:770, paragraphs 109 and 111. See also Case T-286/09, Intel v Commission, EU:T:2014:547, paragraph 186 (and case-law cited therein).

361               Case C-52/09, Konkurrensverket v TeliaSonera Sverige AB, EU:C:2011:83, paragraph 63; Case T- 286/09, Intel v Commission, EU:T:2014:547, paragraphs 88, 149 and 201.

362               Case C-549/10 P, Tomra Systems and Others v Commission, EU:C:2012:221, paragraph 42; Case T- 286/09, Intel v Commission, EU:T:2014:547, paragraphs 117 and 132.

363               Case T-286/09, Intel v Commission, EU:T:2014:547, paragraph 132.

364               Case C-209/10, Post Danmark A/S v Konkurrencerådet, EU:C:2012:172, paragraphs 40-41.

365               Sergey Brin and Larry Page, “The PageRank Citation Ranking: Bringing Order to the   Web”, available at http://ilpubs.stanford.edu:8090/422/1/1999-66.pdf, downloaded on 6 March 2015.

366               Google, “Webmaster tools: Google Basics”, available at http://support.google.com/webmasters/bin/answer.py?hl=en&answer=70897#content, downloaded on 6 March 2015.

367               Google, “Webmaster Tools: Ranking”, available at http://support.google.com/webmasters/bin/answer.py?hl=en&answer=34432&topic=2370570&ctx=topi      c, downloaded on 6 March 2015.

368               Google, “Webmaster Tools: Webmaster Guidelines”, available at http://support.google.com/webmasters/bin/answer.py?hl=en&answer=35769, downloaded on 6 March 2015. See also Google’s reply to […]  the Commission’s request for information […].

369 downloaded at Google, "Webmaster Tools: Webmaster Guidelines", http://support.google.com/webmasters/bin/answer.py?hl=en&answer=35769, 6 March 2015

370  Google, "Webmaster Tools: Webmaster Guidelines" http://support.google.com/webmasters/bin/answer.py?hl=en&answer=35769, 6 March 2015.

371  Google's reply to [...] the Commission's request for information [...].

372  Google's reply to [...] the Commission's request for information [...].

373               […]

374               […]

375               […]

376               […]

377               […]

378               […]

379               […]

380               […]

381               […]

382               […]

383               […]

384               http://googlewebmastercentral.blogspot.com/2011/08/high-quality-sites-algorithm-launched        html, downloaded on 6 March 2015 and Google's reply to […] the Commission's request for information […].

385               http://searchengineland.com/panda-update-rolling-204313, downloaded on 20 January 2015.

386               […]

387               […]

388               […]

389               “Google          search          and          search          engine          spam”,          21        January         2011, https://googleblog.blogspot.be/2011/01/google-search-and-search-engine-spam html,  downloaded  on 22 April 2016.

390               “Algorithm  change  launched”,  28  January 2011,  https://www.mattcutts.com/blog/algorithm-change- launched, downloaded on 22 April 2016.

391               “Finding         more          high-quality              sites           in                           search”,    24         February          2011, http://googleblog.blogspot.be/2011 02 01 archive html - US launch, downloaded on 22 April  2016.

392               “More         guidance           on          building           high-quality                     sites”,              6            May                 2011, https://googlewebmastercentral.blogspot.be/2011/05/more-guidance-on-building-high-quality html, downloaded on 22 April 2016.

393               SSO Response, paragraphs 75-80.

394               The visibility of  websites in  Google's  general search results  reflects  both the  triggering (i.e.   rate of appearance) of websites in Google's general search results pages and their ranking.

395               The trigger rates are defined as the proportion of queries/keywords for which a particular website or service is displayed (“triggered”) among generic search results on Google's general search results pages.

396               […]

397               […]

398               The weekly Visibility Index developed by Sistrix follows the changes in the ranking of  individual sites and a large number of keywords in generic search results on Google's general search results pages. The Index reflects in an aggregate form both the triggering (i.e. rate of appearance) and ranking of sites in generic search results on Google's general search results pages, and it is based on an automatic check of the rankings of millions of keywords every week in more than 10 countries, enriched with additional data and a specially developed algorithm to calculate an individual website's visibility in generic search results    on    Google's    general    search   results   pages.  As    indicated    in    a   Sistrix   blogpost (https://www.sistrix.com/blog/uk-top-100-domains-the-most-visible-websites-in-google-co-uk/, downloaded on 18 May 2017), for the Visibility Index, Sistrix uses a keyword set of  1 000 000 keywords (search phrases) and crawls the first 100 search results on Google's country domains, once a week for both Desktop and Mobile Search. The keywords are chosen in such a way that they reflect the average search behaviour for each country. Week after week, Sistrix obtains 200 million points of data (2x (1 million keywords x the first 100 search results)) which they can attribute to the individual domains. The results are then weighed by position and search volume for each keyword, so that a ranking on the 10th spot for a keyword with a high search volume has a higher bearing on the calculations. Sistrix also incorporates the differing click-through rates for different keywords and then accumulates all the values for all the rankings found for a domain, which results in the Sistrix Visibility Index. The estimation of the click-through rate depends only on the visibility of the generic search  result in Google's general search results page (including its rank and the layout of the page), and not on the website (“domain”) that it links to.

399               Google submission […] and Google's reply to the Commission's request for information […].

400               The SSO complemented the analysis of the SO for the United Kingdom, which had taken into account a sample of six comparison shopping service, with five additional comparison shopping services. Graph 2 takes account of all the comparison shopping services analysed both in the SO and the SSO. The comparison shopping services in the Commission's sample represent approximately […]% of traffic from Google's general search results pages to the 361 SO Response Aggregators active in the United Kingdom in 2014 (see also recital (488)).

401               […]

402               Google submission […] and Google's reply to the Commission's request for information […].

403               The  comparison  shopping  services  in  the  Commission's  sample  represent  approximately […]% of traffic from Google's general search results pages to the 361 SO Response Aggregators active in Germany in 2014. See also recital (488).

404               Google submission […] and Google's reply to the Commission's request for information […].

405               The  comparison  shopping  services  in  the  Commission's  sample  represent  approximately […]% of traffic from Google's general search results pages to the 361 SO Response Aggregators active in France in 2014. See also recital (488).

406  Google submission […] and Google's reply to the Commission's request for information […].

407  Google submission […] and Google's reply to the Commission's request for information […].

408 (…)

409 « EC-featured PC sites » stand for the sample of comparison shopping services the evolution of the total traffic of which was analysed by the Commission in the SO.

410               […]

411               The trigger rates were provided by Google in its reply to […] the Commission's request for information […] and were calculated as a percentage of product queries for which a competing  comparison shopping service is displayed (“triggered”) among generic search results on Google's general search results pages. For the purpose of this Decision, unless specified otherwise, product queries are understood as used by Google for the calculations set out in paragraph 198 of the SO Response and for the data provided by Google in its reply to […] the Commission's request for information […].

412               Google's reply to […] the Commission's request for information […].

413               The individual trigger rates in the United Kingdom, Germany and France have been calculated by the Commission on the basis of Sistrix data. While Sistrix neither has its own definition of product queries, nor follows all product queries as defined by Google, it does calculate for each website that it tracks a number of keywords for which the website is triggered in the top-10 and top-100 ranked generic search results on Google's general search results pages. In order to define a proxy for product queries for the purpose of this analysis, the Commission used the number of keywords for which the Amazon national domains are triggered in the top-10 and top-100 ranked generic search results on Google's general search results pages. This was considered by the Commission as the most appropriate proxy in terms of keyword-sets given the wide coverage of Amazon that could ensure a relatively stable benchmark for the number of product-related keywords. The Commission then analysed the triggering of competing comparison shopping services in generic search results on Google's general search results pages relative to  that  of  the  corresponding  Amazon  domain  for   the  period  between  2   August  2010  and          2 December 2016 by dividing the number of keywords for which they are triggered in the top-10 and top-100 ranked generic search results on Google's general search results pages by the number of keywords Amazon's national domain in the same country is triggered in the top-10 and top-100 ranked generic search results on Google's general search results pages. The comparison shopping services analysed are the top comparison shopping sites competing with Google in terms of volume of traffic identified by ComScore in October 2014 and May 2015 in the United Kingdom, Germany and France.

414               The Sistrix data for the ten other countries in which the Conduct takes place was either unavailable for a representative reference period or, given the more recent launch of the Amazon subdomain in certain

countries and the use of the number of keywords triggering Amazon as a proxy for product queries, did not allow for the same type of analysis of the trigger rates as for the United Kingdom, Germany and France.

415  Google submission [...] and Google's reply to the Commission's request for information [...].

416  The data extracted by the Commission from Sistrix considers the average position for all rankings found for the domain for the current calendar week when the data extraction was made (i.e. 16 November 2015 and 13 February 2017), based on the top-100 generic results. The average ranking based on Sistrix data takes account of instances where the given website is ranked at or above the 100th generic search result. It is therefore a conservative estimate since it does not take account of all instances where the website was ranked beyond the 100 rank.

417  [...]

418  [...]

419 Google submission [...] and Google's reply to the Commission's request for information [...].

420               The  value  '0'  in the  tables  containing average  ranking  data,  as  extracted  from Sistrix,  reflects the situations where the average ranking could not have been observed. The marking 'n/a' stands for 'non applicable' and reflects the situations where the site in question was not active in that particular reference period.

421               Data extracted on 14 December 2015.

422               Data extracted on 14 December 2015.

423               Data extracted on 14 December 2015.

424               […]

425               https://support.google.com/adwords/answer/2454022?hl=en, downloaded on 6 March 2015.

426               […] submission […].

427               Company X's submission […].

428               Google's reply to […] the Commission's request for information […].

429               […]

430               […] reply to […] the Commission's request for information […].

431               […] reply to […] the Commission's request for information […].

432               […] reply to […] the Commission's request for information […].

433               […]

434               […]

435               […]

436     […]

437 Google's reply to […] the Commission's request for information […].

438 Google's reply to […] the Commission's request for information […].

439 Google’s submission […]

440 Google’s submission […]

441               Google's reply to […] the Commission's request for information […].

442               SO Response, […].

443               Google's reply to […] the Commission's request for information […]. In this regard,  Google's founders have stated that: “[S]ince it is very difficult even for experts to evaluate search engines, search engine bias is particularly insidious ... For example, a search engine could add a small factor to search results from ‘friendly’ companies, and subtract a factor from results from competitors. This type of bias is very difficult to detect but could still have a significant effect on the market”; Sergey Brin and Larry Page, The Anatomy of a Large-Scale Hyper-textual Web Search Engine (1998), available at http://infolab.stanford.edu/~backrub/google.html, downloaded on 6 March 2015.

444               Google's submission […].

445               Google's submission […].

446               Google’s reply to […] the Commission’s request for information […].

447               Google’s reply to […] the Commission’s request for information […].

448               Google’s reply to […] the Commission’s request for information […].

449               It can also be positioned at the top of other Google general search results pages.

450               SO Response, paragraph 215; SSO Response, paragraphs 62-63; LoF Response, paragraph 100.

451               The  Commission  has  compared  the  percentage  of  product  queries  for  which  Google displayed  a Shopping Unit in 2014 (SO Response, paragraph 215) with the range of trigger rates for all 361 SO Response Aggregators (out of the domains and sites covered by Google in Annex 3 of the  SO Response) on the first generic search result, or within the first four generic search results on Google's first general search results page in November and December 2014 (Google's reply to […] the Commission's request for information […]). Google could not extract the relevant trigger rates for the competing comparison shopping services prior to November 2014.

452               […]

453               SO Response, paragraph 218.

454               SO Response, paragraph 219.

455               Company X's submission […].

456               SO Response, paragraphs 223-229.

457               SO Response, paragraphs 230-239.

458               SO Response, paragraph 212 and paragraphs 222-225 and 230-239. SSO Response, paragraphs 81-88.

459               […]

460               […]

461               SSO Response, paragraph 96.

462               SO Response, paragraphs 183-191.

463               In the same way, generic search results leading to competing comparison shopping services are not comparison shopping services in themselves.

464  Google's reply to [...] the Commission's request for information [...].

465 SO Response, paragraphs 226-229.

466  Even for the links in Product Universals that went directly to merchants, Google Shopping appears to have benefited since merchants listed in Product Universals are those who signed in for Google Shopping and therefore contribute to the promotion of Google Shopping. Moreover, the results of the Product Universals were subsets of the specialised results from Google Shopping. This Decision, however, does not take a definitive view on these matters.

467  Google's reply to [...] the Commission's request for information [...]. The different reference periods considered are the following: the United Kingdom: January 2008 January 2013, Germany: January 2008-January 2013, France: October 2010 January 2013, the Netherlands: May 2011 January 2013.

468               […]

469               […]

470               […]

471               […]

472               […]

473               […]

474               […] reply to […] the Commission's request for information […].

475               Screenshots downloaded by the Commission on 20 April 2016.

476               From  the  Google  Merchant  blog,  “Google  Shopping  Arrives  in  Australia,  Italy,  Spain,    and  the Netherlands”, 29 April 2011, http://googlemerchantblog.blogspot.be/2011/04/google-shopping-arrives- in-australia html, downloaded on 22 April 2016.

477               “Connect,   Click   and   Convert   Around   the   World   with   Google   Shopping”,   21      May   2013, http://adwords.blogspot.be/2013/05/shopping-global html, downloaded on 22 April 2016.

478               Google   –   Solutions   publicitaires,   Google    Shopping,    “Vendre   vos   produits       sur   Google”, https://www.google fr/ads/shopping/, downloaded on 26 April 2016.

479               Google – Solutions publicitaires, Google Shopping, “Mettez les annonces pour une offre produit à votre service”, https://w.google.fr/ads/shopping/getstarted html, downloaded on 26 April 2016.

480               […]

481               […]

482               […]

483               […]

484               […]

485               SO Response, paragraph 77.

486               “Building        a        better    shopping     experience”,      31         May          2012: http://googlecommerce.blogspot.be/2012/05/building-better-shopping-experience html, downloaded on 21 April 2016.

487               “Sell    your     products    on     Google”,             https://w.google.co.uk/ads/shopping/,             downloaded          on 18 May 2016.

488               […]

489               “Get  Product  Listing  Ads  to  work  for  you”, https://w.google.co.u/ads/shopping/getstarted html, downloaded on 18 May 2016.

490               “Google  Product  Search  To  Become  Google  Shopping,  Use  Pay-To-Play  Model”,  31 May 2012: http://searchengineland.com/google-product-search-to-become-google-shopping-use-pay-to-play- model-122959, downloaded on 21 April 2016.

491               “Google      Product      Search      Changing      Into       Google                     Shopping”,   4                 June               2012: http://www.seochat.com/c/a/search-engine-news/google-product-search-changing-into-google- shopping/, downloaded on 22 April 2016.

492               “Google        Product        Search        is         Now        Google         Shopping”,               4                    June               2012:

https://www.searchenginejournal.com/google-product-search-is-now-google-shopping/56221/, downloaded on 22 April 2016.

493               “The Evolution of Google Shopping”, 19 August 2012: http://larryweeks.com/2012/08/19/the-evolution- of-google-shopping/, downloaded on 22 April 2016.

494               “How     to      get      your      products      noticed      on      Google               Shopping”, 6           September    2013, http://www.alphr.com/realworld/384016/how-to-get-your-products-noticed-on-google-shopping, downloaded on 12 May 2016.

495               LoF Response, Annex 21.16, D. Sullivan, “The Test Begins: Do Google Shopping & Other Shopping Search Engines Give You The Best Deals”, SearchEngineLand, 5 December 2013, available at: http://searchengineland.com/do-shopping-search-engines-give-you-best-deals-178980, downloaded on 12 May 2017.

496               “Google   Shopping   est   disponible   en   Belgique:   comment   démarrer?”,   9       December   2013, http://www.universem.be/fr/google-shopping-belgique/, downloaded on 12 May 2016.

497               “Exemple de recherche avec des annonces Google Shopping dans la SERP”.

498               “Sell     Products     Using       Google       Shopping     Ads”,              Jonathan     Ellins,                        26                  June        2014, https://www.hallaminternet.com/sell-products-using-google-shopping-ads/,      downloaded                 on

13 May 2016.

499               “From  Product  Search  to  Paid  Search:  The  Evolution  of  Google  Shopping”,  31      March  2015: http://bidcops.com/blog/from-product-search-to-paid-search-the-evolution-of-google-shopping/, downloaded on 22 April 2016.

500               “Google:     Google      Shopping”,                     http://searchengineland.com/library/google/google-product-search, downloaded on 12 May 2016.

501               SO Response, paragraph 70, LoF Response, paragraphs 41-43.

502               LoF Response, paragraph 37.

503               […]

504               AdWords          Help,          “About          Shopping          Campaigns          and       Shopping         Ads”: https://support.google.com/adwords/answer/2454022, downloaded on 23 May 2017; “Text Ad”: https://support.google.com/adwords/answer/14093?hl=en, downloaded on 23 May 2017 and “What Makes a Product Shopping Ad”: https://support.google.com/adwords/answer/6275294, downloaded on 23 May 2017.

505               AdWords                Help,                  “Requirements                  for                 Shopping Campaigns”: https://support.google.com/adwords/answer/6275312, downloaded on 23 May 2017.

506               AdWords                    Help,                      “Create                      your                      product data”: https://support.google.com/adwords/answer/188473?hl=en, downloaded on 23 May 2017; “About Shopping Campaigns and Shopping Ads”: https://support.google.com/adwords/answer/2454022, downloaded on 23 May 2017.

507               AdWords Help, “Ad formats”:

https://support.google.com/adwords/answer/3333616?hl=en&ref  topic=24937,                                                                                                  downloaded                                                                                                                  on 21 April 2016.

508               AdWords Help, “Requirements for Shopping campaigns”: https://support.google.com/adwords/answer/6275312?hl=en&ref  topic=6275320,                                                                                                  downloaded                                                                                                                  on 21 April 2016.

509               […]

510               “Google  “Comparison”  Units  Get  New  Look;  Change  Highlights  Paid Inclusion In Some Vertical Search Areas”, 30 April 2012: http://searchengineland.com/google-comparison-units-get-new-look- change-highlights-paid-inclusion-in-some-vertical-search-areas-119865, downloaded on 22 April 2016.

511               LoF Response, paragraph 62, footnote 38, SearchEngineWatch, “Product Listing Ads - An Evaluation”,

21       August        2014,        https://searchenginewatch.com/sew/how-to/2361311/product-listing-ads-an- evaluation, downloaded on 12 May 2017.

512               […]

513               […]

514               […]

515               LoF Response, paragraph 62, footnote 39, Annex 21.6, GoDataFeed, “Why You Should Spend  More on Google Shopping vs. Text Ads”, 17 November 2015, http://www.godatafeed.com/blog/why-you-should- spend-more-on-google-shopping/, downloaded on 12 May 2017.

516               […]

517               […] Review of the likely effects of Google's proposed Commitments […].

518               […].

519               SSO Response, paragraph 96.

520               SO Response, paragraphs 183-191.

521               SO Response, Annexes 1.16 and 1.23.

522               SO Response, Annex 1.19.

523               Document provided to the Commission by the FTC; […].

524               […] reply to […] the Commission's request for information […].

525               […] reply to […] the Commission's request for information […].

526               […] reply to […] the Commission's request for information […]: “Pour avoir le plus d’offres possibles, un des arguments clés auprès des ecommerçants est le volume de trafic potentiellement apporté, d’où la nécessité d’avoir un trafic élevé”.

527               […] reply to […] the Commission's request for information […].

528               […] reply to […] the Commission's request for information […].

529               […] reply to […] the Commission's request for information […].

530               […] reply to […] the Commission's request for information […]: “Plus le trafic augmente, plus le chiffre d’affaires croît et plus l’éditeur du service dispose de moyens pour le développer et se distinguer de la concurrence”.

531               […] reply to […] the Commission's request for information […]: “[L]’existence de trafic est  un facteur déterminant pour augmenter la pertinence d’un moteur de recherche vertical. Chez […], nous utilisons notamment différents « signaux » issus des visites telles que le taux de clic sur nos offres, le taux de clic sur les offres d’un marchand général, en fonction des positions des offres dans nos pages de résultats  de recherche. Des investissements conséquents ont été réalisés par […] pour intégrer des signaux «comportementaux» dans l’amélioration de la qualité de nos résultats de recherche”.

532               […] reply to […] the Commission's request for information […].

533               […] reply to […] the Commission's request for information […].

534               […] reply to […] the Commission's request for information […].

535               Company X's reply to […] the Commission's request for information […].

536               […] reply to […] the Commission's request for information […]: “Nous considérons qu’il y a un lien entre le volume global de trafic et la pertinence des réponses apportées aux internautes (non chiffré). En effet, un effet boule de neige est inhérent au système. (…) Notre moteur de recherche est basé sur la pertinence et la popularité. Plus une offre est cliquée meilleure sera sa position dans les résultats de recherche et la pertinence élevée pour les internautes. Plus le nombre d’offres proposées sur nos moteurs est élevé, plus les internautes ont de chances de trouver une offre totalement pertinente”.

537               […] reply to […] the Commission's request for information […].

538               […] reply to […] the Commission's request for information […].

539               […] reply to […] the Commission's request for information […].

540               […] reply to […] the Commission's request for information […].

541               See the results of the study prepared by GEA Internet Project Consulting (IPC) entitled Eyetracking Search Marketing 2009, available at http://www.geaipc.com/blog/wp-content/uploads/gea-ipc- eyetracking-search-marketing.pdf, downloaded on 6 March 2015. See also the UK Competition and Markets Authority's review of existing literature on “Online Search: Consumer and Firm Behaviour”,   7 April 2017, paragraph 1.6(c), LoF Response, Annex 21.19: “[o]n average, the first three links seem to account for 40-65% of the total clicks on desktop devices. On mobile devices, this tendency is even  more accentuated, with the top three links on average accounting for more than 70% of the total clicks. The evidence suggests that […] consumers seem to display an inherent bias to click on links in higher positions”.

542               See  the  blogpost  of  26  January 2008  by SEO  Design  Solutions  entitled  “Search  Behavior & Why Ranking Above the Fold Matters”, available at http://www.seodesignsolutions.com/blog/articles/search- behavior-why-ranking-above-the-fold-matters, downloaded on 6 March 2015.

543               Google's reply to […] the Commission's request for information […].

544               […] submission […]. See also B.J. Jansen, A. Spink, J. Bateman & T. Saracevic, “Real life,  real users, and real needs: A study and analysis of user queries on the Web”, Information Processing and Management, 1 March 2000, Volume 36 Issue 2, p. 207-227, available at http://www.sciencedirect.com/science/article/pii/S0306457399000564, downloaded on 6 March 2015; and S. J. Yang and A. Ghose, “Analyzing the Relationship Between Organic and Sponsored Search Advertising: Positive, Negative, or Zero Interdependence?”, Marketing Science, April 2010, Volume 29 Number 4 p. 602-623, available at http://mktsci.journal.informs.org/content/29/4/602.abstract. Downloaded on 6 March 2015.

545               This tendency could be for instance a developed habit based on the perception that, in most instances in the search history of the user, the general search engine has ranked the most relevant links in the first generic search results.

546               […] submission […].

547               Ibid., footnote 10.

548               […] submission […].

549               Ibid.

550               SO Response, paragraph 270.

551               As indicated in the study, “satisfied click-through rate” is the number of clicks on a given position where users stay on the landing page for at least 30 seconds divided by the total number of searches, see […] submission […].

552               […] submission […].

553               […] submission […].

554               […] submission […].

555               […] reply to […] the Commission's request for information […].

556               […] reply to […] the Commission’s natural search related request for information […].

557               […] reply to […] the Commission’s natural search related request for information […]. This   applies to all traffic and data in this recital.

558               […] reply to […] the Commission's request for information […].

559               […] reply to […] the Commission's request for information […].

560               […] reply to […] the Commission's request for information […].

561               […] reply to […] the Commission's request for information […].

562               Ibid.

563               […] reply to […] the Commission's request for information […].

564               […] reply to […] the Commission's request for information […].

565               […] reply to […] the Commission's request for information […].

566               […] reply to […] the Commission's request for information […].

567               […] reply to […] the Commission's request for information […].

568               […] reply to […] the Commission's request for information […].

569               […] reply to […] the Commission's request for information […].

570               […] reply to […] the Commission's request for information […].

571               […]  reply  to  […]  the  Commission's  request  for  information  […].  For  32  of  those instances  […] provided detailed numerical information.

572               The Commission's analysis is based on the underlying data submitted by Google in reply to […] the Commission's request for information […] concerning traffic from Google's general search results  pages to the 361 SO Response Aggregators.

573               […].

574               The Sistrix Visibility Index indicates the changes in the ranking of results for individual websites in generic search results on Google's general search results pages and reflects both their triggering and ranking. Data obtained by the Commission from www.sistrix.com. See footnote 398, for more details  on the methodology for the Sistrix Visibility Index.

575               […].

576               SSO Response, paragraph 80 and footnote 139, and LoF Response, paragraph 101.

577               […]

578               […]

579               […]

580               The Commission's analysis is based on the underlying data submitted by Google in its reply to  […] the Commission's request for information […] concerning traffic from Google's general search results  pages to the 361 SO Response Aggregators, as updated by Google in its reply to […] the Commission's request for information […] and complemented […] with data covering the period November- December 2016.

581               In addition to the visits to the standalone websites of the relevant comparison shopping services, the total traffic to competing comparison shopping services also includes clicks on all the items (including, where applicable on links leading to a standalone comparison shopping service's website) included in the product listing units displayed on the general search results pages of […], as well as in the product listing units displayed on third party websites by […], when these units are displayed in reply to a query. This is without prejudice to the question whether one or several of these product listing units can be considered as emanations of comparison shopping services. This is therefore a conservative estimate. Source: replies of […] to the Commission's request for information […]. The responses indicate that the proportion of clicks on all the items displayed in these product listing units account for a minor proportion of the total traffic of the associated competing comparison shopping services. In the case of product listing units displayed on third party websites by […], the clicks taken into account in the Commission's analysis cover both query and non-query related product listing units. This is because the click data provided by […] did not differentiate between the two solutions. See […] reply to […] the Commission's request for information […]. The Commission has also requested the click data on product listing units displayed on third party websites by […]. That data was, however, not included in the Commission's analysis because […] indicated that the product listing units it provides are not query- based. See […] reply to the Commission's request for information […]. The Commission has also requested the click data on product listing units displayed on the general search results pages of […]. That data was, however, not included in the Commission's analysis because (i) […] provided incomplete data; and (ii) […] did not agree to disclose its data to Google's external advisors in a data room (see also footnote 733). See replies of […] to the Commission's request for information […].

582               Traffic data provided by […] (reply of […] to the Commission's request for information […]) is not included in the Commission's analysis because […] did not agree to disclose its data to Google's external advisors in a data room. Instead, consistent with what Google had done in the SSO Response (Annex 11) and in the LoF Response (Annex 17), the Commission has included traffic to […] from Google's general search results pages in the analysis (see also footnotes 608 and 745).

583               The triggering rate of the Shopping Unit is defined as the percentage of queries for which a Shopping Unit appeared out of the total number of queries performed on Google in the relevant country.

584               Google's reply to […] the Commission's request for information […], as updated in Google's reply to […] the Commission's request for information […], and complemented […] with data covering the period November-December 2016. The average triggering rate of the Shopping Unit in 2016 was […]% in the United Kingdom, […]% in Germany, […]% in the Netherlands and […]% in France.

585               Google's submission […].

586               Google's reply to […] the Commission's request for information […], as updated in Google's reply to […] the Commission's request for information […], and complemented […] with data covering the period November-December 2016.

587               The 12 competing comparison shopping services in the sample are the following: […]. Responses of […] to the Commission's request for information […]; […] response to the Commission's request for information […].

588               The 9 competing comparison shopping services in the sample are the following: […]. Responses of […] to the Commission's request for information […].

589               The  6  competing  comparison  shopping  services  in  the  sample  are  the  following:   […]       to  the Commission's request for information […].

590               The 8 competing comparison shopping services in the sample are the following: […]. Responses of […] to the Commission's request for information […].

591               SO Response, paragraph 242.

592               The percentages were calculated based on traffic data in Google's submission […]. Traffic data   of […] and […] were not taken into account as […] is the website of a consumer organisation and […] focuses on product reviews. See footnote 240. The SSO contained a clerical error relating to the percentage of traffic from Google's general search results pages to the 361 SO Response Aggregators in the United Kingdom (see SSO, paragraph (86)(k)(1)). The error has been corrected in the Decision. The corrected percentage is more favourable to Google ([…]).

593               Pulse,  “Google  Velocity:  Froogle  and  Local  are  dying  while  Video  and  Blog  are  surging”, 20 February 2007, available at http://blog.compete.com/2007/02/20/google-properties-froogle-local- video-blog/, downloaded on 6 March 2015.

594               Google's reply to […] the Commission's request for information […].

595               Google's reply to […] the Commission's request for information […].

596               Google's reply to […] the Commission's request for information […].

597               SO Response, paragraph 290.

598               Google's reply to […] the Commission's request for information […].

599               […]

600               […]. Document provided the Commission by the FTC. […].

601               Google    Help,      “About    Shopping     campaigns    and     shopping     ads”,     undated,    available at https://support.google.com/adwords/answer/2454022?hl=en, downloaded on 6 March 2015.

602               Trigger rates as calculated by Google in its reply to […] the Commission's request for information […], for the period preceding January 2013, and its reply to […] the Commission's request for information […], for the period after January 2013.

603               Traffic to Google Product Search  is based  on the sum of  the clicks on  links  that led  the user    to the standalone Google Product Search website. Source: Google's reply to […] the Commission's request for information […]. Traffic to Google Shopping is based on the sum of the clicks on: (i) both links that lead the user to the standalone Google Shopping website, source: Google's reply to […] the Commission's request for information […]; and (ii) links that lead the user directly to the website of one of the merchants whose offer is displayed in the Shopping Units, source: Google's reply to […] the Commission's request for information […].

604               Traffic to Google Product Search is based on the sum of the clicks on links that lead the user to the standalone Google Product Search website, including the Shopping menu link displayed at the top of Google's general search results pages. Source: Google's reply to […] the Commission's request for information of […], as updated in Google's reply to […] the Commission's request for information […]. Traffic to Google Shopping is based on the sum of the clicks on links that lead the user to the  standalone Google Shopping website, including the Shopping menu link displayed at the top of Google's general search results pages, and the clicks on links that lead the user directly to a webpage of a merchant. Sources: (i) for clicks on header links, other click elements and links that lead the user directly to the webpage of one of the merchants whose offer is displayed in the Shopping Units: Google's reply to […] the Commission's request for information […], as updated in Google's reply to […] the Commission's request for information […], and complemented […] with data covering the period November-December 2016; for France, Germany, the Netherlands and the United Kingdom  (data on header links and other click elements for the period February-April 2013 was based on Google's reply to […] the Commission's request for information […]); (ii) for clicks on menu links: Google's reply to […] the Commission's request for information […], as updated in Google's reply to […] the Commission's request for information […], and complemented […] with data covering the period November-December 2016.

605               The analysis concerns the following countries and the following reference periods: the United Kingdom (reference period: January 2008-December 2016), Germany (reference period: January 2008-December 2016), France (reference period: October 2010-December 2016), the Netherlands (reference period: May 2011-December 2016), Italy (reference period: May 2011-December 2016, Spain (reference period: May 2011-December 2016), Denmark (reference period: November 2013-December 2016), and Poland (reference period: November 2013-December 2016). The reference periods correspond to the dates as of which Google began to implement the Conduct. For the 361 SO Response Aggregators, the Commission's analysis is based on the underlying data submitted by Google in its reply to […] the Commission's request for information […] concerning traffic from Google's general search results  pages to the 361 SO Response Aggregators, as updated by Google in its reply to […] the Commission's request for information […] and complemented […] with data covering the period November- December 2016.

606               Traffic to Google Product  Search  is based  on the sum of  the clicks on  links  that led  the user   to the standalone Google Product Search website, including the Shopping menu link displayed at the top of Google's general search results pages. Source: Google's reply to […] the Commission's request for information […], as updated in Google's reply to […] the Commission's request for information […]. Traffic to Google Shopping is based on the sum of the clicks on links that lead the user to the  standalone Google Shopping website, including the Shopping menu link displayed at the top of Google's general search results pages, and the clicks on links that lead the user directly to a webpage of  a merchant. Sources: (i) for clicks on header links, other click elements and links that lead the user directly to the webpage of one of the merchants whose offer is displayed in the Shopping Units: Google's reply to […] the Commission's request for information […], as updated in Google's reply to […] the Commission's request for information […], and complemented […] with data covering the period November-December 2016; for France, Germany, the Netherlands and the UK (data on header links and other click elements for the period February – April 2013 was based on Google's reply to […] the Commission's request for information […]); (ii) for clicks on menu links: Google's reply to […] the Commission's request for information […], as updated in Google's reply to […] the Commission's request for information […], and complemented […] with data covering the period November- December 2016.

607               In addition to the visits to the standalone websites of the relevant comparison shopping services, the total traffic to competing comparison shopping services also includes clicks on all the items (including, where applicable on links leading to a standalone comparison shopping service's website) included in the product listing units displayed on the general search results pages of […], as well as in the   product listing units displayed on third party websites by […], when these units are displayed in reply to a query. See further footnote 581.

608               Traffic data provided by […] (reply of […] to the Commission's request for information […]) is not included in the Commission's analysis because […] did not agree to disclose its data to Google's external advisors in a data room. Instead, consistent with what Google had done in the SSO Response

(Annex 11) and in the LoF Response (Annex 17), the Commission has included traffic to […] from Google's general search results pages in the analysis (see also footnotes 582 and 745).

609               Calculation based on the figures in Table 1, Annex 1. The 12 competing comparison  shopping services in the sample are the following: […]. See footnote 587.

610               Calculation based on the figures in Table 6, Annex 1. The 9 competing comparison shopping services in the sample are the following[…]. Responses of […] to the Commission's request for information […]. See footnote 588.

611               Calculation based on the figures in Table 16, Annex 1. The 6 competing comparison  shopping services in the sample are the following: […]. Responses of […] to the Commission's request for information […]. See footnote 589.

612               Calculation based on the figures in Table 11, Annex 1. The 8 competing comparison  shopping services in the sample are the following: […]. Responses of […] to the Commission's request for information […]. See footnote 590.

613               SSO Response, paragraph 114.

614               SSO Response, paragraphs 123-126.

615               […]

616               SSO Response, paragraphs 130-132.

617               SO Response, paragraphs 133-141 and 330-342; SSO Response, paragraphs 132-135.

618               SSO Response, paragraph 179.

619               SSO Response, paragraph 213.

620               SSO Response, paragraphs 195, 199, 201, 205-207.

621               SSO Response, paragraph 204.

622               Ibid.

623               SSO Response, paragraph 214.

624               SO Response, paragraphs 242, 264 and 307; SSO response, paragraphs 116-120 and 367.

625               […]

626               […]

627               […]

628               SSO Response, paragraphs 138-145.

629               SSO Response, paragraphs 146-156.

630               SSO Response, paragraph 132, graphs titled “Countries with Amazon”.

631               […]

632               […]

633               Joined Cases T-67/00, T-68/00, T-71/00 and T-78/00 JFE Engineering v  Commission, EU:T:2004:221, paragraph 312; Case T-191/06 FMC Foret v Commission, EU:T:2011:277, paragraph 133.

634               Joined  Cases  C-239/11 P,  C-489/11 P  and  C-498/11 P  Siemens  v      Commission,  EU:C:2013:866, paragraph 145.

635               Case T-286/09 Intel v Commission, EU:T:2014:547, paragraphs 1312 and 1406.

636               […]

637               […]

638               […]

639               [...]

640               […]

641               […]

642               […]

643               […]

644               SO Response, paragraph 271.

645               SSO Response, paragraph 156.

646               Pulse, “Google Velocity: Froogle and Local are dying while Video and Blog are surging”, 20 February 2007, available at http://blog.compete.com/2007/02/20/google-properties-froogle-local-video-blog/, downloaded on 6 March 2015.

647  Some respondents have included in this category traffic from other paid search channels, such as paid search traffic from Bing or Yahoo. These figures may therefore overestimate the size of traffic from AdWords, albeit to a limited extent.

648  This category covers both traffic from AdWords search advertisements which are displayed directly on Google's own website and traffic from AdWords search advertisements which are displayed via the AdSense for Search programme on third party websites.

649  [...] data contains small inconsistencies, as the sum of traffic from various sources provided in reply to [...] the Commission's request for information [...], as updated by the reply to [...] of the Commission's request for information [...] is slightly different than the total traffic provided in reply to [...] of the same request for information. However, this does not materially affect the calculations since it affects the percentage of generic search traffic from Google's general search results pages by a few percentage points, at most.

650  [...] data contains small inconsistencies, as the sum of traffic from various sources provided in reply to [...] the Commission's request for information [...], as updated by the reply to [...] the Commission's request for information [...] is slightly different than the total traffic provided in reply to [...] of the same request for information. However, this does not materially affect the calculations since it affects the percentage of generic search traffic coming from Google's general search results pages by a few percentage points, at most.

651  [...] data contains small inconsistencies, as the sum of traffic from various sources provided in reply to [...] the Commission's request for information [...], as updated by the reply to [...] the Commission's request for information [...] is slightly different than that provided in reply to [...] of the same request for information. However, this does not materially affect the calculations since, even when taking the value of the total traffic that is the most conservative, this could affect the percentage of generic search traffic coming from Google's general search results pages by a few percentage points, at most. The total traffic considered in this table is the one provided in reply to [...] and the category Others is calculated as the difference between this total and the sum of the first three categories in the table (i.e. generic search traffic from Google's general search results pages, from AdWords and from direct navigation, including apps).

652  [...] has been unable to retrieve accurate visits data for direct navigation. Direct traffic is therefore included in "Other Sources" in the case of [...].

653  [...] has provided traffic data split by origin only for its [...] websites, [...] from January 2012 and its [...] from July 2013. [...] data contains small inconsistencies, as the sum of traffic from various sources provided in reply to [...] the Commission's request for information [...], as updated by the reply to [...] the Commission's request for information [...] is slightly different than the total traffic provided in reply to [...] of the same request for information. However, this does not materially affect the calculations since it affects the percentage of generic search traffic from Google's general search results pages by a few percentage points, at most.

654  2014 figures do not include the month of December. Later on, [...] stopped or has reduced its operations in Europe significantly; see [...] reply to [...] the Commission's request for information [...].

655   [...] data is available as of January 2012 and contains small inconsistencies, as the sum of traffic from various sources provided in reply to [...] the Commission's request for information [...], as updated by the reply to [...] the Commission's request for information [...] is slightly different than the total traffic provided in reply to [...] of the same request for information. However, this does not materially affect the calculations since if affects the percentage of generic search traffic from Google's general search results pages by a few percentage points, at most.

656   [...] data includes data on [...]. As of November 2015, the websites [...] (no longer active) have been integrated within [...]. [...] data contains small inconsistencies, as the sum of traffic from various sources provided in reply to [...] the Commission's request for information [...], as updated by the reply to [...] the Commission's request for information [...] is slightly different than the total traffic provided in reply to [...] of the same request for information. However, this does not materially affect the calculations since, even when taking the value of the total traffic that is the most conservative, this could affect the percentage of generic search traffic from Google's general search results pages by a few percentage points, at most. The total traffic considered in this table is that provided in reply to [...] and the category Others is calculated as the difference between this total and the sum of the first three categories in the table (i.e. generic search traffic from Google's general search results pages, from AdWords and from direct navigation, including apps).

657  Replies of [...] to [...] the Commission's request for information [...], as complemented or updated by the replies to the Commission's requests for information [...]. The request for information was addressed to 18 comparison shopping services that have the highest traffic in the EEA as identified by ComScore in October 2014 and for the UK those that have the highest total Google traffic. The table includes data for those companies that replied and covers all websites for which the respondents provided the relevant split of traffic. Data is, however, not included from [...] because it is not representative of a comparison shopping service with sustainable operations in the EEA. Indeed, between 2011 and 2016, its total traffic in the EEA decreased from [...] in 2011 to [...] in 2016. See the replies of [...] to [...] the Commission's request for information [...], as updated by the replies to the Commission's requests for information [...].

658 The figures for [...] are those included in Table 9 of the SO. These figures have not been contested by Google.

659  [...] submission [...]; [...] reply to [...] the Commission's request for information [...]; [...] reply to the Commission's request for information [...]; [...] reply to the Commission's request for information [...]. [...] reply to [...] the Commission's request for information [...]; [...] reply to [...] the Commission's request for information [...].

660 The amount of traffic from online search advertising on other online platforms (non-Google) is small. It is covered in the category "Other Sources" in Table 24 which also includes several other categories. For example, [...] states that the amount of traffic from online search advertising on other online platforms was too small to give meaningful calculations for the return on investment. See [...] reply to [...] the RFI [...].

661               See  S.  Yang  and  A.  Ghose,  “Analyzing  the  Relationship  Between  Organic and Sponsored Search Advertising: Positive, Negative, or Zero Interdependence?”, Marketing Science, July/August 2010, Volume 29 Number  4,   p. 602-623,     available at http://mktsci.journal.informs.org/content/29/4/602.abstract, downloaded on 6 March 2015.

662               Groupon Director of Product Management Gene McKenna assessed the impact of Google's changing

the display format of AdWords results and found that three months after such changes, most users reverted to their prior behaviour of scrolling past the AdWords results to click on the generic search results.   http://searchengineland.com/new-google-ad-format-fooled-users-3-months-199818, downloaded on 6 March 2015.

663               See for instance […] submission […].

664               […]

665               […]

666               […] replies to […] the Commission's AdWords related request for information […].

667               […] reply to […] the Commission's AdWords related request for information […].

668               […] reply to […] the Commission's request for information […].

669               […]

670                    Google's submission […]

671See […]      submission    […]

672    […]

673               […]

674               […]

675               As regards […], 2014 figures do not include the month of December.

676               The analysis has been carried out at company level by aggregating traffic data of all comparison shopping websites owned and operated by a given company. The analysis is the net impact on traffic across all websites for a given company because it assumes that resources can be moved across different websites operated in different countries. The five companies (and the corresponding comparison shopping websites) selected for the analysis are […].

677               The  return on investment  of  spending on  AdWords  starts  to  become  negative  after  a certain point because the marginal profitability of traffic from AdWords tends to decrease as the volumes of such traffic increase. See also […] reply to […] the Commission's request for information […].

678               […]

679               Replies to […] the Commission's request for information […].

680               Replies to […] the Commission's request for information […].

681 The conversion rate is calculated by dividing the number of visits which triggered a payment by a partner of a comparison shopping service divided by the total number of visits to the website of that comparison shopping service.

682 [...] data exclude the figures for Italy, Sweden, Denmark and Norway.

683 https://support.google.com/adwords/answer/1722087?hl=en, downloaded on 23 March 2015. See section "How AdWords chooses top ads".

684               […] reply to […] the Commission's request for information […].

685               […] reply to the Commission's request for information […].

686               […] reply to […] the Commission's request for information […].

687               […] submission […].

688 [...] submission [...].

689 [...] reply to [...] the Commission's request for information [...]

690 The Revenue Per Lead is different from the Revenue Per Visit because, during a visit to a comparison shopping service, a user will not necessarily click generally on one link to a merchant partner. Some users may simply go through the information on the comparison shopping service without clicking on any merchant link. Other users may on the contrary click on several merchant links.

691 Graph 77 excludes data for Italy, Sweden, Denmark and Norway for confidentiality reasons.

692               Replies by […] to […] the Commission's request for information […]. Ranges as provided by […].

693               SO Response, paragraph 353.

694               Google  blog,  “Websites  that  may  merit  a  low  landing  page  quality  score”,  18  September  2007, available at http://adwords.blogspot.fr/2007/09/websites-that-may-merit-low-landing.html,downloaded on 24 May 2017.

695               SSO Response, paragraphs 314-315.

696               […] submission […]: “Google´s final proposal is to focus on mobile apps. This is totally misleading for two reasons. First, in vertical markets like Product Search or Local Search, you already need strong brand awareness as a starting point. Without brand awareness, nobody would download such an app”.

697               […]

698 The analysis was carried out for those comparison shopping services, selected by Google, that were included in the graph at paragraph 315 of the SSO Response, and for which data was available in the replies to Question 1 of the Commission's request for information of 20 December 2016, i.e. replies of [...]

699 [...] reply to [...] the Commission's request for information [...].

700 [...] reply to [...] the Commission's request for information [...].

701 [...] reply to [...] the Commission's request for information [...].

702 SO Response, paragraph 355.

703 Translation of [...] submission [...]: “il n'est pas possible de développer un service de comparaison de prix sans bénéficier du trafic apporté par un moteur de recherche généraliste. En effet, les habitudes des utilisateurs sont tellement prégnantes que quelle que soit l'attractivité d'un service de comparaison de prix, les consommateurs démarreront toujours leur recherche sur un moteur généraliste. (..) De la même manière que les utilisateurs ne vont pas directement sur un moteur de shopping pour rechercher le prix d'un produit depuis leur Desktop, ils ne lancent pas une application mobile, mais lancent leur navigateur Web qui les dirige par défaut vers Google. Il n'y a pas de possibilité d'acquérir du trafic de product search sur le mobile en dehors du search de Google".

704 SSO Response, paragraphs 316-318, and LoF Response, paragraphs 285-287.

705               […]

706               […] reply to […] the Commission's request for information […].

707               Google provided visit metrics for: (i) visits from the menu link on the top bar of Google's general search results page; (ii) visits from the header link of the Shopping Unit; and (iii) visits from other links within the Shopping Unit. Google estimates that visits from other sources (including direct navigation to Google Shopping) amounted to […]. See Google's reply to […] the Commission's request for information […].

708               SO Response, paragraphs 356-359, and LoF Response, paragraph 282.

709               […] submission […].

710               […] reply to […] the Commission's request for information […].

711               […] submission […].

712               […] reply to […] the Commission's request for information […].

713               Since April 2014, […] has made cumulated losses of approximately EUR […] . See […] reply to […] the Commission’s request for information […]. […] turnover in 2014 compared to 2013 decreased by 14% to EUR 39 million. See […]. Since October 2013, […] has closed its London and Irish offices and has reduced its Hamburg office. […] has also ceased work on its websites in Italy, Spain and the Netherlands and will cease work on its websites in France and in the United Kingdom. […] response to […] the Commission's request for information […].

714               Pulse, “Google Velocity: Froogle and Local are dying while Video and Blog are surging”, 20 February 2007, available at http://blog.compete.com/2007/02/20/google-properties-froogle-local-video-blog/, downloaded on 6 March 2015.

715               The  Conduct  has  similar  effects  to  the  tying  practice  whereby  Microsoft  used  the  Windows  PC operating system as a distribution channel to ensure for itself a significant competitive advantage on the streaming media player market. See by analogy Case T-201/04, Microsoft v Commission, EU:T:2007:289, paragraphs 982 and 1088.

716               SO Response, paragraph 366.

717               SO Response, paragraphs 94, 95 and 367.

718               Case C-52/09, Konkurrensverket v TeliaSonera Sverige AB, EU:C:2011:83, paragraph 64; Case C- 549/10 P, Tomra Systems and Others v Commission, EU:C:2012:221, paragraph 79; Case T-336/07 Telefónica SA v Commission, EU:T:2012:172, paragraph 272, upheld on appeal in Case C-295/12 P, EU:C:2014:2062, paragraph 124; Case C-23/14 Post Danmark, EU:C:2015:651, paragraph 66.

719               Case T-203/01 Michelin v Commission, EU:T:2003:250, paragraph 245; Case T-219/99 British Airways v Commission, EU:T:2003:343, paragraph 298; Case T-286/09 Intel v Commission, EU:T:2014:547, paragraph 186.

720               Traffic arbitrage consists of buying and selling paid traffic for a profit.

721               […] submission […] .

722               […] submission […].

723               […] submission […].

724               […] reply […] to […] the Commission's request for information […].

725               […] reply […] to […] the Commission's request for information […].

726               SSO Response, paragraphs 228-237.

727               Case C-52/09, Konkurrensverket v TeliaSonera Sverige AB, EU:C:2011:83, paragraph 64; Case C- 549/10 P, Tomra Systems and Others v Commission, EU:C:2012:221, paragraph 79; Case T-336/07 Telefónica SA v Commission, EU:T:2012:172, paragraph 272, upheld on appeal in Case C-295/12 P Telefónica and Telefónica de España v Commission, EU:C:2014:2062, paragraph 124; Case C-23/14 Post Danmark, EU:C:2015:651, paragraph 66.

728               Case C-457/10 P AstraZeneca v Commission, EU:C:2012:770, paragraphs 196 and 202.

729               The Commission concludes that the possible markets comprising both comparison shopping services and merchant platforms would also be national in scope. This is for the same reasons as outlined in section 5.3.1, which equally apply to merchant platforms. See reply of Fnac to Question 1 of the Commission's request for information of 29 October 2015. Google did not dispute this conclusion.

730               SSO Response, paragraphs 178-182, 193-219, 293-304; 181-214.

731               SO Response, paragraphs 254-258 and 330-342.

732               Replies  of  […]  to  […]  the  Commission's  request  for  information  […];  reply  of  […]  to […] the Commission's request for information […].

733               Replies of […] to the Commission's request for information    […]. The Commission has also requested the click data on product listing units displayed on the general search results pages of […]. That data was, however, not included in the Commission's analysis because (i) […] provided incomplete data; and (i)  […] did not agree to disclose its data to Google's external advisors in a data room (see also footnote 581).

734               Replies of […] to the Commission's request for information […]. In the case of product listing units displayed on third party websites by […], the clicks taken into account in the Commission's analysis cover both search-based and non search-based product listing units. This is because click data provided by […] did not differentiate between the two solutions. This is therefore a conservative estimate. See […] reply to […] the Commission's request for information […]. The Commission also requested click data on product listing units displayed on third party websites by […]. That data is, however, not included in the analyses because […] indicated that the product listing units it provides are not search- based. See […] reply to the Commission's request for information […].

735               […]

736               […] Google also listed […] in the SSO Response (Annex 12), the Commission has not included these merchant platforms in its analyses because they are specialised on certain products only, i.e. respectively, home retail products and Do It Yourself products.

737               For the sake of completeness, the Commission also included in the analyses […].

738               Source: Google's reply to […] the Commission's request for information […], as updated in Google's reply to […] the Commission's request for information […].

739               Sources: (i) for clicks on header links, other click elements and links that lead the user directly to the webpage of one of the merchants whose offer is displayed in the Shopping Unit: Google's reply to […] the Commission's request for information […], as updated in Google's reply to […] the Commission's request for information […], and complemented […] with data covering the period November- December 2016; for France, Germany, the Netherlands and the UK (data on header links and other click elements for the period February – April 2013 was based on Google's reply to […] the Commission's request for information […]); (ii) for clicks on menu links: Google's reply to […] the Commission's request for information […], as updated in Google's reply to […] the Commission's request for information […], and complemented […] with data covering the period November-December 2016).

740               Sources: replies of […] to […] the Commission's request for information […]; replies of […] to […] the Commission's request for information […]; replies of […] to the Commission's request for information […]; replies of […] to […] the Commission's request for information […]; reply of […] to […] the Commission's request for information […]; reply of […] to […] the Commission's request for information […]; replies of […] to […] the Commission's request for information […].

741               Sources: reply of […] to […] the Commission's request for information […]; reply of […] to […] the Commission's request for information […]; reply of […] the Commission's request for information […]; reply of […] to […] the Commission's request for information […]; reply of […] to […] the Commission's request for information […].

742               Source: Google's reply to […] the Commission's request for information […], as updated by   Google in its reply to […] the Commission's request for information […] and complemented […] with data covering the period November-December 2016, and […] with data concerning […].

743               In  their  replies  to  the  Commission's  requests  for  information,  comparison  shopping  services  and merchant platforms provided traffic data based on their national presence/(sub)domain(s). For example, […] provided traffic data in those countries in which it has an active national (sub)domain (i.e. United Kingdom, Germany, France, Italy, and Spain); similarly, […] provided traffic data only in Belgium and the Netherlands. Conversely, Google provided data regarding traffic from Google's general search results pages to all the services considered in the First Analysis, based on the IP address of the users  that have visited the websites of the relevant comparison shopping services or merchant platforms, regardless of whether these comparison shopping services and merchant platforms have a local presence/(sub)domain in the relevant country. Therefore, for example, Google provided data regarding traffic from Google's general search results pages to […] also in those countries where […] has no local presence or (sub)domain (i.e. Austria, Belgium, Czech Republic, Denmark, the Netherlands, Norway, Poland, and Sweden). In the First Analysis, with respect to comparison shopping services and merchant platforms, the Commission has used: (i) the data on total traffic provided by the considered comparison shopping services and merchant platforms in the countries in which these comparison shopping services and merchant platforms have a local presence/(sub)domain; and (ii) the data provided by Google on traffic from Google's general search results pages in those countries in which the considered  comparison shopping services and merchant platforms have no local presence/(sub)domain (and for which therefore they did not provide data). Therefore, for example, for […] the Commission used: (i) the data on total traffic provided by […] in United Kingdom, Germany, France, Italy, and Spain; and (ii) the data provided by Google on traffic from Google's general search results pages to […] in Austria, Belgium, Czech Republic, Denmark, the Netherlands, Norway, Poland, and Sweden. This methodology overestimates total traffic to, for example, […] in United Kingdom, Germany, France, Italy, and Spain because traffic reported by […] in these countries also include traffic from users located in one of the countries in which […] has no local presence/(sub)domain. Conversely, this methodology underestimates total traffic to […] in those countries in which […] has no local presence/(sub)domain because it uses as a proxy traffic from Google's general search results pages. However, overall, the use of traffic from Google's general search results pages to complement the data provided by comparison shopping services and merchant platforms allows to obtain a fair representation of users' behaviour in the relevant countries. In any event, in order to address (and check the relevance of) the issue of allocation of traffic across countries, the Commission has also performed the First Analysis in the thirteen EEA countries in which the Conduct takes place in the aggregate. For this aggregate analysis,  to avoid double counting, the Commission used only the data on total traffic provided by the considered comparison shopping services and merchant platforms and, for those comparison shopping services and merchant platforms that did not provide data for the full period 2011-2016, used as proxy for the missing figures traffic from Google's general search results pages, as suggested by Google (SSO Response, Annex 11). The aggregate results of First Analysis in the thirteen EEA countries in which the Conduct takes place are consistent with the results of the First Analysis performed at country level.

744               In order to use total traffic to the Other SO Response Aggregators, the Commission would have needed to send more than 300 requests for information (and receive the relevant replies from the addressees). In the interest of time, the Commission has thus used as a proxy for total traffic to the Other SO Response Aggregators, the data provided by Google regarding traffic from its general search results pages. Overall, the use of total Google traffic underestimates the position of the Other SO Response Aggregators because, although some of them might rely only on traffic from Google or receive the vast of majority of their traffic from Google, others might receive traffic also from other sources. As a result, this calculation underestimates the actual share of comparison shopping services within the possible national markets comprising both comparison shopping services and merchant platforms. This means that the first analysis underestimates the size of the comparison shopping services segments of possible national markets comprising both comparison shopping services and merchant platforms, i.e. the segments affected by the Conduct.

745               The  Commission  also  requested  and  obtained  data  on  total  traffic  to  […]  (reply  of  […]   to  the Commission's request for information […]). That data is, however, not included in the analyses because […] did not agree to disclose its data to Google's external advisors in a data room. Instead, consistent with what Google had done in the SSO Response (Annex 11) and in the LoF Response (Annex 17), the Commission has included traffic to […] from Google's general search results pages in the analyses.

746               The Commission also requested data on total traffic to […] (Commission's request for information […]). That data is, however, not included in the Commission's analyses because these merchant platforms did not respond to the Commission's request. The Commission has therefore, as suggested by Google (SSO Response, Annex 11), used as a proxy for total traffic to these merchant platforms the  data provided by Google regarding traffic from its general search results pages.

747               Submission of […]; reply of […] to […] the Commission's request for information […].   SO Response (paragraph 118) and Annex 6 to the SO Response (paragraphs 49 and 59).

748               SO Response, Paragraph 120 and Annex 6 to the SO Response, paragraph 54.

749               This is the case for […]. Because these merchant platforms have been unable to provide traffic data that does not include single offer webpages, the Commission has used as a proxy, where available, the ratio between the volume (or where the volume was unavailable, the value) of sales concerning the merchant platform's own inventory and the volume of sales of third party products and has applied this ratio to the data regarding total traffic: replies of […] to […] the Commission's request for information […]; replies of […] to […] the Commission's request for information […]. […] did provide sales data concerning 2016; the Commission thus adjusted […] traffic data in 2016 applying the same ratio applied for 2015 (based on the 2015 sales data); reply of […] to […] the Commission's request for information […]. […] provided traffic data that does not include single offer webpages. […] also sells products from its own inventory; however, since the sales of products its own inventory represent less than 1% of its total sales, the Commission has not adjusted […] traffic data. […] also sells products from its own inventory, but it did not provide sales data to be used as a proxy.

750               This is the case for […]. Because these merchant platforms have been unable to provide traffic data that does not include visits to webpages where they display offers from private individuals selling goods or services on a non-professional basis, the Commission has used as a proxy the ratio between the volume (or where the volume was unavailable, the value) of sales of products sold by individuals and the volume of sales of products sold by professional online retailers and has applied this ratio to the data regarding total traffic: replies of […] to […] the Commission's request for information […]; replies of […] to […] the Commission's request for information […]. […] also display offers from private individuals selling goods or services on a non-professional basis, but their traffic was not adjusted because they did not provide sales data to be used as a proxy.

751               LoF Response, Annex 17, pages 4-14.

752               SSO Response, Annex 11, page 14.

753               SSO Response, paragraphs 255-257, and Annex 11; LoF Response, Annex 17, pages 14-21.

754               In practice, the main source of traffic not included in the Commission's Treatment of Missing Figures is direct traffic (i.e. users reaching the destinations sites directly, without going through Google or another general search service).

755               LoF Response, Annex 17, pages 21-24.

756               SSO Response, paragraphs 239-243.

757               This concerns the following services: […].

758               In particular, for competing comparison shopping services and merchant platforms, the Commission has used as a starting point for the First Analysis the variables “visits_unadjusted” created by Google in its Stata dataset named “visits + google clicks.dat”.

759               Reply  of  […]  to  […]  the  Commission's  request  for  information  […];  reply  of  […]  to   […]  the Commission's request for information […];

760               The analyses under points (a), (b), (d), and (e) are based on the Commission's Treatment of Missing Figures, whereas the analysis under point (c) is based on Google's Adjustments for Missing Figures.

761               Certain limited deviations from the results of the analyses as set out in the LoF stem from the correction of the clerical errors mentioned in recital (628) and footnote (757).

762               In the Czech Republic, the figures concerning the analyses under points (a), (b), and (c) refer to the period 2011-2015. This is because, concerning the First Analysis, figures and shares for 2016 have been

replaced with ranges to ensure the confidentiality of the click data on product listing units displayed on […], which are not included in the Commission's analysis set out in the Decision (see footnote 733), but were included in the Commission's analysis as set out in the LoF (LoF Table 37 and LoF Table 64); […].

763               […]

764               […]

765               […]

766               […]

767               […]

768               Case C-7/97, EU:C:1998:569.

769               SO Response, paragraphs 165-182 and Annex 7.

770               SO Response, paragraphs 152-153, 159.

771               SO Response, paragraph 158.

772               Case T-65/98, Van den Bergh Foods, EU:T:2003:281, paragraph 161, upheld on appeal in Case C- 552/03 P, Unilever Bestfoods Ireland v Commission, EU:C:2006:607, paragraphs 113 and 137. See also the Opinion of Advocate General Mazák delivered on 15 May 2012 in Case C-457/10 P, AstraZeneca v Commission, EU:C:2012:293, paragraphs 94-95.

773               Google’s submission […].

774               Google’s submission […].

775               SO Response, paragraph 388.

776               SO Response, footnote 405, Annex 16.

777               SO Response, paragraphs 393-404.

778               SO Response, paragraphs 405-409.

779               Case C-407/08 P Knauf Gips v Commission, EU:C:2010:389, paragraph 91.

780               Case  C-52/09  TeliaSonera, EU:C:2011:83,  paragraphs 20-21; Case T-65/98  Van  den   Bergh Foods, EU:T:2003:281, paragraph 170; Case T-138/07 Schindler Holding Ltd and others v Commission, EU:T:2011:362, paragraph 190.

781               Regarding  the  right  to  impart  information,  see  Case  C-547/14  Philip  Morris  Brands  and Others, EU:C:2016:325, paragraphs 149 and 151. Regarding the right of property and the freedom to conduct a business, see Case C-72/15 Rosneft, EU:C:2017:236, paragraph 148.

782               […]

783               […]

784               Joined Cases 89/85, 104/85, 114/85, 116/85, 117/85 and 125/85 to 129/85, Ahlström Osakeyhtiö and Others v Commission, EU:C:1988:447, paragraphs 11 to 18; Case T-102/96, Gencor v Commission EU:T:1999:65, paragraphs 89 to 101.

785               Case T-286/09, Intel Corp. v Commission, EU:T:2014:472, paragraph 244.

786               Joined Cases 89/85, 104/85, 114/85, 116/85, 117/85 and 125/85 to 129/85, Ahlström Osakeyhtiö and Others (Wood Pulp) v Commission, EU:C:1988:447, paragraph 17; Case T-102/96, Gencor v Commission EU:T:1999:65, paragraph 87.

787               Case T-286/09, Intel Corp. v Commission, EU:T:2014:472, paragraphs 311-314.

788               Case T-286/09, Intel Corp. v Commission, EU:T:2014:472, paragraphs 251-252 and 296.

789               Case T-286/09, Intel Corp. v Commission, EU:T:2014:472, paragraphs 253-255.

790               Joined Cases 89/85, 104/85, 114/85, 116/85, 117/85 and 125/85 to 129/85, Ahlström Osakeyhtiö and Others (Wood Pulp) v Commission, EU:C:1988:447, paragraph 17; Case T-286/09, Intel v Commission, EU:T:2014:547, paragraphs 231-258.

791               Joined   Cases  6/73  and  7/73,   Istituto  Chemioterapico  Italiano  S.p.A.  and    Commercial  Solvents Corporation v Commission, EU:C:1974:18, paragraphs 32-33; Joined Cases T-24/93, T-25/93, T-26/93 and T-28/93, Compagnie Maritime Belge v Commission, EU:T:1996:139, paragraph 203.

792               Case T-141/89, Tréfileurope v Commission, EU:T:1995:62, paragraphs 57 and 122.

793               Case 5/69, Franz Völk v Établissement J. Vervaecke, EU:C:1969:35, paragraph 5/7; Case 322/81, NV Nederlandsche Banden Industrie Michelin v Commission, EU:C:1983:313, paragraph 104; Case C- 41/90, Höfner and Elsner v Macrotron, EU:C:1991:161, paragraph 32; Case T-228/97, Irish Sugar v Commission, EU:T:1999:246, paragraph 170.

794               Guidelines on the effect on trade concept contained in Articles 81 and 82 of the Treaty, OJ C 101, 27.4.2004, p. 81, paragraph 75; Joined Cases 6/73 and 7/73, Istituto Chemioterapico Italiano S.p.A. and Commercial Solvents Corporation v Commission, EU:C:1974:18, paragraph 35.

795               Case 5/69, Franz Völk v Établissement J. Vervaecke, EU:C:1969:35, paragraph 5/7.

796               Case T-65/89, BPB Industries and British Gypsum v Commission, EU:T:1993:31, paragraph 138.

797               Google's submission […].

798               Case C-90/09 P, General Química and Others v Commission, EU:C:2011:21, paragraph 34.

799               Case C-90/09 P, General Química and Others v Commission, EU:C:2011:21, paragraph 35.

800               Case C-90/09 P, General Química and Others v Commission, EU:C:2011:21, paragraph 36.

801               Case C-97/08 P, Akzo Nobel and Others v Commission, EU:C:2009:536, paragraph 57.

802               Case C-90/09 P, General Química and Others v Commission, EU:C:2011:21, paragraphs 37-38.

803               Case C-90/09 P, General Química and Others v Commission, EU:C:2011:21, paragraph 39.

804               Case C-90/09 P, General Química and Others v Commission, EU:C:2011:21, paragraph 40.

805               Alphabet's          press            release            of           7           October            2015,           available at: https://abc.xyz/investor/news/releases/2015/1007 html, downloaded on 13 June 2017.

806               Joined  Cases  6/73  and  7/73,   Istituto  Chemioterapico  Italiano  S.p.A.  and     Commercial  Solvents Corporation v Commission, EU:C:1974:18, paragraph 45; Joined Cases C-241/91 P and C-242/91 P, RTE and ITP v Commission, EU:C:1995:98, paragraph 90.

807               Joined  Cases  6/73  and  7/73,   Istituto  Chemioterapico  Italiano  S.p.A.  and     Commercial  Solvents Corporation v Commission, EU:C:1974:18, paragraph 45.

808               Joined  Cases  6/73  and  7/73,   Istituto  Chemioterapico  Italiano  S.p.A.  and     Commercial  Solvents Corporation v Commission, EU:C:1974:18, paragraph 46.

809               Case T-83/91, Tetra Pak v Commission, EU:T:1994:246 paragraphs 220-21.

810               Joined  Cases  6/73  and  7/73,  Istituto  Chemioterapico  Italiano  S.p.A.  and      Commercial  Solvents Corporation v Commission, EU:C:1974:18, paragraph 45.

811               Joined Cases C-241/91 P and C-242/91 P, RTE and ITP v Commission, EU:C:1995:98, paragraph 93; Case C-119/97 P, Ufex and Others v Commission, EU:C:1999:116, paragraph 94.

812               Case T-69/89, Radio Telefis Eireann v Commission, EU:T:1991:39, paragraph 98, upheld on   appeal in Joined Cases C-241/91 P and C-242/91 P, EU:C:1995:98, paragraph 91; Case T-24/90, Automec v Commission, EU:T:1992:97, paragraph 52; Case T-401/08 R, Säveltäjäin Tekijänoikeustoimisto Teosto  v Commission, EU:T:2008:501, paragraph 46; Case T-410/08 R GEMA v Commission, EU:T:2008:502, paragraphs 46 and 49; Case T-411/08 R, Artisjus v Commission, EU:T:2008:503, paragraph 50; Case T- 422/08 R, SACEM v Commission, EU:T:2008:504, paragraphs 41 and 44; Case T-425/08 R, KODA v Commission, EU:T:2008:551, paragraphs 43 and 46; Case T-433/08 R, SIAE v Commission, EU:T:2008:520, paragraph 37; Case T-167/08, Microsoft v Commission, EU:T:2012:323, paragraph 95.

813               Case T-201/04, Microsoft v Commission, EU:T:2007:289, paragraph 1265

814               Case T-201/04, Microsoft v Commission, EU:T:2007:289, paragraph 1265.

815               Council Regulation (EC) No 2894/94 of 28 November 1994 concerning arrangements for implementing the Agreement on the European Economic Area (OJ L 305, 30.11.1994, p. 6).

816               Case T-83/91, Tetra Pak v Commission, EU:T:1994:246, paragraph 239, upheld on appeal in Case C- 333/94 P, EU:C:1996:246, paragraph 48; Case T-229/94 Deutsche Bahn v Commission, EU:T:1997:155, paragraph 130; Case T-271/03, Deutsche Telekom v Commission, EU:T:2008:101, paragraph 295 upheld on appeal in Case C-280/08 P, EU:C:2010:603, paragraph 124; Case T-336/07, Telefónica SA v Commission, EU:T:2012:172, paragraph 319, upheld on appeal in Case C-295/12 P, EU:C:2014:2062, paragraph 156; Case C-681/11 Schenker & Co. and Others, EU:C:2013:404, paragraph 37; Case T-286/09, Intel Corp. v Commission, EU:T:2014:547, paragraph 1601; Case T- 472/13 Lundbeck v Commission, EU:T:2016:449, paragraph 762.

817               Case 322/81, NV Nederlandsche Banden Industrie Michelin v Commission,  EU:C:1983:313, paragraph 107; Case T-336/07, Telefónica SA v Commission, EU:T:2012:172, paragraph 320; Case T-286/09, Intel Corp. v Commission, EU:T:2014:547, paragraph 1601.

818               OJ C 210, 1.9.2006, p. 2.

819            Paragraph 10 of the Guidelines on fines. 820            Paragraph 13 of the Guidelines on fines.

821            Paragraph 17 of the Guidelines on fines. 822            Paragraph 13 of the Guidelines on fines.

823            Paragraph 21 of the Guidelines on fines. 824            Paragraph 22 of the Guidelines on fines.

825            Paragraph 19 of the Guidelines on fines. 826            Paragraph 25 of the Guidelines on fines.

827            Paragraph 27 of the Guidelines on fines. 828            Paragraph 30 of the Guidelines on fines. 829 SO Response, paragraphs 461-464.

830               Case T-336/07, Telefónica SA v Commission, EU:T:2012:172, paragraph 323.

831               Case T-336/07, Telefónica SA v Commission, EU:T:2012:172, paragraphs 324-325.

832               Case T-219/99, British Airways plc v Commission, EU:T:2003:343, paragraph 315; Case T-461/07 Visa Europe and Visa International Service v Commission, EU:T:2011:188, paragraph 252.

833               Case 322/81, NV Nederlandsche Banden Industrie Michelin v Commission,  EU:C:1983:313, paragraph 107; Case T-321/05, AstraZeneca v Commission, EU:T:2010:266, paragraph 901; Case T-286/09, Intel Corp. v Commission, EU:T:2014:547, paragraph 1602; Case T-472/13 Lundbeck v Commission, EU:T:2016:449, paragraph 782.

834               SO Response, paragraphs 465 and 466.

835               SO Response, paragraph 467.

836               Case C-441/07 P, Commission v Alrosa, EU:C:2010:377, paragraph 40; Case T-491/07 RENV CB v Commission, EU:T:2016:379, paragraph 470.

837               Case T-491/07 RENV CB v Commission, EU:T:2016:379, paragraph 461.

838               Case  C-681/11 Schenker  & Co. and Others, EU:C:2013:404, paragraph 49;  Case  C-499/11     P Dow Chemical and Others v Commission, EU:C:2013:482, paragraph 47.

839               […]

840               This additional amount  is split between Google,  solely liable until 1  October 2015, and    Google and Alphabet, jointly and severally liable as of 2 October 2015, in a pro-rata manner based on the corresponding duration of the Conduct.

841               Case T-472/13 Lundbeck v Commission, EU:T:2016:449, paragraph 782.

842               Case C-408/12 P YKK v Commission, EU:C:2014:2153, paragraph 93.

843               OJ L 362, 31.12.2012, p. 1.