EC, March 11, 2008, No M.4731

THE COMMISSION OF THE EUROPEAN COMMUNITIES,

Having regard to the Treaty establishing the European Community, Having regard to the Agreement on the European Economic Area, and in particular Article 57 thereof, Having regard to Council Regulation (EC) No 139-2004 of 20 January 2004 on the control of concentrations between undertakings (1), and in particular Article 8(1) thereof, Having regard to the Commission's decision of 13 November 2007 to initiate proceedings in this case, Having regard to the opinion of the Advisory Committee on Concentrations, Having regard to the final report of the Hearing Officer in this case, Whereas:

1 INTRODUCTION

1. On 21 September 2007, the Commission received a notification of a proposed concentration pursuant to Article 4 and following a referral pursuant to Article 4(5) of Council Regulation (EC) No 139-2004 (the "Merger Regulation") by which the undertaking Google Inc. ("Google", USA) acquires within the meaning of Article 3(1)(b) of the Council Regulation control of the whole of the undertaking Double Click Inc. ("Double Click", USA) by way of purchase of shares.

2. On 13 November 2007, the Commission concluded that the transaction raised serious doubts as to its compatibility with the common market and with the functioning of the EEA Agreement and decided to initiate proceedings under Article 6(1)(c) of the Merger Regulation. (2)

3. After a second phase investigation, the Commission has concluded that the concentration is not likely to significantly impede effective competition in the common market or in a substantial part of it and that it is therefore to be declared compatible with the common market and the functioning of the EEA Agreement pursuant to Articles 8(1) and 10(2) of the Merger Regulation and Article 57 of the EEA Agreement.

2 THE PARTIES

4. Google, a United States of America (United States) public company listed on the Nasdaq stock exchange operates what has become the most popular internet search engine, whose search capabilities are offered to end users free of charge. It later started to provide online advertising space on its own websites as well as on partner websites (affiliated to the Google "AdSense" network). At the same time, it improved its offer of free functionalities (such as the Google toolbar to facilitate searches) and other web-based software (gmail, Google Earth, Google maps, an office suite, etc) and, more recently especially via the acquisition of YouTube, started to provide content. It derives almost all of its revenue from online advertising.

5. Double Click, a United States non-listed company, mainly sells ad serving, management and reporting technology worldwide to website publishers, advertisers and advertising agencies, in addition to ancillary services. It is launching an intermediation (ad exchange) platform and it owns Performics (a search engine management ("SEM") agency (3)).

3 THE CONCENTRATION

6. On 13 April 2007 Google agreed to acquire Double Click for a purchase price of USD 3.1 billion (approximately EUR 2.3 billion). Pursuant to the transaction, a wholly-owned subsidiary of Google, Whopper Acquisition Corp., will merge with Click Holding Corp., a parent holding company of Double Click Inc.. The parties submitted that Google will acquire all of the shares of Click Holding Corp., which in turns owns 100% of the shares of Double Click. The operation, therefore, constitutes a concentration within the meaning of Article 3(1)(b) of the Merger Regulation.

4 COMMUNITY DIMENSION

7. The transaction does not have a Community dimension as defined in Article 1 of the Merger Regulation. However, on 14 June 2007 the notifying party informed the Commission in a reasoned submission that the concentration was capable of being reviewed under the national competition laws of at least three Member States (Germany, Greece, Portugal, Spain and the United Kingdom) and requested the Commission to examine it. None of the Member States that were competent to examine the concentration indicated its disagreement with the request for referral within the period laid down by the EC Merger Regulation. The case is therefore deemed to have a Community dimension pursuant to Article 4(5) of the EC Merger Regulation.

5 MARKET DESCRIPTION

8. Both Google and Double Click are active in the "online advertising" industry. In this sector the main players are on the one hand, web publishers selling advertising space on their internet pages in order to generate revenues (4), and on the other hand, advertisers, who buy such space in order to place their advertisements ("ads") on the internet and target the audience of internet users.

9. The online advertising market has been growing at an extremely fast pace in recent years: according to estimates, it was expected to grow by 28.2% in 2007, while the rest of the advertising market was to grow at 3.7%. Global internet advertising spending is estimated to have grown from approximately EUR 13 billion (USD 18 billion) in 2005 to approximately EUR 17 billion (USD 24 billion) in 2006, and is expected to reach EUR 26 billion in 2008 (USD 37 billion). Its share of worldwide total advertising spending was estimated at almost 5% in 2005, 5.75% in 2006 and is forecasted to account for nearly 9% of the global advertising spending in 2009 (5).

Types of online advertisements

10. There are different ways to categorize online ads. They can be categorized according to the selection mechanism (the way the ad is selected to appear on a user's screen), their format (text, graphic, rich media) or the distribution channel (direct or intermediation (6)). These categorizations are discussed below.

11. With respect to the selection mechanism, there are two main categories: search ads and non-search ads. Search ads appear next to the results of search queries entered by internet users into internet search engines. They are selected on the basis of the search keywords chosen by the user (7). Non-search ads can appear on any web page and they can either be contextual ads, selected according to the content of the page on which they appear (8), or non-contextual ads. A third category can be identified (on the basis of third party market reports on online advertising, e.g. Jupiter and the Interactive Advertising Bureau (IAB)) which includes classified (9) ads. These classifieds ads are grouped within specific web pages under headings classifying the product or service being offered (e.g. houses for sale in particular suburbs, plumbers active in a specific town) (10).

12. These different selection mechanisms have specific targeting properties. The possibility to discover the individual consumer's interests, even in an anonymous way, makes online advertising particularly attractive to advertisers as they can concentrate their marketing efforts towards users that reveal certain preferences or needs. While for search ads the targeting is based on the self-revealed interests of the consumer (via the search query), for non-search ads the targeting is derived from an indirect and usually less precise definition of the consumer's interests. In other words, when the consumer's interests are clearly identified by search queries, advertisers can profit from a more accurate targeting than when the consumer's profile is revealed by the context of the web page visited (contextual targeting (11)) and/or by simple demographic criteria (e.g. zip code revealed by the IP address of the computer used). There is some consensus among market players that search targeting is more effective than non-search targeting. Non-search advertising has been used predominantly for brand awareness purposes but the increasing, and increasingly sophisticated, use of behavioural targeting (that is based on web surfing behaviour) is closing the gap between search and non-search in terms of effectiveness (12).

13. An alternative way to categorise ads is based on their mere appearance or format. Online ads can take the form of text ads or display (graphical) ads (13). Contrary to text ads (exclusively composed by text), display ads include information beyond text that can be static (a simple graphical "banner") or in "rich media" format, such as video and other dynamic graphics. It should be stressed that ad inventory (the space provided by publishers) is rarely, if not never, restricted to a specific type of ad appearance.

14. There is currently some correspondence between the selection mechanism and ad format in that search ads tend to be almost exclusively text ads, whereas nonsearch ads can be either text ads or display ads. For instance, ads appearing next to the results of a search query usually take the form of a few words and a hyperlink that can be clicked on by the user. Within the non-search category, contextual ads can be either text or display (e.g. an ad advertising a car brand on the "motors" section of an online newspaper can be either a text ad or a graphical ad), while non-contextual ads are in general more likely to be display (e.g. banners on top of the main page of a web journal).

15. There are also two main pricing mechanisms for online ads. First, the "cost per click" ("CPC") is used mainly for text ads (either search ads or contextual ads), i.e. the advertisers only pay the publisher when the user clicks on the text ad. This pricing mechanism reflects the fact that for search and contextual ads, the advertiser generally aims at a direct response from the user (and pays according to that response). Second, the "cost per thousand impressions" mechanism ("CPM") is mainly used for display ads, that is to say that the advertiser pays the publisher when a specific number of ads is displayed to users (each individual ad is one "ad impression").

16. The notifying party estimated the value of the total EEA online advertising market in 2006 at approximately EUR 9.8 billion (14), of which online search advertising accounted for EUR 4.1 billion and online non-search advertising (contextual and display ads) accounts for EUR 4 billion (15). This estimate of EUR 9.8 billion for the size of the total online advertising market significantly exceeds the 2006 market size reported by all available third party studies or reports (16). The market size reported by third party reports varies between approximately EUR 5 billion and EUR 8 billion. The notifying party has maintained that most, if not all, available market reports suffer from a common flaw as they consistently underestimate the size of the online advertising industry, noting, in particular, that they only cover a limited number of EEA Member States and that for some of the reported countries there are other market studies which report a significantly higher market size. Moreover, in a number of countries, Google's search advertising gross revenues exceed the total search advertising market size estimated by at least one market report (17).

17. However, the Commission has not been in a position to fully verify the notifying party's arguments in this respect in the course of its market investigation. Therefore, as a reasonable alternative to the parties' submissions, in the present decision the Commission has taken a more realistic approach and used the market size of EUR 8 billion reported by PWC in a study commissioned by IAB Europe (18). This report covers a large number of European countries and appears to be based on plausible methodological assumptions. If the IAB figure were to correctly reflect the total size of the EEA online advertising market in 2006, search advertising would represent EUR 3.2 billion (19) and non-search advertising (that is to say contextual and display ads, excluding classifieds and e-mail) would represent EUR 2.8 billion. (20)

Intermediation services

18. Ad space can be sold directly by publishers to advertisers through their sales forces or through intermediaries. In very general terms, intermediaries "pool" advertising space made available for sale by publishers and/or advertisers wishing to buy advertising space and facilitate the matching between the supply of ad space and the demand for ad space to place ads.

19. Publishers generally distinguish their available advertising space into "premium" and "remnant" (21). Premium inventory is more valuable to advertisers as it usually consists of the most visible space on a website (e.g. the top space on the most popular pages (such as the homepage) of important online magazines or newspapers). Advertisers are willing to pay a high price for this space and generally purchase it directly from publishers. The less valuable inventory, "remnant inventory", which is more likely to remain unsold, may benefit from intermediated sales as these will increase (through pooling) the probability to find advertisers interested in placing an ad and therefore generate ad revenues for publishers. Both premium and remnant inventory can be sold through either direct sales or intermediated sales. However, premium content is more likely to be sold directly, in particular by large publishers with direct sales teams, whereas remnant inventory tends to be sold indirectly via intermediaries (22). While smaller publishers can also have direct sales team, it is more likely that a small publisher will use intermediation for its entire inventory.

20. Intermediation services are offered by "ad networks" or "ad exchanges" and, to some extent, by "media agencies". An ad network is a two-sided platform serving (i) publishers (websites) that want to host advertisements, and (ii) advertisers that want to run ads on those sites. Ad networks aggregate ad space inventory thus maximizing revenue opportunities and minimizing administrative costs of selling the ad space for the publisher. From an advertiser's point of view, an ad network can be considered as a "single buying point" for online inventory which often also provides handling and performance monitoring of online advertising campaigns. Ad networks generate revenues (paid by advertisers for access to publishers' ad space inventory) that are shared between the network manager (as intermediation fees) and publishers. In the EEA, Google's AdSense, Yahoo!Publisher Network, DrivePM (belonging to the Microsoft group), TradeDoubler, Zanox, AdLink, Interactive Media, AOL, Tomorrow Focus and 24/7 (belonging to the WPP group) and a significant number of other smaller players are active as ad networks.

21. An ad exchange provides a marketplace where advertisers and publishers buy and sell ad space on a real-time basis. The main difference between ad exchanges and ad networks is that ad networks aggregate ad inventory from publishers, which is then re-sold by the platform manager to advertisers, whereas an ad exchange is a (virtual) marketplace where publishers and advertisers can virtually meet to find and execute transactions. Networks tend to be "closed" systems with a finite number of buyers and sellers, whereas exchanges tend to be "open" systems whereby any buyer or seller can access the platform and trade. Open ad exchanges allow both advertisers and ad networks to buy ad space. In the EEA, inter alia Rightmedia (belonging to the Yahoo! group), AdECN (belonging to the Microsoft group), Tomorrow Focus and Quigo are active as ad exchanges.

22. "Media agencies" buy aggregated media space from publishers and resell it to their clients, which are advertisers, usually within the provision of a wider service generally defined as media campaign planning. Most of the agencies already offering this service for offline ads are now also present in online advertising, but some agencies have also developed which are only specialised in online ads. Media agencies can buy online advertising space directly from publishers, but also through ad networks or ad exchanges.

23. Media agencies provide a range of advertising campaign and planning services and act as intermediaries between media owners and advertisers, purchasing media space on behalf of their advertising clients. In this respect their business is at least partially overlapping with the activity of ad networks. In the EEA, WPP, Aegis and Publicis are some of the largest media agencies.

24. Intermediation can take place both for search and for non-search advertising. Ad networks that provide search intermediation either own or outsource a third party search "tool". The intermediation, therefore, specifically takes place in respect of the sale of ad space generated on the search result pages of publishers who use the search tool on their website. In the EEA, Google provides this service through its AFS product, and also syndicates its product to other intermediaries (ad pepper, Interactive Media). Yahoo! and Mediaplex-ValueClick provide search intermediation with their in-house tools.

25. The major providers of non-search ad intermediation include Google, through its AdSense network (notably, with its AFC product), Yahoo! (through its Blue Lithium network, its Yahoo! Publishers Network and its Right Media Exchange), Microsoft (through its DRIVEpm network and its AdECN exchange), AOL/Time Warner (active in the EEA through its networks Advertising.com and Lightningcast), ValueClick, as well as TradeDoubler, Zanox, AdLink, Interactive Media (belonging to Deutsche Telekom) and Tomorrow Focus.

Ad serving tools

26. Once ad space has been sold by a publisher to an advertiser, either directly or through an intermediary, both parties need to ensure that the correct ad actually appears (i.e. is served) onto the publisher website space at the right place at the right time. This step is undertaken by the ad serving tools.

27. The advertiser creates advertisements and uploads them onto an advertiser-side ad server. The publisher enters the campaign terms of the ad (location, price and targeting criteria) into a publisher-side ad server. When a web page is visited by a user, the publisher-side ad server - which records the ad impression generated by the user's visit of the website and determines, in what is called the ad arbitration process, which advertiser to call - enters into communication with the advertiser-side ad server, which then chooses the appropriate ad to deliver on the web page. The relationship between the two servers also enables the advertiser to obtain information relating to the user's online behaviour in the context of the placed ad, via browser cookie technology (23).

28. Several companies offer tools (software) that provide this ad serving functionality for advertisers or for publishers, or for both. These tools can be hosted or non-hosted solutions. Hosted solutions are based on tools granting online access to the main software physically residing on servers owned and managed by the software provider. This is the case of Double Click's tools accessing the overall (DART) infrastructure residing on Double Click's servers (24). Software providers may provide "non-hosted" versions of their tools, as is the case of Double Click's non-hosted version of DFP called DE, which runs on publishers' servers. Intermediaries (e.g. ad networks) may provide routing services to customers adopting non-hosted solutions.

29. These tools enable the publishers to manage their inventory (i.e. choose the ad to place on the ad space) as well as to monitor the financial performance of the ad space sold. Publishers can either build their own in-house technology to serve ads on their sites (e.g. Yahoo!, cNet, Microsoft, AOL, Auféminin/Zanox, WPP, Seat or Disney) or purchase "publisher ad serving tools" from third parties. Ad networks (e.g. ad pepper, advertising.com, TradeDoubler) have also developed their own ad serving tools and use them to serve ads for their clients. Ad serving tools also allow advertisers to find the right ads to be served to the appropriate web pages, as well as to monitor the effectiveness of their advertising campaigns. Advertisers can either build their own in-house technology for this purpose (e.g. eBay) or purchase "advertising ad serving tools" from third parties.

30. Ad serving tools provided on a stand-alone basis by different suppliers and to different customers (publishers and advertisers) present a high degree of interoperability. In other words, there are no major problems of intercommunication between an advertiser using a provider of "advertising ad serving tools" and a publisher using a different provider of "publisher ad serving tools" or its own in-house technology.

Sales channels in the online advertising sector

31. There are currently different ways through which publishers can sell advertising space, using various combinations of intermediation services and ad serving technology. These methods are non-exclusive and publishers tend to use different sales channels at the same time for different parts of their inventory. In particular, intermediation and ad serving tools can be sold independently or as a "bundle" (that is to say an ad network offering both intermediation services in addition to the ad serving tools). The same applies for advertisers (since they can use a variety of distribution channels).

32. These different channels are schematized in the Table 1 below.

<table>

33. In the context of direct sales, advertisers and publishers negotiate the sale of online ad space directly. Direct negotiation is mainly carried out by the largest publishers - such as large newsmagazines, newspapers or news channels, online travel agencies, through their sales forces and generally focuses on their "premium" content (typically, ads on the homepage or banners). Yet, smaller publishers might also have the possibility to organize and manage the direct sale of their inventory. For direct sales, publishers need to purchase ad serving services from providers of ad serving technology (25), such as Double Click.

34. When intermediation comes into play, different combinations of relationships can take place between the players (publishers, intermediaries, ad serving service providers), which can generate "unbundled" or "bundled" solutions (26).

35. As regards the ad serving, depending on the business model developed by the intermediary, different combinations of technology and routing/hosting services are potentially available to publishers. The ad network can make available to its members ad serving capabilities developed in-house (27) or outsourced to a third party tool provider (28). The intermediary may then oblige its customers to use this ad serving technology or allow them to use their own tools. In all cases, publishers and advertisers are free to apply other tools on top of those provided for or imposed by the intermediary e.g. for improved monitoring.

36. Intermediated unbundled sales occur when intermediaries offer no ad serving tool or when they offer it but do not oblige customers to use it (i.e. intermediation services and ad serving are purchased separately by publishers and/or advertisers). For instance, the ad network Oridian uses a third-party ad serving tool but leaves customers free to use their tool. Other examples include ad networks Zanox/Axel Springer, Ligatus, ad pepper (29) and Advertising.com that offer an in-house ad serving solution that is not bundled with their intermediation services.

37. Intermediated bundled sales occur when intermediation services and ad serving technology are jointly offered by the intermediary. For instance, AdLink, Europe's largest ad network, uses Double Click as its ad serving provider for all the publishers belonging to the network and offers ad serving services to its customers. Yet, as mentioned in paragraphs 28 and 30 above, this structure still leaves the possibility for publishers to re-route their ad space through their tool provider, on top of the intermediary's tool, for direct monitoring purposes.

38. The bundled model was initially developed for search ads. Indeed, major search engine providers offering search ads on their own sites have also developed intermediation networks (e.g. Google's AdSense network, Yahoo!'s YPN and BlueLithium). These providers tend to offer intermediation services bundled with ad serving capabilities. The same model is now developing for non-search ads as well. Until recently, suppliers of ad serving tools and intermediaries for non-search ads tended to be separate businesses. However, intermediaries now offer a more complete service by also including in their offer the ad serving technology.

39. There has been a recent tendency towards vertical integration as ad networks as well as media agencies, integrate with ad serving companies. For instance, Double Click, an ad serving technology provider, is launching an ad exchange in order to enter intermediation; Microsoft and AOL (both web publishers and intermediation platform managers) recently acquired ad serving companies (Atlas/aQuantive and ADTECH respectively); WPP, a media agency acquired an ad serving business (OpenAdstream) itself owning an ad network (24/7 Real Media). Atlas-aQuantive, ADTECH and OpenAdstream-24/7 Real Media, nevertheless, are still providing ad serving services on a stand-alone basis.

40. Finally, ad serving can also be "bundled" with the sale of ad space when publishers, which are often "web portals", offer their own inventory directly to advertisers and serve it via in-house ad serving technology provided to advertisers. This model has until now been typically used by internet search engine providers (such as MSN, Yahoo!, Google) to offer ad space on their own web portals (served respectively by AdWords for Google and tailor-made solutions for MSN and Yahoo!).

41. As indicated in paragraph 19 above, large publishers typically tend to use both direct and intermediated channels to sell their ad space, whereas smaller publishers tend to rely mainly on intermediation (30). Both advertisers and publishers tend to use several channels at the same time, in order to get the best deal for their space (publishers) or the best return on the money spent in ad campaigns (advertisers). As an example, the market investigation showed that more than half of the intermediaries who have responded to the Commission questionnaire are aware that their publisher-side customers also sell advertising space directly.

42. The different channels described above are used to a different degree for search and non-search ads. For search ads, the notifying party estimates that around 80% (31) of search advertising is sold directly and 20% is sold through intermediation services. Companies offering search ads directly and via ad networks are Google, Yahoo! and Microsoft as well as a number of smaller competitors. Google offers ad space for search ads both on its own portal google.com (direct sale) and on the result pages resulting from queries in search boxes embedded in its affiliated websites (intermediated sales).

43. For non-search ads (either contextual or display) the supply structure is substantially different. There is no clear leading supplier of ad space but a large number of publishers that sell their ad space to advertisers. Today, roughly 60-75% of all display ad space is sold directly and 25-40% is sold through intermediation (32). However, the market shows a trend towards an increase of intermediation (33), partly resulting from improved targeting technology leading to better monetization of the inventory.

6 RELEVANT MARKETS

6.1. Relevant product markets

6.1.1. Provision of online advertising space

Offline and online advertising

44. The notifying party submits that the relevant product market should encompass the provision of advertising space in all types of media. In this market, publishers are the selling side, whereas advertisers are the buying side. Notably, in the notifying party's view the internet would be just one of the several media channels - among which TV, newspapers, etc. - that can be chosen by advertisers wanting to promote their goods or services. From a demand-side perspective, advertisers would therefore take all media into account when planning their advertising campaigns. Some substitutability among media would be demonstrated through the growth experienced by online advertising in recent years (34), partially at the expenses of other advertising media ("offline advertising"). From a supply-side perspective, this comprehensive definition would be justified by the fact that traditional media publishers as well as major newspapers and magazines offer both online and offline advertising space.

45. This broad market definition cannot be accepted, primarily because the market investigation revealed that offline and online advertising are perceived as separate markets by the majority of respondents. Furthermore, online advertising is used for specific purposes. As opposed to offline advertising, online advertising is considered to be capable of reaching a more targeted audience in a more effective way. Advertisers can precisely target their audience by combining information regarding geographical location, time of day, areas of interest, previous purchasing record of the user and search preferences. This option is not available in the case of offline advertising, for which the amount of "wasted circulation" (35) is undoubtedly higher. In addition to this specific targeting, respondents to the market investigation noted that online advertising has a unique reporting system that enables the advertiser to check exactly how many users have viewed the ad or clicked on it, moreover allowing a rapid "retargeting" of the ad. Hence the measurement of the effectiveness of online ads can also be more precise compared with the traditional measurement systems used in offline advertising.

46. The specific pricing mechanism applied to online advertising also distinguishes the two markets: while offline pricing is in general based on "impressions" viewed by a possible number of consumers (and estimated on the basis of general criteria), online advertising is paid on the basis of the number of internet viewers that effectively establish a contact with the ad. As explained in paragraph 15, online advertising pricing is based mainly on two criteria, both derived from the unique relationship established between the internet user and the web page: the "cost per click" and the "cost per impression" criteria. None of the traditional offline media allows for such a precise connection between reach and the cost of the ad.

47. In its previous Decisions the Commission has also established a distinction between online and offline advertising (36).

Further segmentation within the online advertising market

48. The notifying party also submitted that, if a narrower product market definition was envisaged, that is to say only including the provision of online advertising space, no further distinction should be drawn between the various forms of online advertising. As explained in paragraphs 11 and 13, online ads differ with regard to the way in which they are selected (which leads to a distinction between search and non-search ads) and their visual appearance (text or 'display' ads). Text ads can be search or non-search, whereas display ads are almost exclusively non-search ads.

49. However, the market investigation has revealed to be rather inconclusive on this point. While there are differentiating factors between search and non-search advertising, the extent to which advertisers would switch from one form of advertising to the other in response to relative price changes remains unclear.

50. From an advertiser's perspective, it has been underlined that search and nonsearch ads have different effects and serve different purposes. The targeting characteristic constitutes the essential difference influencing the choices of advertisers. As explained in paragraph 12, while for search ads the targeting is based on the user's precisely revealed interests (via the search query), for nonsearch ads the targeting is connected with a less precise definition of the consumer's interests, determined by means of criteria such as the context of the visited web page and by its geographical location. Some respondents specified that the differentiation between search and non-search ads merely reflects the "triggering mechanism" that determines which ad to place in the inventory space and the different attitude of the viewer (more focused when using a search query), but that this does not necessarily imply the identification of a separate market.

51. Other respondents to the Commission's market investigation also pointed out that all these different types of ads could be considered as substitutable despite the different appearance and targeting properties (reflected in different pricing mechanisms). These respondents submit that the different forms of online ads can be perceived as substitutes as they would compete for the same ad space inventory.

52. The market investigation also showed that, from a technical point of view, the differences between the different types of ads seem to be diminishing. With respect to the targeting dimension, whilst search ads have been traditionally viewed as more effective than non-search advertisements (given that search queries are clear regarding the users' intentions or interests), the ability of nonsearch ads to target relevant consumers is improving. Furthermore, whereas non-search and in particular display ads have largely focused on building brand awareness, a significant part of search advertising expenditure is nowadays also focused on generating brand awareness and not only towards directly generating sales. More generally, the fact that the ad serving tools helping advertisers to assess their return on investment are progressively converging across different types of ads reinforces the conclusion according to which all kinds of ads could be substitutable. The main reason for the limited availability of metrics in some cases seems to originate more from self-imposed policies than technical (37) or regulatory reasons (38).

53. It can, therefore, be inferred that, from an advertiser's point of view search and non-search ads can be considered substitutable to a certain extent.

54. From a publisher's perspective, the possibilities of substitution between search and non-search ads are entirely different. Indeed, publishers can add a search tool on their web page (that is to say a small search box appearing on the homepage of a publisher) and thereby generate additional revenues by sharing the revenues of advertisements appearing next to the search results. Yet these search results generally appear on a new web page not forming part of the publisher's content inventory. Therefore, there is no possible substitution between selling ad space for search and selling ad space for non-search. Some respondents to the market investigation submitted that - from a publisher's perspective - the "needs of customers" change for search and non-search ads. Moreover, the "technical basis" required for the search and non-search ads have been considered to be different. Indeed, it can also be noted that, when a publisher decides to allocate a given space on a web page to a non-search (e.g. display) ad, this would not be substitutable with a "search generated" advertising space, since the latter only appears on the page generated by the search query entered by the user.

55. It can therefore be understood that, from a publisher's point of view, search and non-search ads are considered as rather complementary in the sense that search advertising "completes" or "complements" the sale of non-search advertising space by publishers (39).

56. In view of the above, the existence of a separate market for the provision of online advertising space can be concluded. Furthermore, in this market search and non-search advertising might exert some degree of constraint on each other, especially when considering the advertisers' perspective. From a publisher's standpoint, the distinction between the two categories seems to be clearer. In any case, there is no need to define two separate markets for the provision of search and non-search advertising space as, under any of these market definitions, the transaction would not raise any competition concerns.

6.1.2. Intermediation in online advertising

57. The notifying party submitted that no intermediation market should be identified. In its view, direct sales are effectively constraining the providers of intermediation because a portion of the customers of intermediaries (marginal customers e.g. the large publishers) are able to switch to the direct channel in response to a price increase of intermediation. In the notifying party's view, the existence of these marginal customers would be a sufficient disincentive for intermediaries to raise prices. As a consequence, infra-marginal customers, that is to say the smallest publishers unable to have their own direct sales forces, would also be protected from price increases. Given the effective constraint exercised by the direct sales channel, the market for the sale of online space (or the purchase of online space by advertisers) would include both direct sales and intermediated sales.

58. For the purpose of identifying a separate market for intermediation, the distinction between large publishers and small publishers with no alternative to intermediation would be relevant if a hypothetical monopolist for intermediation services was able to price discriminate between publishers and charge higher prices to publishers not having the direct sales option. In the notifying party's view, this possibility has to be excluded for a number of reasons outlined hereunder.

59. Firstly, publishers do not dedicate a fixed amount of advertising space to direct or intermediated sales but the choice will depend on the value that the publisher is able to extract from the sale of space - if the net profit from selling a given space through intermediation was to increase (for instance if the intermediation fee was to decrease) - a publisher could decide to sell through an ad network rather than directly. Secondly, the costs that an intermediary such as Google would have to bear in order to identify publishers that cannot sell directly and charge them higher prices would be excessively high. Thirdly, small publishers could start selling their ad space directly very easily, on the basis of the guidance and information available on the internet.

60. In fact, Google's AdSense network has two main (identifiable) categories of clients: large publishers with directly negotiated contracts and smaller publishers signing up for Google's standard online contract. Online partners with standard online contracts account for more than [>80%]* (40) of Google's AdSense network partners (41). The revenue generated from online partners accounts for [>80%]* of AdSense for Content (AFC) revenues and [>30%]* of AdSense network revenues in the EMEA region (42). Nevertheless, Google claims that this subdivision would not facilitate price discrimination between publishers as Google is not necessarily aware of the distribution channels used by most of its publishers on standard contract terms.

61. The argument brought forward by the notifying party is not entirely convincing: the fact that publishers are clearly distinguished into two categories according to their economic strength is in itself a key element on the basis of which Google could price discriminate between them, since normally larger publishers with stronger bargaining power sell their premium inventory through direct sales.

62. Furthermore, during the market investigation, the overwhelming majority of Google's competitors (that is to say ad networks and ad exchanges) were actually able to provide precise indications on the share of their publisher customers that are also selling inventory through direct sales.

63. The notifying party also submitted that ad networks are currently developing sales models that are progressively becoming very close to the direct sales channels, by helping publishers to retain more control over what might otherwise be sold by intermediation. On the other hand, publishers, after having fulfilled all their guaranteed sales through direct contracts, would also use their sales forces to place ads throughout their sites in available slots ("run-of-site" sales). However, these developments are still at a very initial stage and cannot be considered to be an established market trend.

64. The notifying party also submitted that currently, "if a publisher is using both the Double Click Advertising Exchange and DFP, the Ad Exchange can dynamically choose the highest revenue ad as between a direct sale by the publisher's website and sale through the Ad Exchange". Double Click could also develop the same functionality between Ad Exchange and another third party's ad serving technology. The fact that the arbitration performed by Double Click's Exchange allows a dynamic choice between the best yielding ads does not imply that other intermediaries apply a similar technology. However, a conclusion regarding the structure of a relevant market cannot be drawn on the basis of a single firm's practice.

65. Several respondents to the Commission's market investigation, both on the customers' and on the competitors' side, supported the distinction: direct sales through a publisher's own sales forces involve high fixed costs that would be difficult to sustain for small publishers (43).

66. On the other hand, the majority of the replies from intermediaries indicate that direct sales and intermediated sales are perceived as competing with each other. It has been observed that, in both channels, the economic purpose is the same, that is, to sell advertising space in order to gain revenues, the presence of an intermediary being irrelevant in this respect.

67. The Commission's investigation also highlighted that, when a web publisher has to decide how to sell his/her inventory, he/she makes a choice based on profit maximizing purposes. It is an uncontested reality in the market that the best yielding inventory ("premium") is more attractive for advertisers and is generally sold directly. Publishers invest in their sales forces in order to gain better revenues, and consequently, try to sell as much ad space as possible through this channel. With regard to remnant inventory, which has a less prominent position on the web pages and therefore has a lesser value, publishers in general choose to sell through the intermediation channel, which involves lower costs. Selling additional inventory, in particular less valuable space, through direct negotiations implies additional costs (as would result from the need for more sales personnel) hence reducing the profitability of the space sold whereas the costs of selling that space through intermediaries is lower. This does not exclude the fact that publishers can "pre-empt" some of their advertising space for direct sales and that, in case the negotiation is not concluded or does not yield sufficient revenues, they can "steer" the sale to an intermediary. Therefore, for larger publishers which use both the direct and the indirect channel, direct sales are not necessarily an alternative for their remnant inventory or, at least, for part of it.

68. A separate market for intermediation in online advertising can be defined in view of the fact that there is no substitute for the service provided by intermediaries for the sale of smaller publishers' inventory and for the sale of (at least) part of the remnant inventory of larger publishers that also use the direct sales channel.

69. However, since the availability of remnant inventory is continuously changed by large publishers (in view of the monetization opportunities), the competitive analysis will take into account the fact that intermediated sales can be constrained by the price of direct sales.

70. Finally, the Commission investigation also revealed that, in the intermediation market, a further sub division may have to be made between search and nonsearch advertising. An intermediary could be offering its service for the sale of ad space resulting from a search query or for contextual or non-contextual nonsearch ads. Some ad networks can provide search intermediation because they own a search "tool" (like Google AdSense, Yahoo! or Vcmedia - Valueclick) while others outsource the search tool from third parties (e.g. from Google or Yahoo!). Intermediaries who do not own or do not outsource a search tool cannot provide such service (e.g. Zanox, Advertising.com).

71. From a demand side perspective, as already mentioned in paragraph 51, advertisers consider that search and non-search ads could be viewed as substitutable despite the different appearance and targeting properties. Hence, in their view, search and non-search intermediation could be substitutable in as much as they both provide ad space for placing their ads.

72. Nevertheless, the Commission investigation provided no elements on the basis of which it could be inferred that the service provided by intermediaries without a search tool is not substitutable with the service provided by intermediaries that have the possibility to provide search advertising.

73. For the purpose of the present decision, it is not necessary to conclude as to the existence of a further subdivision of the market for intermediation in online advertising as, under any of these market definitions, the transaction would not give rise to any competition concerns.

6.1.3. Provision of online display ad serving technology

74. As explained in paragraph 26, a particular feature of online advertising is the need for publishers and advertisers to use ad serving tools. A market, therefore, has developed for the provision of such tools and/or ad serving services to advertisers and publishers.

75. The notifying party confirmed that the market for display ad serving refers to the technology offered to ad agencies and advertisers and to publishers for the delivery of display ads and reporting on the effectiveness of advertising campaigns.

76. Whereas the technical process of serving display ads is not fundamentally different from the technical process of text-based ad serving, the notifying party stated that there are significant differences in the level of functionalities available to customers. It argued that this is confirmed by the fact that display ad serving provides detailed metrics (reach, frequency, conversion) which are not typical of search or context-based text advertising, for which even simple click through rates may go a long way in measuring the effectiveness of ads.

77. The market investigation confirmed the distinction between display ad serving and serving for other kinds of ads.

78. It is worth noting that the supply structure of the online ad serving technology is different for text and display ads. As explained above, serving technology for text ads, in particular, search text ads, is normally provided as an ancillary service to the sale of online ad space, that is to say it is "bundled" with the sale of the space. On the other hand, a number of companies, including Double Click, are active in the provision of stand-alone serving technology for display ads.

79. The notifying party also considered that the provision of display ad serving, management and reporting infrastructure technology could be further distinguished according to whether services are provided to advertisers (and agencies) or to publishers (including self-provisioning). The ad serving for publishers and the one for advertisers are based on the same basic technology which provides trafficking, ad delivery, reporting and optimization. However, these systems are used for different purposes and thus require different functionalities. Advertisers require ad generation, ad hosting and sophisticated reporting technology to measure the effectiveness of the advertising spending, whereas publishers require sophisticated inventory management and sales reservation functions forecasting the likely availability of inventory according to specified criteria and targeting. Thus, there would not be demand-side substitutability between these two types of products.

80. This seems to be confirmed by the structure of supply in online ad serving technology. Indeed, while some providers are offering ad services for publishers and advertisers, such as Double Click, Openads or Atlas, some others are only present on one of the two sides of the market (such as OpenAdStream or CheckM8 for publishers and Bluestreak or Mediaplex for advertisers).

81. The market investigation confirmed that there is a separate market for the provision of ad serving for display ads. The investigation also indicated that such a market could be further subdivided between the provision of such services to advertisers and to publishers.

6.2. Relevant geographical market

6.2.1. Provision of online advertising space

82. The notifying party considers the relevant geographic market to be at least EEA-wide if not worldwide, since the conditions of competition would be largely homogeneous at least across the EEA. In particular, digital advertising would facilitate the provision of ad space across borders. For instance, no technological constraint would impede supplying ads on a given website from abroad, and no technological barrier would exist for customers to purchase advertising space on a cross-border basis. Furthermore, it has been submitted that advertisement campaigns are increasingly conducted on an international scale by global advertisers and agencies. Google also declared that a significant proportion of its customers target an audience outside their home country (44).

83. The results of the market investigation demonstrate that, both for advertisers and for publishers, notwithstanding that technically this market could be EEA-wide, there are many factors which, from a commercial point of view point to a distinction of national or linguistic sub-markets. Supplying or buying of advertising space is differentiated on the basis of national preferences, languages, and cultural specificities. Lastly, in general support and sales networks are locally present in different countries.

84. It can therefore be concluded that the market for online advertising space is to be considered as divided alongside national or linguistic borders within the EEA. The same conclusion would apply to the alternative hypothetical narrower markets for search and non-search.

6.2.2. Intermediation in online advertising

85. The notifying party submitted that the market for intermediation should be considered at least EEA-wide. It argued that, in the light of low technological barriers, the costs of geographic expansion for intermediaries would be low.

86. The Commission's investigation has confirmed that the relevant geographic market for intermediation in online advertising is (at least) EEA-wide. Firstly, from a technical point of view, intermediation services can be provided online on a cross-border basis. Secondly, country or language specificities are of much smaller significance for online ad intermediation than for online advertising, which is considered as commercially distinguished alongside national or linguistic borders. Since intermediaries have an interest in increasing the number of customers belonging to their network or exchange, the intermediation activity frequently aims at reaching and attracting publishers and advertisers in various countries. Such geographic expansion to various Member States successfully occurs because the intermediation service does not depend on the different "content" of the intermediated advertisements.

87. The market investigation has provided several examples of ad intermediators which started their activities in one country and subsequently expanded to a number of Member States. Such is the case of two of the leading online ad intermediation companies, namely TradeDoubler and AdLink, that have expanded their business activities from their respective home countries, Sweden and Germany, across many European countries. The market investigation further indicated that intermediaries in general consider a local presence in each country as advantageous but not necessary. Accordingly, the Commission's market investigation has shown that several EEA based ad networks and ad exchanges operate their servers from one or only a few Member States.

88. For the purpose of the present Decision, it can therefore be concluded that the geographical scope of the market for intermediation in online advertising is at least EEA-wide.

6.2.3. Provision of online display ad serving technology

89. From a geographic point of view, the notifying party submits that the market for ad serving is at least EEA-wide if not worldwide. For instance, Double Click has servers only in the [Four Member States]* but sells in most other Member States and it obtains [>50%]* of its advertiser-side revenues and [>20%]* of its publisher-side revenues from global or pan-EU deals.

90. The market investigation confirmed that the market for the provision of online ad serving technology is at least EEA-wide, essentially for the same reasons brought forward by the notifying party. Depending on the size and business organization, it appears that ad serving is bought on a cross border basis in the EEA.

91. It can therefore be concluded that the market for the provision of online ad serving technology is to be considered at least EEA-wide in scope.

7 COMPETITIVE ASSESSMENT

7.1. Position of the parties in the relevant markets

7.1.1. Google

92. Google is currently active in the online advertising market (i) as a publisher, with its own search engine web page Google.com (and its national web pages such as google.fr, google.it), and (ii) as an intermediary with its ad network (AdSense). Google sells only search-based text ads on its own web pages, while it offers both search-based and contextual text ads on the websites of publishers that participate in the AdSense network (45).

93. Google sells ad space, intermediation services and publisher/advertiser ad serving tools as part of a bundle. For advertisers, Google offers "AdWords", an auction-based advertising programme allowing advertisers to place their ads on Google's search pages or on the websites of the Google AdSense network. For publishers, Google offers "AdSense", which has two product families: "AFC" which focuses on placing contextual ads targeted to the content of the publisher's site and "AFS" which delivers text ads based on the results of queries typed into the Google search box embedded in the publisher's website.

94. Google currently only supplies publisher's ad serving tools for its own Google AdSense network and not on a stand-alone basis. On the advertiser's side, every customer buying ad space from Google (both on Google's website and on inventory provided by AdSense's member publishers) also obtains ad serving from Google via AdWords. No stand-alone, third-party tool can access this inventory controlled directly or indirectly (via AdSense) by Google.

95. Through these direct and intermediated sales channels, Google is the leading provider of online advertising, and in particular of search ad space in the EEA. In the following section, this leading market position will be specified in more detail for each of the possible relevant markets: firstly, on the basis of the market size and market share estimates submitted by the notifying party; secondly, on the basis of the market size as reported by third party market studies and reports and the Commission's market investigation; and thirdly, by briefly indicating those market participants which may be considered as Google's main competitors in the relevant market(s).

7.1.1.1. Market share estimates submitted by the notifying party

96. While there are a number of different third party studies and reports on the overall online advertising market for the worldwide market, various combinations of several EEA Member States and various individual EEA Member States (such as reports from Zenith, Emarketer, PWC, Forrester, Enders, OVK, Gemius (Europe) IAB and Jupiter), the notifying party submits that there is a general tendency for estimates regarding the size of online advertising spending to be understated by these studies and reports. According to the notifying party, this is particularly apparent as far as the relevant reports provide a separate estimate for spending on search advertising and this estimate is lower than Google's revenues from search advertising in some countries (46). On the basis of a combination of the available studies and reports with corrections for those countries where Google's revenues exceed the reported market size, the notifying party estimated the total market size and Google's market shares in the various possible relevant markets.

Total online advertising market

97. The notifying party estimated the size of the worldwide online advertising market (including all different formats and targeting methods of online advertising, that is to say search ads, (non-search) contextual and display ads, classifieds and e-mail) to range between EUR 19.4 billion (47) and EUR 24.1 billion in 2006 (48). Considering Google's 2006 worldwide net revenues of EUR [...]* billion, this would lead to a market share for Google of between [20-30%]* and [30-40%]* in the worldwide online advertising market.

98. In the EEA, the notifying party estimated the size of the total online advertising market in 2006 to range between EUR 5.4 billion (49) and EUR 9.8 billion (50), but strongly suggested that the actual market size would rather be at the upper bound than at the lower bound of this market size estimate. In view of Google's 2006 EEA net revenues of EUR [...]* billion, this would give Google a market share of between [20-30%]* and [30-40%]* in the EEA online advertising market.

99. The notifying party's corresponding estimates of Google's market shares in the overall online advertising market in various EEA Member States in 2006 are set out in Table 2 below (51).

<table>

Search advertising

100. For search advertising, the notifying party has estimated the total worldwide market size in 2006 to range between EUR 8.2 billion (52) and EUR 10.2 (53) billion. Considering Google's total search advertising revenues in 2006 of approximately EUR [...]* billion, this would lead to a market share for Google in search advertising of between [50-60%]* and [60-70%]*.

101. At the EEA level, covering those Member States for which there are third party market studies or reports available (54), the notifying party provided an estimate for the market size in 2006 between EUR 2.6 billion (55) and EUR 4.1 billion (56). In view of Google's 2006 EEA net search revenues of EUR [...]* billion, this would give Google a market share of between [40-50%]* and [70- 80%]* in the EEA search advertising market in 2006 (57).

102. The notifying party's corresponding estimates of Google's market shares in search advertising in the various EEA Member States in 2006 are set out in Table 3 below. (58)

<table>

Non-search advertising

103. In total non-search advertising (excluding classifieds and e-mail, but including both direct sales and intermediated sales of display and contextual ads), Google's market share is limited, primarily because Google currently offers only intermediated sales through its AdSense network and these intermediated sales concern almost exclusively contextual ads and hardly any other types of non-search ads (in particular no display ads). On the basis of the notifying party's estimate of the total EEA online advertising market size of EUR 9.8 billion, the revenue figures provided by the notifying party for non-search advertising (excluding classifieds and e-mail) and the assumption that non-search advertising accounts for 42% of this market (60), Google would have a market share of only [<5%]* (61).

Intermediation in online advertising market

104. With regard to intermediation, considering the notifying party's estimate of a total of EUR 9.8 billion of online advertising spending at the EEA-wide level and assuming that the split in Google's turnover between direct and intermediated sales (62) applies equally to the total intermediation market (excluding classifieds and e-mail), Google's total (gross) revenues deriving from intermediation of EUR [...]* million would represent [40-50%]* of the overall EEA revenue in 2006 for intermediation (63). However, the notifying party notes that this data could overestimate Google's position as it does not include intermediation services provided by media agencies, which might be considered to operate in the same market as ad networks and ad exchanges. Summary of the notifying parties' market share estimates

105. The notifying party's estimate of Google's market shares in the EEA in the various possible relevant markets can be summarized as follows (64):

<table>

7.1.1.2. EEA market shares based on IAB Report and market investigation

106. The starting point for the parties' EEA market share estimates presented in Table 4 above is their assumption that the size of the total online advertising market in the EEA in 2006 was EUR 9.8 billion. This estimate significantly exceeds the 2006 market size reported by all available third party studies or reports. The market size reported by PwC in a study commissioned by IAB Europe (65) is EUR 8 billion.

107. The notifying party has argued that this report also seriously underestimates the total size of the total EEA online advertising market, noting, in particular, that it only covers a limited number of EEA Member States (66) and that for some of the reported Member States there are other market studies which report a significantly higher market size (67).

108. However, in the course of its market investigation, the Commission has not been in a position to fully verify the notifying party's arguments in this respect. Therefore, as a reasonable alternative to the parties' submissions, the Commission has also calculated Google's market shares on the basis of the market size of EUR 8 billion reported by IAB, taking into account possible larger market sizes where this was supported by the Commission's market investigation. Table 5 shows Google's corresponding 2006 market shares in the various possible relevant markets at the EEA level (68).

<table>

109. As is apparent from these market share calculations Google has a leading position not only in the overall EEA search advertising market, but also in the overall intermediation market and/or the two possible sub segments (search and non-search) of the intermediation market.

7.1.1.3. Google's main competitors

110. Google's main competitors in search advertising are Yahoo! and Microsoft with market shares of up to 15% at the worldwide level and at least 5% in the EEA for Yahoo! and approximately 5% for Microsoft both at the worldwide and the EEA-wide levels (74). In addition, there are local search engine providers which concentrate their activities in individual EEA Member States (for instance Seznam.cz in the Czech Republic (75), virgilio.alice.it in Italy, exalead.fr in France Sesam in Sweden and Norway, SAPO in Portugal and miner.hu in Hungary). Yahoo! and Microsoft are both also active in search intermediation. In fact, Yahoo! has significant search intermediation revenues, and Yahoo!'s revenue ratio between search intermediation sales and direct search sales is higher than Google's. Microsoft, on the other hand, focuses more on direct sales of search ads to be published on the results pages of its search engine. In addition, following its acquisition of IMW Group in July 2007, TradeDoubler also offers search intermediation services.

111. In non-search intermediation in the EEA, among others TradeDoubler, Zanox (belonging to Axel Springer), AdLink, Interactive Media (belonging to Deutsche Telekom), Advertising.com and Lightningcast (both AOL/TimeWarner) and Tomorrow Focus are active (approximately 15-20% market share in the case of TradeDoubler, approximately 5-10% in the case of Zanox and around 5% for each of the other (groups of) companies) (76). Smaller players in non-search intermediation in the EEA include not only Yahoo! and Microsoft, but also Oridian, GWB media-marketing GmbH, WPP and VCMedia (belonging to ValueClick). At the worldwide level, apart from the ad networks and ad exchanges active in the EEA, there are numerous other players. Moreover, some of the ad networks and ad exchanges active in the EEA play a much more significant role outside the EEA (e.g. ValueClick, whose worldwide revenues are 20-30 times its EEA revenues; also WPP's worldwide revenues are significantly higher than its EEA revenues). As noted above, to some extent also direct sales of non-search ads also exert a competitive constraint on intermediated sales of non-search ads.

112. As regards intermediation in general, according to comScore, Google AdSense had more than 166 million unique visitors per month (and Google sites more than 171 million) in Europe in 2007. This should be compared with Advertising.com (AOL, more than 106 million), Yahoo! sites (more than 96 million), Microsoft sites (more than 149 million) and TimeWarner network (AOL, more than 78 million). At the worldwide level, broadly the same companies are active: Yahoo!/RightMedia/Blue Lithium; Microsoft/Atlas Network/DRIVEpm/Mediabrokers/AdECN; AOL/Advertising.com/TAcoda/Quigo; ValueClick; and Tribal Fusion.

7.1.2. Double Click

7.1.2.1. Ad serving

113. Double Click offers tools that provide display ad serving functionality, either for advertisers or for publishers. Double Click derives [40-50%]* of its revenues from sales to advertisers and [40-50%]* from sales to publishers. Double Click's tools for publishers are DFP and its "non-hosted" version DE (DART for Enterprises), used by [...]* of its publisher customers. Double Click's tool for advertisers is DFA, which is only provided as hosted application (77).

114. On the advertiser side, Double Click is the leading player together with aQuantive/Atlas (recently acquired by Microsoft). They each have about 35% market share in the EEA and also compete with smaller rivals such as ADTECH, Mediaplex and others. This equal split between Double Click and aQuantive/Atlas also characterises the market at the worldwide level.

115. On the publisher side, the market investigation points to Double Click leading with approximately [40-50%]* market share in the EEA, followed by 24/7 Real Media/OpenAdStream (recently acquired by the advertising agency WPP) with less than 25% and ADTECH/AOL (less than 20%). This leadership of Double Click is also confirmed at the worldwide level.

<table>

116. The notifying party submits that these market shares would be an imperfect indicator as significant variations can be observed from year to year. Hence, Table 7 indicates estimates provided by the notifying party of Double Click's EEA market shares from 2004 to 2006 (weighted according to the countries where Double Click is active) (79):

<table>

117. One explanation for these variations relates to the fact that ad serving tools are often large contracts awarded through competitive bids (80). Double Click usually bids for contracts in the context of a formal Request for Proposal (RFP) or through informal bidding processes. Bidding processes are used not only by new customers, but also by existing customers when renewing their contracts. In 2006, Double Click lost [<50]* existing publisher customers (and [<50]* existing advertiser customers) in bidding contests called by the customer upon expiry of their contract (81). Overall, in 2006, Double Click lost [...]* existing customers ([...]* publishers and [...]* advertisers) operating in the EEA out of a total EEA customer base of [...]* ([...]* advertisers and [...]* publishers)[approximately 13%]*.

118. The dynamic and rapidly evolving characteristics of the ad serving sector has been illustrated by new entry and the rapid growth of this sector, including for Double Click's competitors.

Competition between ad serving tool suppliers

119. The most direct competitive constraint that Double Click faces in the ad serving market comes from rival suppliers of ad serving tools, to which customers could switch in case of a price rise.

120. On both advertiser and publisher sides, Double Click can be considered as a market leader. However, its market shares have to be taken with caution since fluctuations are noticeable from year to year, which could be partly explained by the bidding procedure characterizing these markets. Moreover, in-house ad serving supply should also be taken into account when assessing the market power of Double Click. The market share of Double Click would have been below 20% in 2006 in the EEA if in-house ad serving was included in the market (on the publisher-side, the advertiser-side or both).

121. On the advertiser-side, Double Click and aQuantive/Atlas (recently acquired by Microsoft) are the two market leaders (with a 30-40% market share each).

122. However, many other competitors sell advertiser-side display ad serving technology, including more or less sophisticated management and reporting features: Mediaplex/ValueClick, BlueStreak/Aegis, ADTECH/AOL, Openads, Newtention, Adnologies, Adition, Smart AdServer/Axel Springer and Sapient.

123. Some of these competitors consider themselves as a strong alternative especially for advertisers, with tools offering multi-channel analysis and most importantly real-time actionable functionality in delivery, reporting and billing.

124. The important role played by these competitors on the market is highlighted by the customer base they declare on their websites. For instance, ADTECH manages the ad serving for the BSkyB United Kingdom media group, Himedia, multi.map.com and the editor Class in Italy (82).

125. Some of these competitors are smaller players with worldwide activities (e.g. BlueStreak) or active in a limited number of countries (e.g. Newtention and Adition in Germany, Weborama in France). However, these smaller regional suppliers (amongst others Smart AdServer, subsidiary of Axel Springer group) and rich media ad serving companies (e.g. Eyeblaster) are valuable alternatives competing with Double Click and other global suppliers.

126. This is confirmed by information obtained from the companies' own websites. Smart AdServer provides services to the media agency Publicis and to Zenith Optimedia. The German provider Adition has among its clients Die Bahn, Der Tagesspiele and Netzeitung.de (online journals). Newtention.de provides ad serving to ThomasCook travel agency. Eyeblaster boasts as customers over 2 000 creative and media agencies worldwide, including Euro RSCG Circle, OMD, OgilvyOne, Mindshare, Avenue A/Razorfish, Atmosphere/BBDO, Starcom. It also serves over 2 000 global web publishers, including Yahoo!, MSN, AOL, Weather.com, NYTimes.com, Forbes.com, CBS Sportsline. Moreover, it declares to have run online campaigns for IBM, Disney, Coca-Cola, AT&T, Visa, Warner Bros., Microsoft, General Motors, Unilever, and many more.

127. Once having acquired some customers in the country they are initially active in, smaller players can relatively easily expand their activity to other countries, ensuring the dynamic of the market. For example, one competitor active in France has indicated plans to enter the United Kingdom, Germany and Spain markets by 2008. This willingness for expansion is justified by the sharp increase of the online advertising spending (worldwide as well as EEA-wide and nationally) and eased by the fact that there is neither need to be located in every country to offer services to advertisers (Bluestreak is physically located in the United Kingdom but provides services throughout the EEA) nor barriers to expansion (a company with the technology and services infrastructure in one country can enter a new territory given the speed of international internet lines).

128. On the publisher side, while Double Click leads with around [40-50%]* of the market, Open AdStream/ 24/7 Real Media (recently acquired by the advertising agency WPP) and ADTECH/AOL also have an important presence with around 15-25% and 10-20%, respectively. Following its acquisition of Accipiter, Atlas/aQuantive/Microsoft now also offers publishing services. In addition, a large number of providers (Smart AdServer/Axel Springer, Openads, Newtention, Adition, Exponential, Adnologies, ValueClick, Adnet) are present in the market.

129. Furthermore, some other non-integrated alternatives such as Adify, Zedo, Facilitate Digital, CheckM8, AdJugger are also competing. Finally Mediaplex, Eyeblaster and TradeDoubler also provide publisher-side ad serving solutions in addition to their advertiser-side solutions.

130. The competitive constraint is further demonstrated when considering the customer base of some of these competitors. For instance, Atlas provides publisher side ad serving to AT&T. As for 24/7, in France, Open AdStream manages the ads for AliceADSL.fr, studyrama.com, voyagermoinscher.com, as well as Caradisiac, Editions Neressis, Expedia, France Télévision (France 2, France 3, France 4, France 5), and the media agencies Global Espace et Horyzon MEdia.

131. Some of these smaller competitors have put forward their own advantage compared to Double Click. They have for instance stressed that their publisher customers have access to their campaign reports in real time as well as their earning statistics. They would also be able to use automated sell functions to increase sales of their inventory, or have a unique patent pending technology for integrating ads that is interactive, flexible and real time and can be targeted on the content of online video.

132. Even companies that are not currently active on a given territory are to be taken into account. One of the significant advertiser-side providers considers itself as a credible though much smaller competitor to Double Click in North America on the publisher-side; its product is not actively sold in Europe due to a decision to focus on advertisers. However, as expansion requires limited investment and can be achieved with or without local presence, this company could readily enter and provide a supplementary competitive constraint. This is further illustrated, regarding advertiser-side tools, by publisher-side providers planning to expand their local scope of activity.

133. It is also worth noting that some competitors and large, sophisticated customers (e.g. eBay) have stated that ad serving technology is becoming a commodity and its development or use poses no specific technical problems. This is also admitted by Double Click that has stated in internal documents that "core ad serving is perceived as commodity" and that "competitive differentiation is decreasing" (83). Several companies even use in-house systems to serve ads onto their pages (for publishers) or into their networks and there are even ad serving services provided for free (e.g. Openads tools offered under an open source model).

Alleged "neutrality" of Double Click

134. Some advertisers and publishers have stressed the fact that Double Click, by not being present in the publishing or advertising side of the industry, would be preferred for its neutrality and would have an advantage from not being at the same time an intermediary and a competitor on one side or other of the platform. This would apply for both the advertiser side, where the main competitor (Atlas) is a subsidiary of Microsoft which can be considered as a competitor in selling online ad space for publishers, and the publisher side, where 24/7 Real Media (a subsidiary of WPP) or ADTECH (a subsidiary of AOL) can also be considered as competitors in buying or selling online ad space for advertisers.

135. However, the significance of neutrality appears to be less relevant when considering the profile of some customers of Double Click's current competitors. First, the relevance of this issue is questionable when considering the significant market shares of "integrated competitors" (on both the advertiser and publisher sides; see Microsoft on the advertiser-side and AOL and WPP on the publisherside) which shows that non-neutral players are seen as suitable providers. Indeed, on the publisher side, integrated competitors who have their own website operations do also act as third party ad serving providers to serve publishers which are direct competitors to their ad sales interests. These integrated operators therefore have the incentive to provide a "neutral" service in order to first keep these customers, but also attract new ones. For instance the recently announced online advertising agreement between Microsoft and Viacom (a former Double Click top customer) indicates that the neutrality issue does not seem that relevant: MSN and MSNBC are major competitors of Viacom and its online media.

136. Therefore, it appears that both large and small competitors are able to exert a significant competitive constraint on Double Click in the display ad serving market.

Switching costs

137. While there are numerous ad serving rivals, switching between ad serving suppliers entails some costs. Switching platforms involves three major steps, which are the training of personnel, the deployment (including re-tagging) and the post deployment transition.

138. The perception of the switching process in terms of time and cost varies significantly throughout the market players. Some consider switching to be a straightforward process. Others put forward that switching may represent a non insignificant cost for publishers and advertisers (re-tagging web pages, need to transfer past data from the former system to the new one, training of personnel). It has to be noted that such contradictory statements were indistinctly made by publishers or competitors irrespective of their size (for advertisers, the switch is considered simpler as they usually use several ad serving providers). Nevertheless, such switches may represent a non insignificant cost for some publishers and advertisers. The market investigation has identified some relevant switching costs.

Re-tagging

139. Publishers have to re-tag web pages, which could be a time consuming task depending on the number of sites involved, the number of pages with tags and whether the sites use a content management system. Replacement can be done manually or through an automated process. Manually, customers replace their tags by searching HTML files to text strings representing their tool provider ad tags and replacing them with new tags. Through a content management system, the publisher can replace the tags automatically thus reducing time considerably (84).

Transfer of data

140. Advertisers may need to transfer past data from one system to the other. However, an insignificant proportion of Double Click's customers - under 1% - have actually required migration of ads historical delivery data when they switched from Double Click to a competing ad server (and Double Click does not migrate historical delivery data from new advertiser-side or publisher-side ad serving customers onto its ad serving products). However, publishers that seek to have their historical delivery migrated to their new ad serving platform can do so with minimal time and expense by allowing Double Click to perform a data transfer or using an outsourced ad operations company that would assist in the migration process.

Training

141. Both advertisers and publishers identify training of personnel in new ad serving tools as a relevant switching cost. However all ad serving systems accomplish the same goals using similar concepts based on similar factors. Therefore experienced users can adapt to a new system. Moreover, there is generally only a limited number of publisher's staff involved in campaign management and ad trafficking that needs to be trained (and ad serving providers propose these training sessions, sometimes even for free).

Time constraint

142. Finally, the timing issue could be particularly costly if switching to a competing ad serving supplier implied a disruption in advertising revenues (which are particularly crucial for online publishers relying on this sole source of revenues). However the speed with which switching may occur primarily depends on the amount of time IT personnel require to complete the task.

143. It is possible for customers to transfer information relating to their ad campaigns without disrupting their business operations. This can be done through "object migration". Object migration is typically done by data migration tools provided by the ad serving provider, or by the customer or by an outsourced ad operations company manually re-trafficking the campaigns. Customers with existing ad sales relationships that will continue to be met through their new ad server can migrate current "objects" when changing ad serving platforms.

144. Some respondents have provided evidence about actual switches that took place in a relatively short period of time. The timing involved in performing such tasks can therefore be rather limited. For instance, a major European incumbent telecommunication operator indicated that it took 3 months to switch from Double Click to Falk, before the latter was acquired by Double Click; a significant shopping website switched from Double Click to a combination of a "home grown" solution combined with Yahoo! in 4 months; a leading pay-TV operator stated that switching would take 4 working days, whereas NRJ switched from Double Click to Smart Ad Server in 1 month (NRJ switched because DFP was too complicated); lastly, Paru-vendu said that the re-tagging procedure takes about 15 days. Ad serving operators are also ready to cover part of the switching costs in order to attract customers or to help in the switching process. Moreover, ad serving providers (CheckM8 amongst others) assist publishers to manage the switching procedure (85). Internal emails from Double Click show that a client's switch to OAS only took a few days (86).

145. Furthermore, Double Click's standard DFP migration plan foresees that it takes between [<1 Month]* to be implemented. Of the [...]* European DFP implementations completed in 2007, 70% were completed in less than 30 calendar days, and the remaining in less than 65 calendar days.

Actual switching

146. While the market investigation provided mixed answers regarding the theoretical level of switching costs, there is evidence that a large number of publishers and advertisers have switched from Double Click to other service providers (and vice-versa) in the past few years. This would tend to confirm the view of the notifying party that switching costs are manageable.

147. As indicated in paragraph 117, the switching data provided by the notifying party indicates that in 2006, Double Click lost [<100]* existing customers operating in the EEA ([<60]* publisher customers and [<40]* advertiser/agency customers). Given that Double Click's EEA customer base consisted of [<300]* advertisers and [<300]* publishers in 2006, the switching data implies a churn rate (87) of 12.6% in 2006 (16% for publishers and 8.5% for advertisers) (88). These customers represent about USD [...]* million dollars in lost revenues (based on the customer's revenue in 2005), that is, about [<10%]* of Double Click's 2005 turnover in the EEA. (89)

148. Based on data for the Americas and the EMEA, Double Click lost a total of [<150]* existing customers ([<100]* publisher customers and [<50]* advertiser/agency customers) in 2006 out of approximately [<700]* advertiser/agency customers and [<700]* publisher customers worldwide. This suggests an overall churn rate of about 9% at the worldwide level. The market investigation has revealed that these churn rates may be on the high side when compared with other ad serving providers, although they are similar to the information provided by another significant competitor.

149. Since the beginning of 2007 (and up to October), Double Click has lost [<150]* existing customers ([<100]* publishers and [<50]* advertisers). These customers represent about USD [...]*in lost revenues (based on the customer's revenue in 2006), that is, about [<10%]* of Double Click's 2006 turnover in the EEA (90).

150. These figures demonstrate that several of Double Click's European and worldwide customers have migrated from DFP (and DE or DFA) and switched to rival display ad serving providers and moreover, that switches occur within the industry.

151. Similarly, Double Click has also acquired numerous customers from competitors. In total, Double Click has won [<100]* new customers ([<40]* publishers and [<60]* advertisers) in 2006 and [<120]* in 2007 ([<70]* publishers and [<70]* advertisers).

152. Considering Double Click's competitors, they have recently also announced switches to their services by major customers. For instance Microsoft announced that more than 20 publishers switched from other ad serving providers or solutions to the Atlas Publisher Suite, since the company unveiled its plans to acquire aQuantive Inc. New customers include SmartBrief Inc., Reunion.com Inc. and Entrepreneur.com Inc (91).

153. It is also worth noting that competitors put forward the ease of switching in their promotional arguments. For instance, Accipiter, before being bought by Microsoft, and at a time where Double Click was to acquire Falk, offered a special ad serving migration package to Falk customers stating that "Accipiter's customers find it easy to migrate from their existing ad server to AdManager and to integrate the technology with their existing workflow and billing infrastructure" (92).. Accipiter also argued that its "automated migration tools to make the transition as painless as possible" and advertised the fact that 30% of its new customers in 2005 migrated from competing platforms (93). AOL/ADTECH (94), amongst others, also argues that switching is not a detrimental hurdle for ad serving customers.

154. This is also confirmed by the market investigation through the numerous examples of customers, whether publishers, advertisers or intermediaries, having switched their ad serving provider.

Switching by large customers

155. Some market participants have indicated that these switches usually relate to smaller customers for which switching costs may be less significant or in other words that switching customers account for a non significant part of the turnover (95) of the ad serving provider since larger customers would be reluctant to switch their ad serving tool provider.

156. Larger customer may require more planning and resources to accomplish a switch because they may have integrated their ad servers with their internal systems or developed software to directly call the ad serving tool interface in order to create new ads or access campaign forecasting data. However the switching process involves the same steps (training, deployment and post deployment) regardless of whether a customer has integrated its ad server with other technologies. Furthermore, large customers, who are sophisticated enough, can integrate their ad server with other systems to streamline certain aspects of the ad sales process as they have resources and technological tools to switch ad serving platforms without disrupting their online advertising operations (using for instance workflow automation tools, interoperable with all ad serving platforms, which maintain user interface).

157. Moreover, these larger customers are likely to be more sophisticated from a technological point of view and thus better equipped to implement a switch. Furthermore, these large customers are more likely to exert some degree of bargaining power and have a greater ability to secure competitive pricing and features they desire when contracts are renewed. They are also more likely to consider developing an in-house solution.

158. Migrating a large number of individual sites is typically more time consuming than migrating a single site or a number of sites owned by the same parent company. Thus, relative to an individual publisher, ad networks, which operate numerous independently-owned sites, may take longer to migrate due to the logistical challenges of coordinating a large number of publishers. However, the complexity of any particular switch is more a function of the size, complexity and sophistication of the customer rather than a function of whether it is an ad network or a publisher. This is illustrated by the fact that some of Double Click's former customers were online ad networks ([...]* and its 4000 websites, [...]* and its 592 websites, [...]*, a German-based ad network). Double Click has also been able to attract ad network clients such as [...]* (from an in-house solution) and [...]* (from a competitor).

159. Ad networks could also be considered as large customers. When a publisher decides to switch ad networks or to add a network, it usually contacts other networks and negotiates new representation. If a publisher decides to discontinue the relationship with an existing network, the publisher then terminates the contract with this network, usually at the end of the same month. At that point, the newly appointed ad network handles the transition, following the same three deployment phases used in a direct publisher implementation. The switching process can be standardized and networks have created tagging automation tools to assist in this process. The amount of automation varies depending on the network since larger ad networks have automated booking, while others may use a mix of in-house and outside resources to move campaigns. Although the new network may use Double Click to help with ad serving, Double Click takes no part in the transition process (96).

160. Recent examples illustrate the fact that even large customers can switch at reasonable cost. On the advertiser-side, [...]* (a United States-based advertising company, which represents major advertisers such as [...]* and generates 2-3 billion impressions per month) was able to switch its large volume of business from Double Click to Atlas in less than 30 days. [...]*(AT&T Wireless) also switched from Double Click to a buy-side competitor. [...]* completed its switch from Double Click to Atlas in less than two weeks. [...]*, a German ad agency, switched to Falk (now owned by Double Click) in less than one week.

161. On the publisher-side, switches have been accomplished in a matter of weeks or even days. [...]*, a publisher that began its ad serving experience with 24/7 Real Media's OpenAdServer platform, stayed only six months before switching to Double Click (despite the existence of a three-year contract), in just two weeks (97). In Europe, in particular, [...]*, generating 600-800 million impressions per month, switched from Double Click to ADTECH in less than 30 days. [...]*, with four billion impressions per month, was able to switch from Accipiter to Double Click in just one week.

162. [...]*, successfully migrated a significant volume of impressions from DFP to Yahoo! in a short period of time. On May 25, 2006, Yahoo! and [...]* announced a multi-year deal for Yahoo! to sell and serve all graphical ads on the [...]* site. [...]* informed Double Click that under the terms of the deal, Yahoo! would begin selling ads immediately on [...]* behalf and that all new campaigns would be served by Yahoo! [...]* reduction of its DFP volume from over 1 billion to fewer than 40 million in the span of a week illustrates that even large publishers that generate significant impression volumes can quickly switch ad serving platforms.

163. [...]*, informed Double Click on 19 December 2007 that it would switch to Microsoft upon the expiration of its contract with Double Click on 28 December 2007, suggesting a 10 day period to be enough to complete the migration (in the last three years, [...]* has switched from 24/7 Real Media to Falk, from Falk to DFP, and now from DFP to Atlas). The same applies for [...]*, an advertising sale network for radio and internet (97% of its over 4 million daily DFP impressions were migrated over 3 business days). [...]*, Europe's leading online comparison shopping service operating in 9 countries reaching 10% of all European internet users, informed Double Click of its migration 13 calendar days before the expiration of its contract; within 5 calendar days, [...]* reduced its DFP impression volume by 99% (98).

164. Competitors of Double Click also illustrate that large customers can switch ad serving tool provider. For instance on 19 December 2007, Microsoft and Viacom announced a major online advertising agreement including Viacom's switch of display ad serving provider from Double Click (Viacom was amongst Double Click's top 10 largest publisher-side customer in 2006) to Microsoft's Atlas AdManager (99). The media unit of General Electric Co., CNBC (leading business news provider to more than 300 million homes worldwide which has more than 2.6 million unique monthly visitors), announced on 10 December 2007 that Microsoft would replace Double Click as its ad serving provider and that the migration would be completed in March 2008 for display advertising. IAC, a network of online websites that could be considered as a complex customer (each of the websites comprising the IAC network generates significant impression volume) (100) has also decided to migrate from Double Click to Microsoft Atlas (101).

165. It is also important to note that switching is not prevented by contractual relationships. In this industry, contracts tend to be concluded for short terms. According to the notifying party, in 2006, [>70%]* of Double Click's contracts, representing [...%]* of its revenues for that year, had a duration of 2 years and less (102) and [>50%]* of Double Click DFP contracts only have a duration of one year. This suggests that customers frequently have the opportunity to renegotiate contracts and to switch if they obtain better terms with a competing supplier of ad serving tools.

166. Finally, the ability to switch at manageable cost and the frequent renegotiation of contract terms implied by short contract durations is consistent with the evidence that prices have been consistently decreasing in this market (see details below in paragraphs 168-175). As explained in paragraphs 168-169, the parties have provided evidence that Double Click is offering significant price reductions to prevent customers from switching to competitors.

167. In view of the evidence presented in paragraphs 137-166, it appears reasonable to conclude that, while switching costs are not insignificant, they do not prevent publishers/customers from actually switching between providers.

Evolution of prices in the provision of ad serving tools

168. The fact that the ad serving market is currently competitive is also evidenced by the significant price decline of Double Click's products for advertisers and publishers, during a period of increasing demand: between January 2003 and May 2007, prices fell by around [>70%]* on the advertiser-side and [>60%]* on the publisher-side (103).

169. The data on price decreases relates to the average CPM paid by Double Click customers. A complainant argued that it is not surprising that average CPMs have been decreasing as the market is characterised by volume-tiered usage based discounts and publishers have been growing. Moreover, CPMs are in any case "not an economically meaningful measure of price in this market" (while CPMs can decrease with volume tiers, the value of money extracted from publishers can increase) (104). The Commission considers that the evidence provided by the parties is convincing in showing that Double Click has had to reduce prices in response to competitive constraints by its rivals. As explained below, the parties' data shows that CPMs have decreased within volume-tiers (and hence the fall in average CPM is not primarily driven by increases in volumes) and data on price reductions offered during renegotiations with specific customers at the time of contract renewals suggests that Double Click has responded to competitive pressures.

170. The price decreases observed since 2003 are not primarily driven by volume effects due to customers getting larger and benefiting from better conditions at higher volume-tiers. The development of DFP prices since 2003, divided into four volume-tiers shows that the decline in DFP prices remains pronounced when attention is restricted to a particular volume tier, such as the group of customers with a volume of 0 - 500m page impressions (the same is true whatever the tier considered). The drop in prices is therefore consistent with the intesity of competition.

171. Double Click has been offering substantial discounts and short term contracts when negotiating renewals, rather than "locking" clients on long term contracts and fixed prices. Considering Double Click European customers at the point of renewal during the period from November 2006 to October 2007, divided into four volume tiers representing the range of small to large customers, the largest renewing publisher-side customers, with more than one billion monthly impressions, were on average offered a price reduction of [20-40%]*. The next largest tier of customers was offered a reduction of [30-50%]*, followed by [10- 30%]* and [30-50%]* for the third and fourth tiers respectively. The largest nonrenewing publisher-side customers, with between 300 million and one billion monthly impressions, were on average offered a price reduction of [20-40%]*. The next largest tier of customers was offered a reduction of [5-20%]*, followed by [20-40%]* for the smallest tier. The largest publisher-side customers, who are already offered lower prices, received price reductions that were similar in proportion to smaller customers.

172. For instance, before [...]* announced its intention to move to Microsoft's Atlas AdManager, Double Click actively sought to keep [...]* business. In the most recent renewal negotiations, Double Click offered significant price reductions from the previous contract price of EUR [...]* CPM, including an initial offer of EUR [approximately 24% lower]* effective CPM and a second offer of EUR [approximately 39% lower]* effective CPM after the client informed Double Click that it could save EUR [...]* by moving to Microsoft's Atlas AdManager. Other similar examples of bidding competition (with [5 customers]*) demonstrate that other ad serving companies have increased their competitive pressure on the market.

173. Both the evidence on switching and price reductions is consistent with the fact that the majority of Double Click's contracts in the Community have a duration of one to two years or less. The short duration of ad serving contracts enables customers to obtain lower prices through frequent renegotiations. Indeed, the fact that customers insist on short contract durations is evidence that they do not consider themselves to be susceptible to price increases at the point of renewal due to the presence of switching costs. This point is valid both for small and large publishers (arguably more susceptible to face high switching costs). For example, [4 large customers]* are customers with more than 1 billion impressions per month and 1-year contracts with Double Click while [4 large customers]* have contracts having a duration of two years (105).

174. The price development for selected long-term DFP customers (that remained in the same volume tier throughout the period) shows that the CPMs charged by Double Click have remained relatively stable over the duration of a contract, only to drop significantly whenever a contract came up for renegotiation It is thus clear that Double Click's customers have been able to obtain significant price reductions whenever a contract expired.

175. The market investigation has confirmed that customers have been renegotiating terms and obtaining lower prices from Double Click (106). In addition, Double Click's internal documents (107) clearly describe the presence of price pressure in the market for ad serving.

Competitive constraints from in-house provision of ad serving tools

176. As indicated in paragraph 29, ad serving can also be provided in-house. Major advertisers and publishers have opted for this solution. For instance, the market investigation has confirmed that on the advertiser side, Aegis, eBay or Napster (108) have developed their in-house technology. On the publisher side, cNet, Microsoft, Yahoo!, AOL, Auféminin and WPP, are also using such an inhouse solution. Furthermore, a number of ad networks have also developed their own in-house ad serving technology (Quigo, ValueClick, TradeDoubler, Adpepper, Zanox for instance).

177. The ability to develop in-house solutions imposes a constraint in two ways. First, publishers could develop their in-house solution for their own use in response to a price increase by third-party ad serving tool providers. Second, inhouse solutions can ultimately also be marketed and sold to third parties as well (that is to say customers become competitors). Some publishers have already demonstrated their ability to commercialize their in-house ad serving solution to third parties (Microsoft, Auféminin) and have become actual competitors on the market (Yahoo! is also turning its in-house solution into a solution for third parties). Ad networks that use in-house solutions to deliver display ads to third party sites on which ad space is sold can also offer their ad services as a third party provider (TradeDoubler for instance has started selling its ad serving tool). Therefore, the constraint exerted by the development of in-house solutions for ad serving cannot be ignored when assessing competition in the online advertising industry.

178. If in-house solutions are taken into account, the present position of Double Click in the market diminishes substantially. According to information provided by the notifying party, the market shares of Double Click including inhouse would be:

<table>

Collection and current use of data by Double Click and related direct network effects in ad serving

179. Another aspect that determines Double Click's market position in ad serving is the extent to which Double Click can benefit from network effects in the ad serving market due to the large amounts of customer-provided-data (hereafter, CPI data) it collects on its servers hosting its products DFP and DFA on behalf of its customers (compared to the more limited amounts of CPI data collected by its competitors in the ad serving market) (109).

180. Ad serving naturally produces a number of data records each time an ad is served. The server serving the ad may store these records. As billions of ads are served on the internet every day, the total number of such records is large. Double Click, a market leader on the markets for advertiser side and publisher side ad serving, collects, on behalf of its customers, only a fraction of all such data on its own servers hosting its DFP and DFA products.

181. In general, two types of data have to be distinguished: (1) Data created on the advertiser side by use of DFA; (2) Data created on the publisher side by the use of DFP.

Data created by the use of DFA

182. Each time a user visits a publisher's web page on which an ad from an advertiser using DFA is to be displayed, DFA automatically receives information about the user (by way of Double Click's ad tag embedded in the source code of the web page) (110). It needs this information because it must be able to send back the actual ad to the user's computer. The data collected by DFA therefore contains information about a subset of the web-browsing behaviour of the user across the websites of all the publishers who carry ads of DFA customers, that is to say advertisers. This information as a whole would be highly valuable for advertisers, because it would allow better targeting of ads and consequently an increasing return on investment (ROI) for advertisement spending (indeed, if a user's behaviour can be followed across several publishers' websites, this would provide additional, valuable information on the user). Advertisers would be drawn to an ad serving provider that could leverage this data to serve them better. Publishers may also be interested in this data, because, incorporated in the ad arbitration mechanism of DFP, it could result in better targeting of ads, thus better ad performance on the publishers' sites and consequently a higher attractiveness of these sites for advertisers.

183. Double Click currently does not use the data it has collected in the past to offer better targeting to new advertiser customers. This is because Double Click is contractually prohibited from using data created through the use of DFA except for limited purposes, none of which involves using data about user behaviour for the purpose of improving ad serving to publishers or advertisers other than the one advertiser on behalf of which the data was generated and collected.

184. Thus, Double Click can currently only use the data created through DFA to improve the service to the advertiser whose ads were served when the data was first recorded. For example, it can offer advertisers the ability to specify a sequence of ads that are to be shown to each user across all the web pages onto which the advertiser's ads are served. It is important to note that for this service to work, it makes no difference whether DFA has 190, 10 000 or just 1 advertiser customer. The technical working of the ad serving and the results for each advertiser would be the same in each case. Consequently, in this contractually restricted framework, the data created through the use of DFA does not bring about a network effect that could attract other advertisers to Double Click's ad serving tools. The same is true for publisher customers. For this scenario to work, it is not necessary that the web pages on which the advertiser's ads appear use DFP. They could use any competing publisher-side ad serving technology. As a result, data about user behaviour as it is created through the use of DFA does not bring about a network effect on the publisher side. As discussed in more detail below (cf. paragraphs 258 et seqq.), the merger is unlikely to change this to any significant extent.

Data created by the use of DFP

185. When the user requests the web page of a publisher using DFP, the publisher's web server calls DFP to add the ad tags to make sure ad space is filled. DFP executes its ad arbitration mechanism to decide from which advertiser, or from which ad network, an ad should be taken.

186. As a result of this way of operating, for each ad, DFP has a record about which advertiser or ad network has been selected to fill a specific ad space at a given time for a given web page of the publisher's whole website as well as the IP address of the user who had requested the web page. This record can sometimes be supplemented by information derived from what is known about the advertisers and the ad networks insofar as this information is provided by them, in particular the price they agreed to pay for the selected ad.

187. These records are then aggregated and presented to the publisher in order to track the performance of his/her website. For example, the publisher could track the development of prices paid by different ad networks or, more plausibly, he/she could analyse which of the individual pages of his website created the most ad revenue. This type of analysis could be used as an input to uncover web page design flaws or to create new web pages that are optimised in order to bring in advertisement revenues.

188. This data contractually belongs to the publisher whose website it relates to. Presently, Double Click is limited in the use it can make of this data. In particular, it cannot be made available to other publishers or advertisers or be used to improve ad targeting for other publishers or advertisers.

189. For the individual publisher, as far as data generated by DFP and registered on Double Click's servers is concerned, it makes no difference whether other publishers are also using DFP or another tool for their ad serving. Thus, the data stemming from the use of DFP is not instrumental in bringing about any network effect benefiting Double Click's ad serving tools. Again, as explained in more detail below (cf. paragraphs et 258 seqq.), the merger is unlikely to change this to any significant extent.

190. As is evident from the above descriptions, it also makes no difference to Double Click's advertisers or publishers whether the publishers or advertisers they are dealing with use DFP or DFA. While apparently some minor interoperability issues pertaining to the way ad impressions are counted (and billed) exist, they merely pertain to counting conventions and could relatively easily be avoided by publishers and advertisers agreeing on a common definition of exactly when an ad is supposed to be counted for billing purposes.

7.1.2.2. Intermediation

191. Double Click is a technology company that so far does not sell ad space. Double Click is launching a new ad exchange service named Double Click Ad Exchange that will allow advertisers to bid for space on publisher websites. This Ad Exchange commenced beta testing in June 2007. Such an exchange, however, has not yet achieved full commercialization. The current participants ([<50]* buyers and [<50]* sellers in the United States) conducted a total of [...]* million transactions in November 2007 (when for instance Right Media Exchange achieved 145 billion transactions in July 2007, that is to say Double Click's Ad Exchange achieved transaction volumes equal to less than [1%]* of Right Media's exchange).

7.2. Horizontal effects

7.2.1. Actual competition

192. The description of the parties' activities in the previous paragraphs indicates that they are not direct competitors given that Double Click does not sell advertising space whereas Google is present in the market for the provision of online advertising space. The parties do not directly compete either in the market for the provision of display ad serving technology. While Double Click is a leading supplier of ad serving technology to third-parties, Google only provides ad serving technology as an ancillary service to its offer of online ad space (and not on a stand-alone basis) (111).

193. While the "products" that the parties offer do not compete directly, the combination of these products are substitute solutions from the point of view of publishers and advertisers. As explained above, publishers (and advertisers) can sell (purchase) advertising space through different channels: direct sales, intermediated unbundled sales and intermediated bundled sales (e.g. the AdSense package). There is some substitution between these channels and therefore, within this framework, Google's "bundled" solution competes with the "unbundled" combination of stand-alone ad serving technology (such as Double Click's technology) with intermediation. Given that Double Click is a leading provider of stand-alone ad serving technology and Google is a major provider of the "bundled" solution, the Commission assessed whether, by being a supplier of a component of the unbundled solution, Double Click is currently constraining the competitive behaviour of the bundled solutions provided by Google (112).

194. The main reasons why the Commission does not believe that the parties' constrain each other in a "horizontal" sense are set out below:

(i) the cost of ad serving services represents a very small part of the total cost of unbundled solutions and therefore, the degree to which the parties' product did constrain each other's pricing pre-merger was minimal;

(ii) Double Click faces strong competition from other ad serving companies and these direct competitors are the main constraint on Double Click's pricing; and

(iii) Google's bundled solution and an unbundled solution including Double Click's products are not close competitors.

7.2.1.1. Ad serving services represent a very small part of the total cost of unbundled solutions

195. The fact that different combinations of ad space and intermediation or/and ad serving tools compete in the same market does not necessarily imply that the price of tools for ad serving is constrained by providers of ad space or vice versa. This is because in choosing between different advertising channels, advertisers and publishers consider the total cost of advertising in one channel versus the (every customer buying ads from Google also obtains ad serving from Google and there is no category of customers buying ads from Google but using non-Google ad serving technology to serve these ads). When Google sells ad space, the purchase price includes the ad slot and the serving of the ad. total cost of advertising in another (113) and display ad serving tools account for only a small fraction of the total cost of advertising.

196. Advertisers would respond to a small but significant increase in the total cost of advertising through one channel by re-allocating expenditure to another one. However, since ad serving constitutes a small proportion of the total cost to the advertiser, small but significant changes in the price of stand-alone ad serving can only cause very small changes in the total cost of the unbundled channels relative to the total cost of another channel, such as the bundled one provided by Google. Such price changes are therefore very unlikely to precipitate much (if any) switching from one channel to another.

197. For publishers, ad serving represents a small proportion of the cost of using the unbundled solution (and it represents an even smaller fraction of the total cost of the ad for the advertiser). A large majority of respondents, being customers (advertisers and publishers) or competitors, confirms that the cost of ad serving represents a minor proportion of total advertising costs, typically only 2 to 5% (114) (with the cost of the advertiser tool being higher than the cost of the publisher tool; however the amount an advertiser pays to use an advertiser-side serving tool is not relevant to a publisher's decision in choosing its sales channel). Therefore a 5-10% price increase for stand-alone serving tools would lead to a very minor increase, 0.5% at most, of the total cost of the unbundled solution (115). Such a small increase is unlikely to trigger a switch to the bundled Google AdSense solution. For similar reasons, following the merger, Google would not have the ability and/or the incentives to significantly raise prices for the bundled AdSense solution as Double Click did not previously constrain its pricing.

198. One complainant argued that the analysis of the cost structure should not be done on the basis of the impact of the cost of the ad serving technology as a proportion of the overall cost incurred by an online advertiser, that is to say the value of the final product, but on the basis of the costs of using the intermediation channel, that is to say the value of the intermediation services. On the basis of an example illustrating the significance of ad serving costs in intermediation, the complainant claimed that the input provided by the ad serving technology provider (Double Click) represents a significant input into the intermediation market and therefore, the impact of a change in Double Click's prices would affect the choice of the distribution channel by a publisher (or an advertiser) (116). In particular, the pricing of Double Click's ad serving tools was constrained by the threat of switching to the AdSense bundled solution. The merger will remove this constraint and consequently, according to this complainant, the new entity will coordinate pricing and significantly alter competition between "unbundled" ad networks supported by Double Click's ad serving tools and Google's "bundled" AdSense network.

199. In order to evaluate these arguments, it is important to establish how publishers and advertisers choose between the various alternatives. First, publishers (and advertisers) decide on whether to use a third-party ad serving tool at all. This depends - amongst other thing - on whether part of the inventory will be sold through direct sales, in which case a third-party ad serving tool will be necessary. Once the publisher (or advertiser) uses a third-party ad serving tool, the allocation of the inventory between the various distribution channels will depend on the value at which online space can be sold on each of the channels. The publisher will compare the net profit realized on each channel and distribute its inventory in order to optimise its monetization. Based on the example provided by the complainant (117), if the merged entity was to increase the price of DFP for serving ads on competing ad networks by say 10%, this would trigger a decrease in the publisher's net profit of about 0.3% when serving ads on those networks. (118) The publisher would switch inventory to AdSense if the net profit of serving an ad through AdSense is at least 0.3% higher than the profit initially made on the competing network. In other words, switching would occur if the publisher was almost indifferent between AdSense and the competing ad network in the first place (119). Hence, in order to induce any significant switch between ad networks, a substantial relative price increase would be necessary (e.g. a very significant price increase for DFP used on other networks).

200. Similar considerations apply on the advertiser side. Based on the example provided by the complainant (120), if the merged entity was to increase the price of DFA for serving ads on competing ad networks by 10%, this would trigger an increase in the advertiser's total costs of about 0.4% (121). As with the publisher side, the price increase in ad serving necessary to cause advertisers to switch networks would need to be quite substantial.

201. Accordingly, the market investigation confirmed that an increase of prices of stand-alone ad serving tools that would make customers, that is to say publishers, advertisers and intermediators, switch to other forms of advertising would have to be quite large, "fairly dramatic" and "unlikely" and would ultimately trigger switching to other ad serving solutions.

202. Finally, even if the complainant's argument were to be fully accepted, the ad serving costs for unbundled platforms still account for a relatively low proportion of the total costs of intermediation incurred by unbundled platforms (122). In other words, the market investigation established that the degree of substitution between Double Click's ad serving tools and Google's intermediation services is limited. It is therefore unlikely that the parties would have unilateral incentives to increase prices after the merger.

7.2.1.2. Double Click faces competition from other ad serving companies that could be part of the unbundled solutions

203. Based on the analysis above, the Commission concluded that an eventual post-merger increase of the price of Double Click's ad serving tools is unlikely to lead to a switch of publishers and advertisers from unbundled solutions towards the Google bundled solution. More realistically, such a price increase would lead to switching towards competing suppliers of stand-alone ad serving tools present in the market. Given the competitive constraint that Double Click faces on this market, such price increase would be unlikely. Indeed, the market investigation has confirmed that a number of viable providers of stand-alone ad serving tools are today present in the market and exert a competitive constraint on Double Click.

Actual competitors

204. As has been stated in paragraphs 121-122, on the advertiser-side these competitors include aQuantive/Atlas (recently acquired by Microsoft), which shares the market equally with Double Click (around 40% each). Other competitors are Mediaplex/ValueClick, BlueStreak/Aegis, ADTECH/AOL, Openads, Newtention, Adnologies, Adition, Smart AdServer/Axel Springer and Sapient.

205. On the publisher side, as stated in paragraphs 128-129, Double Click faces competition from Open AdStream/ 24/7 Real Media (recently acquired by the advertising agency WPP) and ADTECH/AOL, which also have an important presence with around 20% market share each. Atlas/aQuantive/Microsoft also offers publishing services following its acquisition of Accipiter. In addition, a large number of other providers (Smart AdServer/Axel Springer, Openads, Newtention, Adition, Exponential, Adnologies, ValueClick, Adnet) and some non-integrated alternatives such as Adify, Zedo, Facilitate Digital, CheckM8, AdJuggler are present in the market. Finally, Mediaplex, Eyeblaster and TradeDoubler also provide publisher-side ad serving solutions in addition to their advertiser-side solutions.

206. These advertiser side and publisher side ad serving providers are able to exert a significant competitive constraint on Double Click in the ad serving market.

Switching

207. These competitors are present on the market and do exert a competitive constraint on Double Click. As explained in detail in paragraphs 146-167, there is evidence that a large number of publishers and advertisers are switching to and from Double Click, the corresponding churn rates are significant, Double Click's competitors have also recently announced switches of major customers from competing ad serving providers to their services and have put forward the ease of switching in their promotional arguments.

208. Data on lost and won EEA customers provided by Double Click confirms these dynamics. Double Click lost [10-30%]* existing advertiser-side customers in 2006 and 2007. Of these, Double Click was able to identify the new supplier in [>10%]* cases. In half the cases, customers were lost to ADTECH (AOL), while the rest were lost to Adition (in Germany), Atlas and Smart Ad Server. Double Click also competed in a number of bids for prospective customers against 24/7 Real Media, Atlas, ADTECH, CheckM8, Eyeblaster, Mediaplex, Smart Ad Server and TradeDoubler. During these two years, Double Click won [<50]* new advertiser-side customers which were previously using ad serving tools from 24/7 RealMedia, Atlas, Eyeblaster, Mediaplex, Smart Ad Server and TradeDoubler.

209. On the publisher-side, Double Click lost [<100]* existing customers in 2006 and 2007. Of these, Double Click was able to identify the new supplier in [>30]* cases. A larger number of customers were lost to Adition in Germany and ADTECH (AOL), while the rest were lost to Smart Ad Server, Accipiter and 24/7 RealMedia. Double Click competed for prospective customers in a number of bids against 24/7 RealMedia, Accipiter, Adition, ADTECH, CheckM8, Mediaplex, Smart Ad Server and Zedo. During these two years, Double Click won [<50]* new publisher-side customers which were previously using ad serving tools from 24/7 RealMedia, Accipiter, ADTECH, CheckM8 and Mediaplex

210. In view of the evidence presented, it appears reasonable to conclude that, although switching costs are not insignificant, they do not represent an important obstacle to publishers/customers actually switching ad serving providers. The existence of actual switches does provide evidence of this conclusion.

Evolution of prices in the provision of ad serving tools

211. The fact that the ad serving market is currently competitive is also evidenced by a significant price decline of Double Click's products for advertisers and publishers, as described in detail in paragraphs 168-175. Double Click has been offering substantial discounts and short term contracts when negotiating renewals, rather than "locking" clients on long term contracts and fixed prices. The price development for selected long-term DFP customers (that remained in the same volume tier throughout the period) shows that the CPMs charged by Double Click have remained relatively stable over the duration of a contract, only to drop significantly whenever a contract came up for renegotiation. As noted in paragraph 175, the market investigation has also confirmed that customers have been renegotiating terms and lowering prices of the supply contracts with Double Click

7.2.1.3. Google's bundled solution and an unbundled solution including Double Click are not close competitors

212. Firstly, Google is a direct provider of space for search ads and an intermediary with its ad network (AdSense) for search and contextual ads. These features make it a distant competitor of unbundled solutions where Double Click would be providing ad serving tools. Double Click is mainly a provider of serving tools for display ads. Indeed Google's presence in intermediation for display ads is minimal (far less than 1% of the display advertising segment). In other words, the parties' products are strong for different types of ad formats.

213. Google allows display ads to be placed through its AdSense network (but not on Google.com). However, advertisers generally choose other networks than AdSense to place display ads because the price that they would have to bid on AdSense for a display ad, as compared to text ads, would not be justified in view of the fact that Google does not provide the metrics required by advertisers purchasing display advertising. Indeed, when Google text ads appear on a web page, Google records whether the user has clicked on an ad, which is necessary because Google charges on a CPC basis. Google also records the destination to which the ad took the user, that is to say typically the advertiser's website, but there is no additional tracking function beyond that initial link that reports or tracks the user's subsequent web behaviour, including, whether the user ultimately buys a product, requests information or takes some other desired action. Google cannot provide those metrics in part because, unlike Double Click, it does not deliver cookies to users upon the serving of the impression that allow monitoring which users have viewed which ads.

214. This is something that most display advertisers require, and for which more sophisticated ad serving, management and reporting technologies such as Double Click's are used. Accordingly, Google has had very little demand for display ads on its system, its provision of display ad space accounts for a very small percentage of all display ads on the internet. The insignificant turnover achieved by Google with this activity illustrates that direct competition between Google and unbundled solutions including Double Click ad serving tools is today virtually non-existent.

215. The fact that Google had decided pre-merger to develop a new ad serving product for advertiser-side display ad serving, management, and reporting, Google Ad Manager for Advertisers, "GFA", which is still in the early stages of development and not currently available, confirms that Google's current offering is not interchangeable with Double Click's products. Google also only recently decided to begin developing a publisher-side ad serving solution, Google For Publisher, "GFP". GFP is still in development and not currently available (123).

216. Secondly, Google's solution operates as a closed network, imposing a number of restraints to participating publishers and advertisers. Indeed, in order to use AdSense, publishers must join Google's Publisher Network. Ad network providers, such as Google, share the advertising revenue generated by its network with the publishers providing the inventory. The revenues transferred by the intermediary to the publisher are considered as a traffic acquisition cost (TAC) and depend on the relative bargaining power of the intermediary and the publisher. The negotiation may be subject to some degree of asymmetry of information in that the publisher may not know exactly the total amount of ad revenues generated by its inventory via the platform and the agreement may not refer to a precise share of total ad revenues (124).

217. On the contrary, Double Click ad serving tools could be used in unbundled solutions involving direct sales of ad space or intermediation via open ad exchanges that can be used on an as-needed basis, with no guarantees or minimums (an "open" network). With such solutions, advertisers have more control over which inventory they wish to purchase, while publishers have more control over which advertisers will be allowed to buy their inventory. Such solutions seem to involve a minor transfer of revenues to intermediaries, or no transfer at all in case of direct sales.

218. For the reasons elucidated in paragraphs 212-217, bundled solutions seem to be preferred by smaller players, or larger players interested in the acquisition or sale of remnant inventory, while unbundled solutions are normally preferred by the large players and, in particular, for the acquisition/sale of premium content.

219. A complainant has provided a survey of the top EU and United States websites (on the basis of page view) showing that a large proportion of these top websites uses both AdSense and Double Click. In addition, the survey indicated that AdSense and Double Click ads were sometimes served at similar positions on the web page across all websites surveyed that is to say there was no systematic tendency for Double Click ads to be in one position and AdSense ads to be in another position on the surveyed web pages. As a consequence, the complainant concluded that ads served through AdSense and Double Click are substitute channels for distributing ad inventory to advertisers. In fact, the notifying party provided the list of Double Click's EEA publishers that also served ads through AdSense in 2006 and identified only [...]* publishers out of Double Click's publisher customer base of [...]* (i.e. [15-20%]*) which used both Double Click and AdSense. The revenues generated by most of them are very low, representing a tiny fraction of AdSense's gross revenues. Therefore, not only is the pool of common customers limited, but also, the fact that there are overlapping customers does not necessarily mean that the ad space used for AdSense and the ad space used to serve ads with Double Click tools either directly or through other intermediaries are the closest alternatives to one another for any given publisher (125).

220. Google's closest competitors are likely to be other major search engines that are also providers of ad space and have developed similar networks such as Yahoo! or Microsoft. These operators will be the closest choice for advertisers with a preference for search ads (and publishers able to display such ads) and a preferred choice for bundled solutions. Consequently, other advertising closed networks are likely to be the next best alternative. Some of these networks, such as Advertising.com, have reached already sizeable proportions in the market. The unbundled solutions using third-party ad serving technology such as Double Click are likely to constitute more distant alternatives.

Conclusion on actual horizontal competition

221. In view of this analysis, it does not seem likely that Google and Double Click are today exerting a significant competitive constraint on each other's activities. Therefore, the operation does not seem to significantly impede effective competition with regards to the elimination of actual competition.

7.2.2. Potential Competition

7.2.2.1. Double Click as potential competitor in online ad intermediation

222. Several complainants assert that the merger would eliminate potential competition between Google and Double Click. While the various theories of harm brought forward by these complainants differ with regard to the details and nuances, they all rely on the presumption that Double Click has a number of advantages that would, absent the merger, allow it to develop into a key competitor of Google in the market for online ad intermediation and, by extension, in the market for the provision of bundled online ad intermediation and ad serving services.

223. As explained in paragraph 191, Double Click already launched an ad exchange which it is testing with some of its customers. Therefore Double Click had planned to develop into an ad intermediation operator. Such plans have been put forward and described in detail in Double Click's internal documents about the ad exchange.

224. The relevant question however is whether the elimination of Double Click as a potential competitor in the market for online ad intermediation and in the provision of bundled online ad intermediation and ad serving services may significantly impede effective competition. To this purpose, it must be noted that the guidelines on the assessment of horizontal mergers (126), in paragraph 60, clearly indicate that two conditions must be fulfilled: first, there must be a significant likelihood that the potential competitor would have grown into an effective competitive force and second, that there would be an insufficient number of other competitors left which could maintain sufficient competitive pressure after the merger (127).

225. In order to perform this assessment, Double Click's possible relevant advantages will be analysed and compared with the situation of other companies operating in the market for ad intermediation. Double Click's alleged advantages can broadly be divided into three types:

(i) First, advantages deriving from the possible integration of Double Click's ad serving technology with ad intermediation services offered by Double Click;

(ii) Second, advantages deriving from the current customer base of Double Click among both publishers and advertisers; and

(iii) Third, the existence of direct network effects which could give Double Click a strong competitive advantage through the use of information about consumer behaviour collected through ad serving.

7.2.2.1.1. Advantages deriving from the possible integration of Double Click's ad serving technology with ad intermediation services offered by Double Click

226. As explained above, Double Click enjoys a strong position with regard to ad serving technology for advertisers and for publishers, even if it is constrained by other players, both large and small competitors. Several respondents have mentioned the superior quality of its technology and some even indicated that it has become the standard for parts of the industry, especially for sophisticated users such as for publishers with a strong share of direct sales and ad agencies.

227. Some advertisers and publishers have also stressed the fact that Double Click, by not being present in the publishing or advertising side of the industry, is preferred for its neutrality and would have an advantage from not being at the same time an intermediary and a competitor on one or other side of the platform.

Current position in intermediation

228. With regard to intermediation services, the Double Click Advertising Exchange has not yet developed any significant market position.

229. In June 2007, [<50]* publishers and [<50]* advertisers had agreed to participate in the test in the United States. Of these original [<100]* participants in the United States, only about [27%]* have been participating at any given time. In the Community, Double Click has signed on [<50]* customers to participate to the Ad Exchange, only [<10]* of which are publishers. By October 2007, the exchange had not yet earned any revenue in Europe (the model of an ad exchange as opposed to an ad network is not developed in Europe and must compete with established ad networks), and in the United States it has generated only about USD [...]* in revenue (far less than 1% of the net revenues of one of the main EEA intermediation player).

230. It could not be excluded that the announcement of the merger with Google may have had an impact on the development of the exchange, by creating uncertainty for potential users and possibly slowing down the efforts by Double Click to develop its ad exchange platform. However, even assuming such an impact, the numbers provided in the previous paragraph suggest that Double Click Ad exchange is unlikely to have become in the short term a significant intermediation operator.

Combination of Ad Exchange with Double Click's ad serving technology

231. In the medium term, it could not be excluded that the combination of Double Click Ad exchange with the leading ad serving technology currently provided by Double Click would have allowed this company to obtain a non minor presence into intermediation services. But Double Click's internal revenue projections anticipated revenue in 2008 of only USD [<10]* million in the EU and USD [<20]* million worldwide which represents a tiny fraction of the indirect sales channel128. So even if the Ad Exchange achieved its revenue's projections, this would be equivalent to less than [<5%]* of the size of just one of the networks belonging to one of the notifying party's main competitors.

232. In addition, such a combination of an intermediation platform with ad serving services does not seem to be unique to Double Click. In fact, there has been a trend in the market towards vertical integration between companies present in intermediation services and companies providing ad serving tools.

233. Yahoo!, which sells ad space through the Yahoo! Publisher Network and the Right Media Exchange, and recently acquired major behavioural targeting ad network BlueLithium, also developed a third-party display ad serving solution, Yahoo! For Publishers (129).

234. Microsoft sells ad space through Microsoft AdCenter and DRIVEpm (both ad networks) and the ad exchange AdECN (130). Microsoft has an internally developed technology solution that provides publisher-side display ad serving for ads placed on Microsoft properties and also owns the Atlas third-party ad serving provider. Microsoft recently also launched "AdMarket," a new technology designed to allow publishers to sell premium inventory without using a direct sales force or an ad network (AdMarket allows publishers to set up their own proprietary ad exchanges to facilitate the sale of inventory). In addition, Atlas offers its Network Control Panel, which is integrated with its publisher-side ad serving product and is intended for mid- to large-size publishers that redirect remnant inventory to ad networks. It allows publishers to manage relationships with multiple ad networks and optimize inventory by choosing the network that will provide the highest revenue.

235. AOL owns the leading display ad network (Advertising.com), a major behavioural targeting ad network (Tacoda), a display ad serving provider (ADTECH) and a mobile phone advertising network (Third Screen Media). In addition, AOL recently announced the creation of Platform A, which integrates its leading display ad network, Advertising.com, with its newly acquired technologies and businesses, including ADTECH's display ad serving business.

236. WPP, through its acquisition of 24/7 Real Media, operates a major ad network (Global Web Alliance), and a publisher-side ad serving platform (OpenAdStream). WPP plans to offer an integrated online advertising service (ad buying, ad selling, ad network, ad serving).

237. Axel Springer, a major German online publisher and seller of online and offline ad space owns Zanox.de, a leading affiliate network, and a major French ad serving provider, Smart AdServer.

238. Aegis, another major ad agency, acquired BlueStreak, an advertiser-side ad serving provider.

239. In addition, and contrary to Double Click, both Microsoft and Yahoo! also have sophisticated search ad businesses that provide a very good platform for selling search ads, which they could integrate into their bundled offering, thereby increasing its attractiveness in relation to advertisers which in turn would attract more publishers to their networks. Double Click would not be able to replicate such a larger bundle in the absence of the merger with Google.

240. To the extent that neutrality is perceived as an advantage of an intermediation service provider, all vertically integrated providers have a strong incentive to ensure that their intermediation services are neutral from their publishing or advertising activities. This is evidenced by the number of publishers who have succeeded despite the lack of such perceptions of neutrality. Google has its own publisher site and also provides content through YouTube, which has not impeded the growth of Google's AdSense network. The same is true for Yahoo!, Microsoft and AOL, which are major publishers also operating substantial ad networks or ad exchanges. Their affiliation with a publisher or advertiser (in the case of WPP) has not prevented them from operating some of the largest intermediation platforms.

241. In view of these considerations and developments, it seems reasonable to conclude that the combination of Double Click's leading and neutral ad serving technology with an ad exchange would probably have allowed this company to become a player in the online ad intermediation market. However, it cannot be established that such combination would have placed Double Click in a unique position with regards to other intermediation players that would also benefit from a similar combination of ad serving and intermediation activities and are already established in the intermediation market with a stronger position than Double Click.

7.2.2.1.2. Advantages deriving from the current customer base of Double Click among both publishers and advertisers

242. To assess whether Double Click's customer base would be likely to contribute to Double Click becoming a strong competitor to Google without the merger, it is necessary to assess the importance of this customer base.

243. Firstly, the simple numerical size of the present customer base of Double Click as ad server ([<300]* advertisers and [<300]* publishers at the EEA level in 2006) does not suggest that it enjoys a significant advantage compared to its future competitors in intermediation. For instance, when considering ad networks or ad exchanges active in the EU (such as AdLink, TradeDoubler, Zanox, Advertising.com, ValueClick or Yahoo!), they have several thousands of customers on both the publisher and advertiser sides. Consequently, even if all of Double Click's customers decided to use its future ad intermediation services, its ad exchange would still have a much smaller number of participants than competing ad intermediation services.

244. With regard to the size and importance of Double Click's customers, [<20]* of Double Click's Top 100 EU publisher customers are ranked within the Top 50 websites across EU Member States by Nielsen/NetRatings. At the EU level, Double Click serves [10-20%]* of the top 50 publishers total ad impressions ([30-40%]* if DE, the non-hosted version of DFP, is included) (131). On a worldwide basis, and based upon comScore data for September 2007, [<20]* of the top 50 worldwide web properties are within Double Click top 100 publisher customers (132).

245. However, the figures provide an incomplete indication with regards to the importance of Double Click's customers. Publishers with quality traffic and having premium content tend to use the direct sales channel more than other publishers such as smaller publishers and blogs whose needs are better suited by intermediation platforms. The DFP publisher base has less unsold inventory than do smaller sites and ad networks, to which ad exchanges are more valuable. The principal customers of ad exchanges are ad networks and small publishers with large amount of inventory, not large publishers with a high ranking in Nielsen/NetRatings. Indeed, the top sites usually relying on direct sales tend to have limited amounts of the lower-value remnant or unsold inventory that is typically sold through an ad exchange. Neither of these two customer groups constitutes a significant proportion of Double Click's customer base. Such customers are unlikely to serve significant impression volumes through an ad exchange, even though they are likely to use an ad network or ad exchange for at least some of their ad sales.

246. Even taking these considerations into account, the actual customer base for ad serving technology would only be relevant to evaluate the future position of Double Click in intermediation services if it could be demonstrated that ad serving clients could easily be converted into intermediation clients.

247. This scenario is not convincing, however, as it disregards how the market currently works. Both publishers and advertisers, especially middle and large companies, that is to say those that are similar to Double Click's customer base, typically are using more than one intermediation platform (ad exchange or ad network) at the same time (multi-homing). Data provided by the notifying party shows that more than half of Double Click customers used at least 2 ad networks (133) (and sometimes more than 5) (134).

248. As explained above, currently, around 75% of display ad space is sold directly from publishers to advertisers. The remainder of the space is sold through ad exchanges (which facilitate the sale of ad space) or ad networks (which remunerate the publisher for being allowed to fill its ad space with ads collected from advertisers participating in the network).

249. As ad exchanges are online markets, it is not guaranteed that a publisher will be able to find buyers for its entire remaining inventory on one single ad exchange. It therefore makes sense for publishers to be active on two or more ad exchanges at the same time. The same is true for advertisers who can also expect to acquire the desired ad space on better conditions if they are active on more than one ad exchange. Moreover, as inventory in online advertising is dynamic, publishers are not able to sell it in its entirety on an ad exchange before it is created by the users' behaviour. Such "remnant", unpredictable ad space is also unlikely to have characteristics that would make advertisers buy it directly from publishers.

250. As explained above, the ensuing gap of unsold ad space is filled by ad networks. Ad networks pool advertisers on the one side and publishers on the other and therefore can provide ads for space that otherwise would have gone unsold because no advertiser was explicitly interested in this space. Advertisers may be interested to work with several ad networks at the same time because this enlarges the potential ad space available for the advertiser's ads. Publishers are also interested in working with several ad networks because these offer different remuneration schemes for participants and may also be thematically specialised which publishers may like since it helps them getting thematically diverse ads onto their web pages.

251. All these considerations show that it is very likely that any given website requested by a user could be filled with ads from different sources, e.g. from one or more advertisers who directly or indirectly (through an ad exchange) bought ad space or from different ad networks. At some point, it must be decided which of these ads to transmit to the user's browser. This functionality, the so-called "ad arbitration", is provided by Double Click's DFP. It is designed to take into account a large number of possible targeting criteria and conditions of different advertisers and ad networks to select the ad that is most beneficial for the publisher.

252. If most customers of Double Click already use one or more ad networks or exchanges to sell or buy ad space, there is no apparent reason for them to leave these networks and exchanges and only work with Double Click's intermediation service rather than only adding the Double Click ad exchange to gain access to even more market liquidity. In practice, this would mean that publishers would see whether they could get a better price for their ad space on the Double Click ad exchange than they could on other ad exchanges. If this would be the case, they would sell there, if not, they would sell elsewhere. A DFP customer would therefore have an incentive to move a portion of its inventory to the Ad Exchange only if the Ad Exchange generates higher revenues or proves a better use for the inventory than these other alternatives.

253. Even if they joined Double Click's ad exchange, Double Click's current customers would therefore be likely to go on selling ad space through different exchanges and continue to rely on DFP's ad arbitration to select which ad to serve onto each requested web page. There would therefore be no exclusivity for Double Click's ad exchange in dealing with these customers and its advantage of already having commercial relations with many large customers would not help it much in competing with other ad exchanges.

254. In conclusion, while Double Click's present business relations with publishers and advertisers represent an important asset, it has been shown that it cannot be established that such relations are so unique and can be so easily leveraged into intermediation services that they would give it any better chance of competing in the market where there are many established players, and in particular where such players are already able to offer intermediation and ad serving services.

7.2.2.1.3. Network effects deriving from Double Click's use of information about consumer behaviour collected through ad serving

255. As explained in more detail in paragraphs 181-189, Double Click, as a market leader on the markets for advertiser-side and publisher-side ad serving, collects, on its own servers that host its products DFP and DFA on behalf of its customers, considerable amounts of data records each time an ad is served through DFP or DFA.

256. If this (ad serving) data allowed Double Click to offer a service to its ad intermediation customers that is superior to the service offered by its competitors in the intermediation market which do not have access to this data, advertisers and publishers would inevitably flock to Double Click's ad serving and, by extension, to its newly-created ad intermediation service, by virtue of a direct network effect and Double Click's bundled offering (ad serving plus ad intermediation) could be very well placed to compete with Google's bundled offering (which would be weaker on behavioural targeting but stronger on search capabilities and established as a successful integrated platform).

257. As has been explained in paragraph 183, Double Click is currently contractually prohibited from using data created through the use of DFA for the purpose of improving ad serving to publishers or advertisers other than the one advertiser on behalf of which the data was generated and collected. Similarly, data created through the use of DFP contractually belongs to the publisher whose website it relates to and Double Click cannot therefore make this data available to other publishers or advertisers nor use it to improve ad targeting for other publishers or advertisers. Consequently, the availability of considerable amounts of CPI data does not bring about network effects benefiting Double Click in the ad serving market, let alone in the intermediation market.

Potential future use of DFA and DFP by Double Click

258. These contractual restrictions were created when Double Click was only an ad serving provider and this neutrality is considered as an asset to be able to guarantee each customer the benefit of the same service as all other customers. Taking into account the potential development of Double Click into an integrated intermediation provider, some market respondents expressed concerns that the interests of Double Click to increase the cross use of this data could lead to a modification of the contractual conditions proposed to customers or a pressure upon customers to grant waivers of existing conditions in order to allow this new use and then aim at increasing the collection of data.

259. Indeed, in view of the new role that Double Click could play by developing a bundled solution combining both its (still very minor) intermediation services and its current ad serving activity, it cannot be excluded that such contractual constraints would be modified as Double Click would have a strong interest in increasing its collection of data. Some third parties have also suggested that Double Click could, for instance, offer financial incentives to some of its customers in order to extend the use of data collected through them. Such customers could also be offered a reciprocal use of data collected from other customers to incite them to modify their current contracts. Thereby a direct network effect in the ad serving market to the benefit of Double Click could arise, which could ultimately also benefit Double Click in the intermediation market.

260. However, there is no evidence that Double Click possesses market power that would allow it to impose such far-reaching contractual changes on its customers. Indeed, in the recent past, Double Click has had to make considerable efforts to retain customers that were contemplating a switch to another platform. To prevent such customers from switching, Double Click offered them substantial decreases in the price of its services.

261. Similarly, to prevent customers (advertisers or publishers) from switching to other ad serving providers, Double Click will also likely have to refrain from modifying the contractual terms regarding the use of CPI data. Indeed, on the basis of the market investigation it seems unlikely that, in the future, advertisers and publishers will agree to share much more data with Double Click than they do now. In fact, the submissions of a considerable number of advertisers and publishers rather indicate that they are clearly reluctant to allow their competitors to benefit from the use of 'their' data (135). Advertisers obviously have no interest in other advertisers having access to their data and consequently learning competitively important information such as information about the pricing of ads across different websites. Similarly, publishers would be reluctant to share their data with other publishers competing for the same advertising budgets.

262. In view of all these circumstances, Double Click therefore probably lacks the ability to force upon its current customers a change in contractual relations.

263. Moreover, it is doubtful whether Double Click would have an incentive to try to do so because stopping to be a neutral service provider would require farreaching changes in the company's business model. At present, Double Click's business is focused on its service to customers. As confirmed by the market investigation, advertisers and publishers value Double Click's service in the area of ad serving in the sense that it is time-tested and reliable, scalable and secure. Indeed, many of the respondents to the Commission's market investigation indicated that these characteristics had been the main reason for them to choose Double Click as their ad serving provider. Double Click's incentives to continue to provide such services coincide with its customers' preferences. The better Double Click's service becomes, the better each of its clients will be served. In view of these circumstances, it appears that Double Click would not have any interest in discriminating between its ad serving customers by offering different service levels or by favouring some customers to the detriment of others. This neutral stance is coherent with Double Click's contractual engagements discussed above.

264. Changing these contractual provisions would only make sense if Double Click would subsequently actually make use of the data of individual advertisers to the benefit of other advertisers (or publishers). That would, however, jeopardise Double Click's current position of neutrality. For instance, data gathered from one advertiser may be very important and useful to another advertiser, possibly a competitor of the first one, or to Double Click itself if it develops as an integrated intermediation provider. If Double Click could sell information gathered from such data as it pleases, its customers could no longer be sure that Double Click would treat them fairly and equally compared to other customers. This would probably be taken into account by actual and potential customers of Double Click, when choosing their ad serving provider in the future (whereas at present, customers base their choice mainly on technical characteristics).

265. Given the available alternatives to Double Click's ad serving technology, such a fundamental change in Double Click's business model would be likely to cause a considerable number of current Double Click ad serving customers to switch over. Therefore, it is doubtful whether Double Click would have an incentive to change its contractual engagements in the future.

266. From this analysis it follows that Double Click is unlikely to have any particular competitive advantage in the ad serving market and, ultimately in the intermediation market, due to network effects deriving from the CPI data it collects.

267. As regards the scenario in which Double Click offers its customers some compensation for the right to use their data, it must be pointed out that nothing prevents other companies from also offering Double Click's customers compensation for the right to use their data. Consequently, the possibility that new models of joint use of data relevant for ad targeting may emerge in the future is not restricted to Double Click and its customers or, indeed, to the type of data under discussion here.

Data gathered by other online operators

268. It is also useful to recall that several Double Click's competitors are currently already in the position to offer services to their customers that equal those that could be achieved by Double Click if it was not bound by the contractual constraints regarding data analysis and use. Just looking at the technical situation, Yahoo!, Microsoft (both with their recent acquisitions) and even Google without the merger could offer them, building on data collected throughout their ad networks (136). This shows that even in the unlikely situation that Double Click could somehow free itself from the contractual constraints this would hardly allow it to offer unique functionality (137).

269. Moreover, even the merged entity, let alone Double Click alone, would not have access to unique, non-replicable data because the type of information collected by Double Click is relatively narrow in scope. Other companies active in online advertising have the ability to collect large amounts of more or less similar information that is potentially useful for advertisement targeting. The following paragraphs sketch a few examples of such current or potential data collection.

270. For instance, portals and other major web publishers typically require their users to register and provide personal information, and they design their websites to keep users on their web pages for the greatest length of time possible, all the while observing and recording their users' behaviour. As a result, major web publishers can (and do) collect a vast amount of information about their users' interests and actions. Following its proposed acquisition of BlueLithium, Yahoo! will be able to collect behavioural data across both Yahoo!'s own sites and BlueLithium's ad network to target display advertising based on demonstrated user characteristics (BlueLithium and Yahoo! have the third and fifth largest global web audiences, respectively).

271. Internet service providers can track all of the online behaviour of their users, following them to every website they visit. Particularly large internet service providers could thus try to team up with advertisement companies to make use of this data under the restrictions imposed by privacy rules, but they could also try to use this data with their customers' consent, for instance in exchange for lower prices. Several companies offer appliances for "deep packet inspection" of network traffic routed through internet service providers in order to extract information that is meaningful for ad targeting (138). Data collected in this way is potentially much broader and richer than data collected by Double Click (or even the merged entity) or any of its rivals because contrary to the merged entity's data collection efforts, this data source covers every page on every site a user visits.

272. Microsoft also has access to vast amounts of user data through its internetbased update service (139) as well as through its ad agency Avenue A. Its portals, websites and services can derive information from user registrations and behaviour in connection with content pages (the cookies dropped on these content pages are a source of targeting information). In addition, Microsoft gained behavioural targeting technology through its acquisition of Atlas/Accipiter and has been actively publicizing this functionality to potential customers. The partnership with Facebook further increases this database that can be used in combination with its behavioural ad targeting technology (140).

273. It is important to note that much of the information described in the previous paragraphs cannot be replicated by Double Click or even the merged entity because it has other structures and types than the information collected by Double Click or Google. Competition based on the quality of collected data thus is not only decided by virtue of the sheer size of the respective databases, but also determined by the different types of data the competitors have access to and the question which type eventually will prove to be the most useful for internet advertising purposes.

274. To sum up, the current situation would be unlikely to change as Double Click without the merger, would not acquire any additional means to put pressure on its customers to agree to less constraining contract provisions. As a consequence, advertisers contemplating the use of DFA are not compelled to switch to this service because it already serves many other advertisers. There is no network effect in the ad serving market on the advertiser's side. The same is true on the publisher side of the ad serving market for the use of DFP. Apart from the comforting knowledge that other customers also trust the quality of Double Click's services, publishers switching to DFP are not better off even if more publishers join.

275. This finding deprives Double Click's ad exchange of one of its alleged main advantages, this being that publishers and advertisers would supposedly be drawn to it by virtue of its unique network effect based on Double Click's ad serving business and the use it makes of the data it collects through it. While it must be recognised that Double Click collects a given amount of data, this does not give it a strong advantage with regard to its entrance into the ad intermediation business. For the same reasons for which there are no network effects in the ad serving market, there are also no network effects in the intermediation market, as any network effects in this related market would be derived from those which could arise in the ad serving market through the cross use of data created through the use of DFA or DFP.

276. Advertisers and publishers use third-party ad serving technology to enable them to monetize their inventory and would not want information related to their sales made available to competitors. The ability to record and report on the performance of an ad campaign in a way that protects the confidentiality of an advertiser's data is an important part of the business model of an advertiser-side ad server, and it is equally important to publishers that the publisher-side ad server protects confidentiality.

277. Additionally, it is worth noting that the performance of an ad campaign is not comparable across entities. Not only does Double Click have no incentive or contractual right to look at ad campaign performance across entities, the data itself is also of limited utility because each individual client defines its own metrics within the Double Click products. Both advertisers and publishers insist upon strict confidentiality provisions in their ad serving contracts to prevent their rivals from benefiting from their data. The required contractual provisions prohibit the use of the data by Double Click, including the sharing of such data with other customers or any other third party, other than for purposes that the customers themselves expressly permit. The fact that customers do not allow ad servers to share their data with, or to use it for the benefit of, their rivals or their own customers indicates that any value to them from enhanced ad serving performance would be outweighed by the actual or perceived advantages that such data sharing would confer upon their competitors. If Double Click did not agree to, or violated, the confidentiality provisions upon which its customers insist, there would be alternative ad serving providers to which these customers could and would switch. Indeed, it is notable that Double Click did not try to, and in fact never even considered trying to, change the contracts in order to introduce more successfully Ad Exchange.

7.2.2.1.4. Conclusion regarding the elimination of Double Click as a potential competitor

278. It can be concluded from the above that, despite Google's strong market position in intermediation, the proposed merger does not have significant anticompetitive effects with a view to the elimination of Double Click as a potential competitor of Google. While it is not excluded that Double Click would have grown into an effective competitive force in online ad intermediation services, it is rather likely that a sufficient number of other competitors would be left in the market which could maintain sufficient competitive pressure after the merger. Double Click would still be a new entrant which would face not only Google, but also other companies such as Microsoft, AOL, Yahoo!, TradeDoubler, Zanox, AdLink, Interactive Media, Tomorrow Focus, Oridian and ValueClick who also possess strong advantages, including, for some of these players, (in-house) ad serving technology, and are far more advanced in the development of their intermediation services.

7.2.2.2. Google as potential competitor in ad serving tools

279. As indicated above, Google permits display ads to be placed through its AdSense network (but not on Google.com). But advertisers generally chose networks other than AdSense to place display ads because Google is unable to provide the metrics required by advertisers purchasing display advertising. Google therefore decided to develop a new ad serving product for both advertiser- and publisher- side display ad serving (GFA and GFP), which is still in the early stages of development and is not currently available.

280. [...]*, there is no indication of any likelihood that it would grow into an effective competitive force. Google has no significant experience with display advertising or the advanced metrics required by customers and thus needs to develop these products from scratch. There is no indication that it is more likely to enter successfully than others.

281. The technology used in Google's existing advertising business is different from the technology used in publisher and advertiser ad serving products like GFA and GFP. When Google began to develop GFA and GFP, it did not start by adapting existing Google products. With respect to publisher-side ad serving, Google's AdWords system is entirely auction-based, which has little to do with the reservation-based technology that would allow a publisher's direct sales team to make commitments to particular advertisers as to when and where their ads would appear on the publisher's site.

282. Google's relationships with publishers and advertisers do not provide detrimental advantage for the launch of a new product such as GFA and GFP. This relationship is based on the provision of text ad space. Google would have no prior display ad track record with the customer. At the same time, these advertisers and publishers already have established business relationships with other providers of display ad space and display ad serving. Other potential entrants, such as rich media companies, may be better placed in terms of customer relationships, as they may already be providing their customers with rich media ads (including serving of the ad).

283. There are at least two other major categories of well-positioned potential entrants into ad serving, ad agencies and portals. Ad agencies have long experience working with advertiser-side ad serving solutions and dealing with major display advertisers. On the other hand, portals are major web publishers with their own direct sales teams that sell display ad space on their websites. Portals have relationships with major advertisers, and have long experience using publisher-side ad servers to show display ads on their websites. Some of the many recent entrants into ad serving include ad agencies (in particular WPP) and web portals (in particular Microsoft, Yahoo!, and AOL). Google would not have advantage over other ad networks, including those of Microsoft or Yahoo!, that have more unique visitors, more sophisticated display advertising technology, more relevant data and an historical focus on display advertising.

284. In the near future, Google could have developed its technology. Respondents to the market investigation considered that once available, GFA and GFP could become competitors of other display ad services, such as Double Click's.

285. However there is no reason to believe that it would have provided such features that would have made it a competitor particularly better placed than the numerous already present in the market. Even if Google's display ad serving products under development were to succeed, they would become one of many competitors acting in the display ad serving.

7.3. Non-horizontal effects

286. As indicated above, the proposed concentration combines Google's online advertising and online advertising intermediation activities with Double Click's display ad serving technology. The merger can be described as having a vertical dimension given that Google currently offers a bundle consisting of both intermediation and ad serving tools (that is to say the AdSense network "bundled" solution) while Double Click only offers stand-alone ad serving tools that can be used as a stand-alone product ("unbundled") in the indirect distribution channels or the direct distribution channel. With the acquisition of Double Click, Google would therefore control the leading supplier of a key input into distribution channels that compete with its own ad network (AdSense). Moreover, the merger can also be described as having a "conglomerate" aspect given that intermediation and ad serving tools are products that publishers and advertisers can purchase together (though not necessarily). With the acquisition of Double Click, Google would therefore acquire the leading supplier of a product that publishers and advertisers can combine with intermediation services such as those offered by Google's ad network (AdSense).

287. A number of respondents to the market investigation have raised concerns about the potential non-horizontal anti-competitive effects of the merger. In particular, competitors have been vocal about the potential for anti-competitive foreclosure in online ad intermediation by the new entity.

288. The relevant non-horizontal theories of harm the Commission has investigated in this regard can be grouped in three main categories, namely (i) foreclosure scenarios based on Double Click's market position in ad serving; (ii) foreclosure scenarios based on Google's market position in search advertising and online ad intermediation services; and (iii) foreclosure scenarios based on the combination of Double Click's and Google's databases on customer online behaviour (CPI) (141). These theories are discussed in turn below.

7.3.1. Foreclosure strategies based on Double Click's market position in ad serving

289. Concerns have been voiced that the new entity could implement a wide range of exclusionary price and non-price strategies. The exclusionary strategies arising from the "conglomerate" dimension of the merger include (i) increasing the price of Double Click tools when used by publishers or advertisers with competing ad networks or selectively increasing the price of Double Click tools to customers less likely to switch to other ad serving tools suppliers (142), (ii) degrading Double Click tools' quality when used with competing ad networks (143), (iii) bundling Double Click tools with Google's intermediation services (either through pure or mixed bundling (144)) and finally (iv) "tweaking" (145) the ad arbitration mechanism to serve ads in favour of AdSense. The exclusionary strategies arising from the "vertical" dimension of the merger include input foreclosure (that is to say the refusal to sell or raising rivals' costs) in the sale of ad serving tools to ad networks that compete with AdSense.

290. All the theories of harm have a similar thread: if the merged entity was to increase the relative price (146) of using DFP (147) on competing networks, bundle DFP with AdSense or use other non-price strategies to attract inventory, most Double Click customers would move part of their inventory to Google AdSense thus reducing the inventory available to other ad networks. Indeed, in view of (allegedly) high switching costs, DFP customers would be unable to respond to such foreclosing strategies and would be compelled to move part of their inventory to AdSense or stop using other networks altogether. Given that online advertising is a two-sided market characterized by network effects, scale and access to user data are important ingredients of success. Through the foreclosure strategies, the merged entity would deny sufficient scale and liquidity or, in other words, the ability to find easily and quickly a counterpart with which to trade (148), to competing networks which would consequently be weakened. As the network of the merged entity would become larger and "information-richer", it would attract more publishers and more advertisers up to the point where the market would "tip" in favour of the network of the merged entity, enabling it to raise the price of its offering. In view of its large scale and access to CPI, the network of the new entity would be protected by high barriers to entry as no other network would be able to reach a similar size, in particular with the disadvantage of not having access to the same amount of data on users.

291. For publishers, the ability to manage ads through the direct channel and/or a variety of ad networks or ad exchanges is a determining factor for the choice of using the ad serving technology (rather than rely on a bundled solution such as AdSense). In particular, the use of a stand-alone (unbundled) ad serving tool enables the publisher to determine the rules and priorities for selecting ads that will offer the highest monetisation of its ad space. This ability to instruct the ad serving tool is a major attraction of the product and is generally associated with participation in a number of ad networks or ad exchanges. A degradation of the conditions of use of DFP with competing networks (that is to say higher prices, lesser quality, "tweaked" ad arbitration or an outright refusal to provide ad serving services on competing networks) would imply a degradation of the product for the users of DFP. The ability to foreclose will therefore depend on how customers respond to such degradation. In response to product degradation, publishers could switch to alternative suppliers of ad serving tools unless switching costs are high or they could move their inventory to AdSense (this is the effect predicted by most third-party complainants). The effect of such strategies will therefore hinge on whether switching is easy, whether degrading DFP has an impact on the choice of ad network and whether credible alternatives for ad serving tools do exist.

292. It should be noted that attracting publishers and increasing the sale of remnant inventory through the AdSense network is one of the objectives of the acquisition of Double Click by Google. The fact that Google wishes to increase the size of its ad network and its intermediation revenues by improving the monetization of Double Click's customers' remnant inventory might be a concern if rivals were foreclosed to the detriment of consumers. Establishing whether the merged entity's potential post-merger strategies will lead to anti-competitive foreclosure requires a careful consideration of the facts in this case.

293. For anti-competitive foreclosure to arise, a number of conditions must be present as some of the strategies that the new entity might put in place can also have pro-competitive effects (such as lower prices for users or improved quality of ad serving on AdSense, provided such quality improvements are not outweighed by price increases). As indicated in paragraph 94 of the Non- Horizontal Merger Guidelines (149), the (i) ability to foreclose; (ii) the incentives to foreclose; and (iii) the overall impact on effective competition need to be assessed.

294. All the theories of harm presented to the Commission rely on a number of crucial assumptions about the market characteristics. In particular, the existence of high switching costs for ad serving tools, the degree of market power held by the merging parties and the existence of strong network externalities (both direct and indirect) are central ingredients of all these foreclosure theories. Given that the likely foreclosure effects of these exclusionary strategies rely on these assumptions, the Commission's analysis focused on establishing whether these assumptions were confirmed by the market investigation.

7.3.1.1. Ability to foreclose

295. The theories of harm based on the vertical and conglomerate dimensions of the merger assume that the merged entity would have the ability to foreclose competing networks (through the strategies described) because Double Click has significant market power in the supply of ad serving tools, publishers face high switching costs with respect to the ad serving technology, ad serving tools are an important component of the indirect distribution channel for online advertising and there are strong direct and indirect network effects.

296. First, the market investigation has revealed that Double Click faces a number of competitive constraints and is not able to exercise any significant market power. As discussed in paragraphs 119-175, while Double Click is the leading supplier of publisher and advertiser ad serving tools in Europe and worldwide, it faces strong competition from a number of rivals as evidenced, in particular, by the price pressure in recent years leading to price reductions for existing and new customers and the extent of actual customer switching (both large and small customers). Indeed, on the publisher side, Double Click competes with 24/7 OpenAdstream (owned by WPP), ADTECH (owned by AOL), aQuantive/Atlas (owned by Microsoft) and a number of smaller competitors, some of which are strong on a national basis (150) (Smart AdServer/Axel Springer, Openads, Newtention, Adition, Exponential, Adnologies, ValueClick, Adnet, CheckM8, Mediaplex, Eyeblaster and TradeDoubler). On the advertiser side, Double Click competes with aQuantive, ADTECH and several smaller competitors (Mediaplex/ValueClick, BlueStreak/Aegis, Openads, Newtention, Adnologies, Adition, Smart AdServer/Axel Springer, Sapient). Moreover, the threat (by large publishers or ad networks) of developing and switching to an in-house ad serving solution constitutes an additional competitive constraint on ad serving tool suppliers such as Double Click.

297. In view of these competing alternatives, the level of switching costs is relevant for the competitive assessment because high switching costs may enhance the degree of market power of the parties (by raising barriers to entry or expansion) and facilitate foreclosure as customers are less able to react to foreclosure strategies (such as price increase, quality degradation or bundling). As indicated in paragraph 138, the market investigation has revealed mixed opinions. Switching costs are considered to be significant by a number of customers (including customers that have never switched) and modest by others. Yet, the evidence on actual switching suggests that the obstacles to switching are indeed manageable. The notifying party has also provided evidence on price decreases achieved by Double Click's customers over time through the renegotiation of contract terms at the renewal of the contract, as referred to in paragraphs 168-175. Even larger publishers (for whom the cost of switching is arguably more significant) have either switched suppliers or renegotiated contract terms to achieve lower CPMs. Contracts also tend to be renegotiated relatively frequently and contract duration is relatively low. This is significant because it suggests that customers consider the threat of switching to be credible. Indeed, if competition is strong and switching is possible, customers will tend to prefer short-term contracts allowing for frequent renegotiations. While a third party claimed that contracts for ad serving tools tended to last about 3 years, the notifying party has provided evidence that the majority of Double Click's contracts do not last more than 2 years ([...]* of Double Click's European contracts last up to 2 years - measured by 2006 revenues - and [>70%]* - measured by number of contracts) (151). Double Click's largest European customers (with more than 1 billion impressions in 2006) have either [...]* contracts (152). Any significant price rise would therefore likely trigger switching towards competing ad serving tool suppliers or threats to do so as customers have the frequent opportunity to renegotiate prices.

298. A third-party complainant has also raised the issue that post-merger, the access to CPI by the new entity will create an additional switching cost. As discussed in the discussion regarding the data combination in paragraphs 359- 366, the combination of the Google and Double Click customer "databases" is unlikely to provide a considerable additional competitive advantage to the new entity.

299. The issue of whether price rises in ad serving tools could trigger significant switches to Google's ad network was also considered. Ad serving tools are an input into the delivery of ads and reporting services for publishers and advertisers involved in the direct and the indirect "unbundled" channels. However, as discussed in paragraphs 195-202, the cost of ad serving only represents between 2 and 5% of online advertising costs/revenues for advertisers and publishers (153). From the point of view of the publisher, once the decision to use a stand-alone ad serving tool (for direct and/or indirect sales of inventory) has been made, the allocation of the inventory between the various distribution channels will depend on the value at which online space can be sold on each of the channels. The publisher will compare the net profit realized on each channel and distribute its inventory accordingly. The fact that ad serving represents a small fraction of net ad revenues has implications for the likely switching behaviour with respect to ad networks. Indeed, a price increase of 10% of the ad serving tool would lead to minimal variations in the net profit received by the publisher on AdSense's competing networks and this is unlikely to lead to any significant switch towards AdSense. In order to induce any significant switch between ad networks, a substantial relative price increase would be necessary (for instance, a very significant price increase for DFP used on other networks). However, in view of the evidence on switching costs and the degree of competition on ad serving tools, this seems rather unlikely.

300. With respect to ad networks, ad serving tools are a key input (154). However, while ad networks also use third-party ad serving tools (and could switch if the price of DPF or DFA was to increase), they also have a greater tendency to develop their own in-house technology as illustrated by the examples outlined in paragraphs 176-177. Indeed, both large and small ad networks have developed their in-house ad serving tools such as Edintorni, Ligatus, ValueClick, Quigo, BlueLithium/Right Media, Advertising.com, ad pepper or WPP (for the publisher-side). There are examples of ad networks and ad exchanges having switched ad serving provider such as [...]* that switched from OpenAds to Double Click or [...]* that moved from Double Click to Atlas. The ad network ad pepper indicated that they switched "several times" before developing their inhouse solution (in 2007). The ability to foreclose ad networks by either refusing to sell Double Click tools to ad networks or increasing the price of these tools when sold to ad networks will therefore also depend on whether ad networks have other alternatives. In view of the evidence on switching costs and the degree of competition on ad serving tools, it can be concluded that ad networks would have alternatives to the parties' ad serving tools.

301. In the absence of significant switching costs and in view of the possibility to switch to alternative suppliers of ad serving tools (including the possibility to develop in-house solutions), the ability of input foreclosure therefore also appears to be rather limited.

302. The presence of network effects would also be another crucial ingredient of the theories of harm. Indeed, if the merged entity was able to attract customers to its own intermediation platform through lower DFP prices when used with AdSense or bundling (or other strategies), the ability to foreclose could be reinforced by the presence of network externalities as competitors might be deprived from a sufficiently large customer base (that is to say from sufficient "liquidity") or from the benefit of direct network externalities or in other words, from the access to superior CPI allowing of the improvement of targeting.

303. One third-party complainant has put forward the presence of direct network effects which would give the new entity a comparative advantage, difficult to replicate by competitors. As Double Click is the leading provider of ad serving tools, it has a wide customer base both on the publisher and the advertiser side. This large publisher customer base improves the quality of services through the information gathered by Double Click's servers, in particular, the targeting of adverts. As adverts are better targeted, more publishers and advertisers are attracted by Double Click's services (155) and ad targeting improves even further. Few respondents to the market investigation confirmed the view that the access to more data about user behaviour should improve the targeting (though one of these respondents indicated it was not clear how the ad server could use this data across publishers as this is forbidden by the contracts). Indeed, the notifying party has indicated that such use of the data was contractually not feasible as publishers themselves requested such use to be limited. Publishers insist that data use restrictions are included in their contract terms. The theory of harm relying on data combination (and access to CPI) implies that the new entity would have to renegotiate these terms despite the fact that customers seem to value highly the confidentiality of their datasets (or alternatively that the new entity would have to fraudulently use this data). In any case, the type of behavioural targeting that lies at the core of these direct network effects is an emerging technology which neither Double Click nor Google have developed, contrary to a number of competing firms (such as Yahoo!'s ad network BlueLithium or AOL's Tacoda network).

304. The presence of strong indirect network effects lies at the core of most thirdparty complainants' theories of harm. These network effects are such that an ad network becomes more attractive to advertisers as the number of publishers increases (and vice versa). The reason put forward is that liquidity is key to success in online ad intermediation and more liquidity is achieved through scale. With a higher number of publishers and advertisers participating in an ad network, the probability and expected value of a match increases. Through the foreclosure strategies described above, if the AdSense network is able to attract additional publishers (or inventory), it will reach a critical size while denying the necessary scale to competing ad networks. According to some complainants, the presence of indirect network effects provides additional incentives to engage in foreclosure strategies as rival networks are more likely to be weakened. While the presence of these network effects is theoretically compelling, the evidence gathered during the investigation suggests that these may not be as strong (or at least, not strong enough to lead to "tipping").

305. The market investigation has provided evidence that there is significant entry and competition in online ad intermediation as well as evidence on the prevalence of multi-homing and the ability of ad networks to compete even with a relatively small number of partners on the publisher side. The prevalence of multi-homing suggests that the participation by a publisher or an advertiser to an ad network (e.g. AdSense) does not imply that they are unable or unwilling to participate in another ad network, that is to say their participation to an ad network is not exclusive. This may result from the fact that the fixed cost of joining an ad network is either very low or non-existent (as indicated by the notifying party). Multi-homing is also enabled by the interoperability of the ad serving technology allowing publishers and advertisers to provide instructions across several networks. For example, RightMedia's Direct Media Exchange is a free ad network manager allowing publishers participating in several ad networks to automatically display ads on the network paying the highest price per impression (in real time bidding).

306. Data provided by the notifying party shows that more than half of Double Click's customers used at least 2 ad networks (156) (and sometimes more than 5). A survey of European and United States publishers provided by the notifying party indicated that the median number of ad network participation by European publishers was 2 (while it was 5 in the United States) (157). The notifying party also provided (partial) information on the participation to ad networks of the top 50 EU websites (based on NielsenNetRatings). The data indicates that about half of the publishers that use intermediation platforms use more than one ad network and/or ad exchange.

307. The prevalence of multi-homing is consistent with the fact that many ad networks and ad exchanges have developed in parallel and are growing. The market investigation has revealed that a large number of ad networks have experienced significant growth in recent years (growth in net revenues has been more than 50% for a large number of ad networks). Both small and large networks have also seen the number of publishers and advertisers participating grow in recent years.

308. The fact that several ad networks are able to coexist and grow is further illustrated by data on the reach achieved by a number of ad networks. Reach, that is to say the number of unique visitors per month relative to the online population, is an important parameter for joining a network, however, it is one amongst other parameters such as the targeting capabilities or the quality of advertisers/publishers participating in the network. The notifying party has provided data on the reach achieved by a number of ad networks (relative to the United States online population) showing that ad networks such as Advertising.com, ValueClick, Tribal Fusion, Casale Media Network, Blue Lithium, Specific Media, DRIVEpm and others, each achieved reach percentages of more than 50% showing that many co-existing ad networks are able to reach a large online population with their members (158).

309. In any case, both the parties and respondents to the market investigation seem to agree that the intermediation market is growing (as targeting improves, additional inventory can be optimally sold through ad networks).

310. To sum up, post-merger the new entity will control both a large intermediation platform (AdSense) and a leading supplier of ad serving tools for display ads. It will thrive to attract additional publishers to its intermediation platform but the evidence reviewed does not support the view that the new entity would have the ability to foreclose competitors in intermediation markets through leveraging of its leading position in ad serving. In view of the presence of credible alternatives to which customers (publishers/advertisers/ad networks) can switch and the evidence that network effects are not strong enough to induce tipping, any strategy to attract publishers/advertisers to AdSense through input foreclosure or a variety of bundling/tweaking strategies is unlikely to be able to foreclose rivals in intermediation markets.

7.3.1.2. Incentives to foreclose

311. The various strategies that have been put forward by third-parties raise the question of the merged entity's incentives to engage in such strategies. This section discusses in particular the incentives to increase the price (degrade the quality) of DFP/DFA when used with competing ad networks, to "tweak" the ad arbitration process or engage in input foreclosure.

312. The notifying party claims that one of the objectives of the acquisition is to render DFP more attractive to customers in order to achieve one of the major rationales of the merger (i.e. use Double Click's customer relations to expand its intermediation services in particular for inventory that is currently unsold). While the merged entity's incentives to attract additional traffic to its own platform are clear, the incentives to engage in the foreclosing strategies put forward by competitors are less straightforward.

313. Third-parties have claimed that the merged entity would have the incentives to engage in the foreclosing strategies described above because intermediation profit margins are significantly higher than margins from selling ad serving tools (159). The margins lost on DFP sales (that is to say from serving ads from competing networks) would be minimal relative to the additional margins earned on intermediation through the AdSense network. Under the assumption that switching costs are high, third-parties conclude that the merged entity will have incentives to retain DFP customers and attract them to the AdSense network through the pricing and bundling strategies described above.

314. With respect to mixed bundling (160), the parties have provided margin information for DFP customers that also use AdSense. It appears that average revenues on AdSense are significantly higher than those on ad serving. However, the relevant revenues (i.e. those served on AFC) are about the same as those from ad serving services (161). The figures provided by the parties suggest that DFP customers spend relatively limited amounts on Google's AdSense nonsearch platform and therefore, the incentives to offer a mixed bundle might be limited as margins earned on additional sales through AdSense would merely (if at all) compensate the opportunity cost of reducing the price of Double Click tools (or offering it for free). In any case, even if the intermediation margin was substantially larger, it is in fact unlikely - as discussed earlier - that price variations (even significant) for ad serving tools would trigger significant switching between ad networks (162).

315. Pure bundling (163) (that is to say bundling of DFA or DFP with intermediation through AdSense) is likely to be unprofitable. While the margins earned on additional sales of the bundle would exceed those lost on sales of the ad serving tools, the strategy is likely to be unprofitable in view of the switching this might entail. As discussed above, the ability and freedom to input instructions are key dimensions of the quality of the ad serving tool. Any restriction of this ability would amount to a degradation of the product (either through bundling or refusal to serve ads from competing networks) likely to trigger switches from current customers to alternative ad serving tool suppliers. As a third-party recognizes, such strategies would also have long-run implications as degrading the quality of DFP (or DFA) to current customers would lead to fewer publishers and advertisers wishing to install Double Click products in the first place.

316. With respect to the "tweaking" strategy, one third-party has expressed concerns that the merged entity would discontinue the neutrality of the ad arbitration algorithm used by DFP. After the acquisition of Double Click, Google would be in a position to alter the internal workings of DFP, including those governing ad arbitration, in such ways that more ads would be served through AdSense (relative to the pre-merger situation where the algorithm would have selected another network or direct sales).

317. However, there seem to be several disincentives to doing so. Firstly, this strategy would constitute a breach of the merged entity's contractual obligations towards its customers. Indeed, the notifying party submitted that such a use of Double Click's ad server algorithms would constitute fraud to the detriment of its customers because the data that publishers enter into DFP, including priority levels and selection criteria used to select which ads to serve, belong to the publisher and reflect the publishers' view of the best way to optimise its inventory. The strategy would imply overriding these instructions and may constitute a breach of contract. Secondly, such strategy, if carried out on any meaningful scale, would probably be detected. Indeed, a detailed description of the ad arbitration mechanism suggests that such "tweaking" against the instructions of the publisher could not remain undetected for long.

318. The terms agreed between publisher and advertiser in their contract (such as location on the website and time of day) govern the first stage of the ad arbitration mechanism. These terms operate as a first screening criterion to determine which ad to serve. It is only after this initial screening stage that the DFP, more general, criteria (priority, satisfaction and eCPM) pursue the arbitration mechanism.

319. In Double Click's system, publishers manually input the priority level they wish to assign to particular types of ads. Publishers base their priority placements on the return on inventory that various categories of ads are likely to produce. The DFP user interface allows for entry of sixteen different priority levels; DE offers over 100 priority fields. If Google were to change a Double Click publisher's priority ranking, this would be immediately visible at the user's interface. Not only would publishers demand to know why their settings had been altered, but Google would be unable to profit from the tampering because publishers could manually correct their priority rankings to move AdSense to a lower priority.

320. Following this, the satisfaction index criterion comes into play. Among ads at the publishers' designated priority level, the ad server determines, on the basis of the number of impressions guaranteed by the publisher's ad sales contract, whether each advertiser's ads are on schedule to be served the agreed number of times (164). If DFP was altered in order to interfere with the satisfaction level, publishers would fail to meet their contractual obligations to show ads for which their advertisers had paid. Repeated under-delivery of a given advertiser's ads would be clear to both advertisers and publishers through the reports Double Click generates and would cause Double Click's publishers to be in breach of their ad sales contracts with their advertisers.

321. Lastly, although rarely considered during the ad selection process, the eCPM criterion is, in the notifying party's view, fully transparent because publishers know the value of their ad sales relationships and closely monitor their ad serving reports. If more than one ad is available at the highest available priority level, and if there is no satisfaction deficit to make up, the CPM criterion selects ads in order of revenue generation for the publisher. DFP determines which ad pays the highest CPM and delivers that ad.

322. Publishers are aware of the revenues they should be earning from each of their selected distribution channels because they individually negotiate the terms for their directly sold inventory and ad networks state their effective CPM rates when pitching for business. The notifying party observed that Double Click's history of relationships with customers shows that customers notice even small changes in ad delivery algorithms (165).

323. The ad arbitration mechanism indicates that the probability of detection of such tweaking would be quite high. Given that "neutrality" is a key quality of the ad serving tool, the merged entity would unlikely have the incentive to engage in a practice that, even with a modest risk of detection, might strongly affect its business in the long-run by inducing customers to switch from DFP altogether or to exclude AdSense from their list of intermediaries. Reputational costs would also be likely to be large if the new entity was discovered to engage in such fraudulent conduct.

324. With respect to the possibility of input foreclosure, it would appear that incentives are limited. The issue is whether the cost of refusing to sell (or increasing the price) of Double Click tools to competing ad networks would outweigh the benefit of increased intermediation sales. However, it is unlikely that such a price increase or refusal to sell would trigger a significant increase in the inventory sold through AdSense as ad networks would have other ad serving suppliers to turn to (or they could develop their own in-house solution) and moreover, competition in intermediation is fierce. In this context, it is unlikely that the strategy would result in competing ad networks becoming more expensive thus triggering a (profitable) switch towards AdSense.

7.3.1.3. Anticompetitive effects of the merger arising from foreclosure strategies

325. The discussion above has illustrated that the new entity is unlikely to have the ability and incentives to foreclose intermediation competitors through a variety of price and non-price strategies.

326. However, even assuming that all these strategies were to be implemented by the merged entity and succeeded in attracting inventory sales from rival networks to its own ad network, it remains questionable whether rivals would be marginalised and customers would be harmed in the long run, or in other words, whether the overall impact on competition would be negative, taking efficiencies into account.

327. Firstly, post-merger, the new entity would continue to compete with a number of platforms offering the same product combination (bundle competition). As a result of a recent wave of acquisitions, the market has evolved to a situation where "bundled" platforms (intermediation + ad serving tools) now coexist with independent suppliers of inputs for online advertising (ad networks and ad exchange offering intermediation only and suppliers of standalone ad serving tools). In response to bundling strategies by the new entity, the competing platforms could respond by offering similar bundles. As discussed above, these platforms include large players such as Microsoft (166) (with aQuantive, Atlas, DRIVEpm and AdECN), Yahoo! (YFP, Yahoo! SmartAds RightMedia and BlueLithium), AOL (167) (ADTECH, Advertising.com, Tacoda) and WPP (24/7 RealMedia, OpenAdstream). Moreover, the market investigation has also identified a number of intermediation platform that have developed their own in-house ad serving technology, such as ValueClick, Zanox, Quigo and TradeDoubler.

328. Finally, the evidence on network effects and the competitive advantage that the combination of data may confer, as well as the evidence of entry and growth into intermediation are not consistent with the view that the new entity would be able to exert market power in the long-term due to high barriers to (re)entry.

329. To sum up, the Commission has examined the potential for anti-competitive foreclosure arising from the acquisition by Google of Double Click, a leading ad serving tool provider. The theories of harm were based on both the vertical and conglomerate dimensions of the merger and focused on strategies through which the new entity might use the leading position of Double Click to enhance the market position of Google's intermediation platform at the expense of rivals and ultimately of consumers. These theories of harm relied on a number of assumptions regarding the characteristics of the markets under consideration, such as Double Click's degree of market power, the level of switching costs, the presence of direct and indirect network effects and the share of ad serving in the costs of intermediation. The market investigation has revealed that Double Click is not able to exercise a significant degree of market power, that switching costs are manageable, that ad serving represents a small proportion of ad costs/revenues and a limited proportion of intermediation revenues and that network effects are not strong enough to induce tipping. In such circumstances, it is unlikely that the new entity would have the ability or incentives to engage in foreclosing strategies and that any such strategy, if implemented, would lead to the marginalisation of rival ad networks and ad exchanges to the detriment of consumers.

7.3.2. Foreclosure strategies based on Google's market position in search advertising and ad intermediation services

330. The second category of non-horizontal concerns described above takes Google's market position in search advertising and (search) ad intermediation services as a starting point assuming that Google may attempt to foreclose rivals by bundling its sales of search ads or its intermediation services for the sale of search and/or non-search ads with Double Click's ad serving technology. Google's search ad services or its (search) ad intermediation services would thus be the bundling services and Double Click's ad serving would be the bundled service (168). These strategies can be conceived if search and non-search advertising are deemed to be in separate markets.

331. Practically, this would mean that advertisers wanting to place search ads via Google or via Google's (search) ad intermediation (AdWords) would be (contractually) required to make a certain minimum use of DFA in case they use display ads at all. Equally, publishers wanting to use Google's (search) ad intermediation could be obliged to use DFP, either on a contractual basis or by means of a technological tie, whereby publishers could only market their inventory on AdSense if they use DFP. Alternatively, Google could use its pricing to induce advertisers and publishers who utilize AdWords or AdSense to (voluntarily) utilize DFA or DFP. The main concern in this context is foreclosure in the sense that the acquisition of Double Click may confer on Google the ability and incentive to leverage its strong market position with regard to the provision of online advertising space for search ads or the provision of (search) ad intermediation services into the market for the provision of ad serving for display ads, thereby reducing the ability and incentive of actual or potential rivals in the ad serving market to compete. Ultimately, the merged entity's strategy could be to use its strengthened position in the ad serving market to impose an even wider bundle on advertisers and publishers, which would include also Google's nonsearch intermediation services, thereby foreclosing also its actual and potential rivals in non-search intermediation.

332. The Commission analysed these concerns and found that the proposed transaction would not bring about such a degree of foreclosure that competition would be significantly impeded. While it cannot be excluded completely that Google may have the ability to foreclose its rivals by bundling the provision of online advertising space for search ads or the provision of (search) ad intermediation services with Double Click's ad serving technology (and to ultimately extend such strategy by also including non-search intermediation in the bundle), the Commission found that the merged entity would most likely not have an incentive to adopt such a strategy. In any event, such a strategy would not have a significant detrimental effect on competition because a number of financially strong, vertically integrated competitors would not be foreclosed.

7.3.2.1. Ability to foreclose

333. As recognised by the Non-Horizontal Merger Guidelines, in order to be able to foreclose competitors, the new entity must have a significant degree of market power (which does not necessarily amount to dominance) in one of the markets concerned. In particular, the Guidelines note that the effects of bundling or tying can only be expected to be substantial when at least one of the merging parties' products is viewed by many customers as particularly important and there are few relevant alternatives for that product. (169)

334. On the basis of these general rules set out in the Guidelines, the notifying party has claimed that Google lacks the market power it would need to be able to foreclose rivals. More specifically, in the notifying party's view the narrowest possible relevant market is the overall market for the provision of online advertising space without any further division between search and non-search advertising space. In this broader market Google's market position is much more limited in the various Member States and in the EEA as a whole than its market position in the narrower search ad segment. According to the notifying party, in the overall online advertising market Google therefore does not have a sufficient degree of market power to be able to foreclose competitors in the ad serving market.

335. The notifying party further argues that, even if search advertising were to constitute a separate product market, this market is a dynamic market in which market positions can change quickly as a result of new technological developments (170). The notifying party points out that, in search advertising, Google faces competition from companies such as Microsoft, Yahoo! and Ask.com (171), all of which have only recently developed several improvements in their search advertising services and have announced further investments and improvements relating to their search advertising offering during the coming years. In this context, the notifying party also notes that there are virtually no costs for users of search engines or advertisers associated with switching among search engines or using multiple search engines. Accordingly, the use of multiple search engines is common for advertisers (172).

336. Moreover, the notifying party submitted that in a number of regions worldwide, Google faces competition from successful local search providers (173). In addition, the notifying party noted that vertical search providers focusing on providing information to users interested in a particular type of search result or topic (such as job seekers, recruiters, house hunters, travellers, doctors and patients) would further constrain the merged entity, in particular because vertical search engines have the potential to provide more relevant and focused results to users than broad-based search engines and moreover, they have the potential to provide advertisers with audiences that are particularly interested in the products and services they offer.

337. Despite these arguments put forward by the notifying party, on the basis of the results of the market investigation the Commission cannot exclude that Google has a sufficient degree of market power to be able to foreclose rivals in the ad serving market by way of the described bundling strategy, at least in search advertising or search intermediation (if search intermediation were to constitute a separate market), but most likely also in a possible overall intermediation market comprising both search and non-search intermediation (174). Google is in a very strong position in the search ad segment with market shares ranging from 50-80% in the various Member States, both in direct sales (relevant for advertisers) and intermediation (relevant for advertisers and publishers). Google's search ad services are considered as those with the best ad targeting. It therefore seems that many advertisers depend on Google's search ad services and that the revenues derived from Google's (search) ad intermediation make Google an almost irreplaceable source of income for many publishers. In most cases, Google's search ad competitors (including search ad networks) do not seem to be a real alternative as they do not achieve the same level of ad targeting. This is also reflected in Google's constantly growing market share in search ads.

338. Only with regard to non-search intermediation is Google's market position weaker. Although Google also has a significant non-search intermediation market share of at least 40%, Google seems to face considerable competition in this segment. Other players offering non-search intermediation such as TradeDoubler, Zanox (belonging to Axel Springer), AdLink, Interactive Media (belonging to Deutsche Telekom), Advertising.com and Lightningcast (both AOL/TimeWarner) and Tomorrow Focus, but also Yahoo! and Microsoft, Oridian, GWB media-marketing GmbH, WPP and VCMedia (belonging to ValueClick) exert significant competitive pressure on Google. In addition, there are competitive constraints from direct sales. It therefore appears that, assuming that non-search intermediation is a separate market, Google could lack the necessary market power to foreclose rivals in the ad serving market by bundling its non-search intermediation services with Double Click's ad serving technology.

339. However, for a number of other reasons the merged entity's ability to foreclose competitors by adopting the described strategy of pure bundling may be limited, not only in case of Google's non-search intermediation services as the bundling product, but also in case Google uses its search ad and its (search) ad intermediation business as the bundling product, and, more importantly, the merged entity most likely lacks the incentive to engage in the described strategy, both on the advertiser side and on the publisher side.

340. Firstly, there may be practical difficulties in requiring advertisers wanting to place search ads via Google (or search ads and/or contextual ads via Google's ad intermediation (AdWords)) to use DFA. Search ads and contextual ads sold by Google or through Google's ad intermediation services are priced on a cost-perclick basis. The prices charged to advertisers are determined by an auction. As a result, the terms according to which Google provides (search) advertising may be set with an individual advertiser on a daily basis. Advertisers can vary their bids on different search terms as often as they wish, or withdraw altogether from advertising with Google at any time. On the other hand, the terms according to which Double Click provides display ad serving are set by contracts that typically have a duration of one to two years. As a result, it would be difficult to set the terms for search advertising or (search) ad intermediation and display ad serving simultaneously. Pure bundling, however, is very unlikely to be possible if products are not bought simultaneously (175). There would therefore have to be a substantial change in the way the merging parties carry out their business if bundling were to be feasible, which makes bundling more difficult to put into practice, and less likely to occur. The same practical difficulties would apply in case the merged entity decided to extend the bundle so as to include also nonsearch intermediation. Indeed the only conceivable manner for Google to engage in pure bundling on the advertiser side without the described practical difficulties would be by making the use of DFA a precondition for advertisers to participate in the AdWords auction at all. As will be seen below, however, given the low margins on DFA compared to the margins on search advertising, Google would clearly lack the incentive to engage in such an extreme form of pure bundling.

341. On the publisher side, the practical difficulties in bundling Google's (search) ad intermediation with DFP appear to be more limited because for the provision of both display ad serving and (search) ad intermediation for (larger) publishers contractual arrangements of a similar nature and duration apply. As regards display ad serving, Double Click's contracts with publishers typically have a duration of one to two years. Regarding Google's contractual arrangements with its AFS partners, Google enters bidding contracts for the inventory of certain larger publishers, known as "direct partners" (176). The contracts with these European direct partners have a similar duration as Double Click's ad serving contracts with publishers, the average duration being approximately [<3]* years. Even the parties therefore concede that the contractual differences between AFS partners and DFP customers are relatively limited. The practical difficulties described for the advertiser side thus do not appear to apply to the publisher side to any similar extent. Such practical difficulties therefore do not appear to affect the merged entity's ability to engage in the described bundling strategy on the publisher side.

342. Secondly, the extent to which the described bundling strategy results in foreclosure depends, amongst other things, on the extent to which demand for the merged entity's rivals' ad serving products is reduced by the tie. In this respect, the extent to which there is a common pool of customers that purchase both products may be relevant. If there are lots of users of DFA or DFP or competing ad serving products who do not sell or purchase search ads or whose sales or purchases of search ads are very low in value, then the effect on demand is less likely to be significant, and foreclosure less likely to occur. On the other hand, foreclosure is more likely to be effective, if there is considerable overlap between advertisers and publishers that use Google's search ads or its (search) ad intermediation services and advertisers and publishers that use DFA or DFP or competing ad serving products.

343. According to the notifying party, only [less than 0.1%]* out of Google's more than [...]* advertiser customers in the EEA which purchase search ads from Google (either directly or through intermediation) also use DFA. The notifying party estimates the share of these overlapping customers on the advertiser side represented in Google's search advertising revenues (including direct sales and intermediated sales) to be approximately [<20%]*. On the publisher side, the corresponding numbers and percentages are even lower. In fact, only [less than 0.1%]* out of Google's more than [...]* publisher customers in the EEA which sell advertising space through Google AdSense (only [...]* of which sell search ads whereas [...]* publisher customers sell only contextual ads, and [...]* sell both search ads and contextual through Google) also use DFP and the share of the overlapping customers on the publisher side represented in Google's AdSense net revenues is only around [<10%]*. Corresponding figures for advertisers and publishers which sell or purchase (search) advertising space through Google and which are customers of competing ad serving technology providers are not available. However, as Double Click is the leading provider of ad serving technology to both advertisers and publishers, it is unlikely that such figures would change the overall impression that the common pool of customers of Google and Double Click is currently fairly limited (177). This at least reduces the merged entity's ability to foreclose rivals in the ad serving market.

344. Thirdly, the merged entity will face vertically integrated competitors that could replicate the strategy of bundling search advertising and advertiser side display ad serving as well as search intermediation and ad serving for publishers. These competitors include financially strong (groups of) companies such as Microsoft, Yahoo! and AOL (178). According to the notifying party, the threat of such counter-strategies being employed by its competitors constitutes an additional factor making it difficult, if not impossible for the merged entity to foreclose its rivals by engaging in the described bundling strategy.

345. Given these circumstances, there are notable indications that, practically, Google may not be able to foreclose competitors in the ad serving market, at least on the advertiser side where the products to be bundled are not being sold simultaneously and the pricing mechanism for these products is very different. In view of Google' strong position in the search ad segment and in the (search) intermediation market, however, it cannot be excluded completely, at least on the publisher side, that the merged entity may have the (limited) ability to foreclose rivals by bundling the provision of online advertising space for search ads or the provision of (search) intermediation services with Double Click's ad serving technology.

7.3.2.2. Incentive to foreclose

7.3.2.2.1. Bundling Google's sales of search ads and (search) intermediation services with DFA and/or DFP

346. However, the Commission found that, for the following reasons, the merged entity would not have an incentive to adopt such a strategy because that strategy would not be profitable.

347. Firstly, by requiring advertisers or publishers wanting to place search ads via Google or wanting to make use of Google's (search) ad intermediation services to make a certain minimum use of DFA or DFP, the merged entity would run the risk that some customers would no longer be willing to purchase search ads or (search) intermediation services from Google, either because they would rather not purchase a display ad serving solution at all or because they would rather purchase an alternative display ad serving solution.

348. As margins on DFA and DFP are low compared to margins on Google's direct sales of search ads and intermediated sales of (search) ads, even small volume losses in search advertising and (search) intermediation would outweigh the gain in profits from customers taking up DFA or DFP.

Publisher side

349. On the publisher side, Google's 2006 average revenues (net of traffic acquisition cost (TAC)) from those [...]* publisher customers which make use of both DFP and Google's ad intermediation services for the sale of search ads (i.e. AFS partners) were around EUR [...]*, whereas Double Click's average revenues from these overlapping publisher customers in 2006 was only around EUR [less than 1% of this figure]*. In other words, the merged entity would need to gain more than 35 bundled publishers (which did not use DFP previously) to compensate the loss in search ad intermediation revenues caused by only one publisher refusing the bundle.

Advertiser side

350. On the advertiser side, the disincentive for the merged entity to engage in the described bundling strategy is comparable. Whereas Google's 2006 average revenues from those [...]* advertiser customers which use DFA and also purchase search ads from Google (either directly or through intermediation) was around EUR [...]* million, Double Click's average revenues from these overlapping advertiser customers in 2006 was only around EUR [less than 5% of this figure]*. This means that the merged entity would need to gain more than 20 bundled advertisers (which did not use DFA previously) to compensate the loss in search ad and search ad intermediation revenues caused by only one advertiser refusing the bundle. Similar figures would apply, if the bundle were limited to direct sales of search ads on Google.com (thus excluding intermediated sales).

351. Another hypothetical bundle would include DFA and only intermediated sales of (search) ads, but no direct sales on Google.com. Under such a strategy, the merged entity would force its AdWords customers to opt out of AFS (179), unless they also purchase DFA for use both with Google and third parties. Even though the revenues that Google generates from certain groups of intermediation customers on the advertiser side do not differ significantly from Double Click's average DFA net revenues, it is unlikely that the merged entity would have an incentive to impose this limited bundle. The majority of advertiser spending on AdWords is accounted for by ads appearing on Google.com, with AFS ads accounting only for around [<10%]* of advertiser spending. Currently, most advertisers purchasing AdWords choose not to opt out of AFS because there is no significant extra cost allowing ads to be displayed on AFS. However, this would change if the merged entity imposed the condition that any advertiser wishing to allow their ads to be displayed on AFS, in addition to Google.com, would also have to purchase DFA. Given the small proportion of their total expenditure on AdWords accounted for by AFS, advertisers would be unlikely to switch from DFA's rival products to enable their ads to continue appearing on AFS. The bundling strategy limited to AFS would thus likely result in a large degree of "opt-out" from AFS and a subsequent loss of revenues from ads that would have been placed by AFS (180).

352. In the light of these figures and circumstances, the merged entity would be unlikely to risk losing even only a few customers in its core business of search advertising and (search) ad intermediation, where the vast majority of its revenues are earned and where the revenues from each large customer are high, in an attempt to force its low-margin ad serving products upon those larger customers.

353. Secondly, in the online advertising environment transactions often involve customised solutions or services that are uniquely priced. Bundling is usually an attractive and profitable strategy in order to discern customer's willingness to pay in a context where prices are posted and uniform across customers. In the online advertising industry, bundling would not enable the new entity to increase profits because prices are highly individualized. On the advertiser side, both Google through its auction mechanism for keywords and Double Click through its direct negotiation with customers have the ability to vary the price of their products according to customers' willingness to pay. Similarly, on the publisher side, both Google through its negotiations with its direct partners (which account for around [>80%]* of its AFS revenue) and Double Click through its direct negotiation with publishers, have the ability to vary the price of their products according to publishers' preferences. In such context, one of the attractions of bundling usually disappears.

354. To sum up, the large losses that the described bundling strategy would likely produce in the merged entity's core business, coupled with the limited gains from revenues in ad serving would render any such strategy unprofitable to the effect that the merged entity would lack the economic incentive to engage in such practice.

7.3.2.2.2. Including non-search intermediation in the bundle

355. As regards the possible extension of the bundle so as to include also nonsearch intermediation, the incentives may be different as revenues achieved by Google through non-search intermediation are much more significant than the revenues that Double Click achieves through the sale of its ad serving technology. As a consequence, the merged entity could more easily compensate losses of search customers which refuse to accept the bundle by gains from nonsearch intermediation customers (181). However, as regards the bundling of Google's search ads business with non-search intermediation, the proposed concentration does not bring about any significant change of incentives. Indeed, Google could already engage in this type of bundling, especially with regard to intermediation of contextual ads, but also in relation to the intermediation of display ads, namely by making use of the required ad serving technology under a contractual arrangements with Double Click or any of Double Click's competitors. As has been explained above, not all ad networks and ad exchanges have their own in-house ad serving technology, but a substantial number of ad networks and ad exchanges make use of third party ad serving technology, including DFA and DFP. As Google could do the same already today, the merger does not change Google's incentive to engage in this wider form of bundling to any significant extent.

7.3.2.2.3. Overall likely impact on competition

356. Finally, even in the very unlikely event that the merged entity nevertheless engages in a foreclosure strategy involving the bundling of Google's search ad services or (search) ad intermediation services with Double Click's ad serving (and possibly including also non-search intermediation in the bundle), such a strategy would be very unlikely to have a significant detrimental effect on competition. As recognised by the Non-Horizontal Merger Guidelines, it is only when a sufficiently large fraction of market output is affected by foreclosure resulting from the merger that the merger may significantly impede effective competition. If there remain effective players in either market, competition is unlikely to deteriorate.

357. The proposed concentration is unlikely to result in the elimination of a sufficient number of competitors in ad serving so as to significantly impede competition. Even if, despite the obstacles and disincentives described above, (i) the merged entity in the present case decided to bundle Google's search ad services or (search) ad intermediation services with Double Click's ad serving, and (ii) this foreclosure strategy caused most or all smaller, non-integrated competitors in the ad serving market to exit the market, the implementation of this strategy by the merged entity would still be very unlikely to cause competitors such as Microsoft, Yahoo! and AOL to cease offering ad serving or search ad services. Each of these competitors is vertically integrated and has access to considerable financial resources, which will enable it to continue to exert significant competitive pressure on the merged entity after the proposed concentration has been completed.

7.3.2.2.4. Conclusion

358. In the light of this analysis, bundling of Google's search ad offering with Double Click's ad serving technology seems unlikely to occur as a result of the proposed transaction. In any event, even if it did occur, it would not result in a significant impediment to effective competition.

7.3.3. Foreclosure based on the combination of Google and Double Click's assets

359. Finally, in the case of the third category of foreclosure scenarios set out above, third-parties have alleged that the mere combination of Double Click's assets with Google's assets, and in particular the combination of customer provided information CPI data (generated by the use of internet) obtained by both of them, would allow the merged entity to achieve a position that could not be replicated by its integrated competitors (mainly Yahoo! and Microsoft) or "point" product competitors. As a result of this combination, Google's competitors would be progressively marginalised which would ultimately allow Google to raise the prices for its intermediation services.

360. It is not excluded that, from a factual point of view, the merged entity would be able to combine Double Click's and Google's data collections. Such a combination, using information about users' IP addresses, cookie IDs and connection times to correctly match records from both databases, could result in individual users' search histories being linked to the same users' past surfing behaviour on the internet. For instance, after such a match, the merged entity may know that the same user has searched for terms A, B and C and visited web pages X, Y and Z in the past week (182). Such information could potentially be used to better target ads to users.

361. The notifying party submitted that Double Click's current contracts with advertisers do not allow the use of data regarding which web pages a user visited, in order to better target ads from other advertisers than those that were instrumental in bringing this data into existence, that is to say, the advertiser that had served an ad to the user when the user was visiting the web page. By extension, the merged entity would also be contractually prevented from using that part of its enlarged database originating from Double Click to improve, for example, targeting of search ads on Google's sites or contextual ads in the AdSense network.

362. However, these contracts could be waived, modified or renegotiated either as a result of Google/Double Click's new position or in exchange for any sort of compensation given to the customers concerned by the merged entity.

363. However, as has been noted in paragraphs 258-265, advertisers have no interest in other advertisers having access to their data and thus getting insight into competitively important information such as information about the pricing of ads across different websites. Given this probable lack of ability to force a change in contractual relations, it is also doubtful whether Double Click would have an incentive to try to do so since stopping to be a neutral service provider might prompt customers to switch over.

364. Even if Google's and Double Click's data collections were available as input for Double Click it would therefore be unlikely that its competitiveness would be enhanced in a way that would confer on the merged entity a competitive advantage that could not be matched by its competitors.

365. In this regard it must be noted that the combination of data about searches with data about users' web surfing behaviour is already available to a number of Google's competitors today. For instance, both Microsoft and Yahoo! run search engines and offer ad serving. Competitors may also purchase data or targeting services from third parties such as comScore, a global internet information provider which maintains extensive proprietary databases that provide a measurement of the various ways in which the internet is used. Data is also available from internet service providers, which can track all of the online behaviour of their users, following them to every website they visit. Several companies offer appliances for "deep packet inspection" of network traffic routed through internet service providers in order to extract information that is meaningful for ad targeting (183). Data collected in this way is potentially much broader and richer than data collected by Double Click (or even the merged entity) or any of its rivals because contrary to the merged entity's data collection efforts, this data source covers every page on every site a user visits.

366. From this, it follows that the possible combination of data of Google and Double Click post-merger is very unlikely to bring more traffic to AdSense so as to squeeze out competitors and ultimately enable the merged entity to charge higher prices for its intermediation services.

VII. CONCLUSION

367. For the reasons set out above, it is concluded that the proposed concentration would not significantly impede effective competition in the common market or in a substantial part of it. The concentration should therefore be declared compatible with the common market pursuant to Article 8(1) of the Merger Regulation and with the EEA Agreement pursuant to Article 57 thereof.

368. This Decision refers exclusively to the appraisal of this operation with Community rules on competition, namely whether the merger is compatible with the objectives of the Merger Regulation in that it does not impede effective competition in the common market. As enshrined in Recital 36 of the Merger regulation, the Community respects the fundamental rights and observes the principles recognised in particular by the Charter of Fundamental Rights of the European Union184. In any event, this Decision is without prejudice to the obligations imposed onto the parties by Community legislation in relation to the protection of individuals and the protection of privacy with regard to the processing of personal data, in particular Directive 95-46-EC of the European Parliament and of the Council of 24 October 1995 on the protection of individuals with regard to the processing of personal data and on the free movement of such data 185 and Directive 2002-58-EC of the European Parliament and of the Council of 12 July 2002 concerning the processing of personal data and the protection of privacy in the electronic communications sector (Directive on privacy and electronic communications)186 and Member States implementing legislation, which apply to the processing of personal data activities performed by the parties to the merger and by the entity resulting from the merger operation. Irrespective of the approval of the merger, the new entity is obliged in its day to day business to respect the fundamental rights recognised by all relevant instruments to its users, namely but not limited to privacy and data protection.

Has adopted this decision:

Article 1

The notified concentration whereby Google Inc. ("Google", USA) acquires sole control within the meaning of Article 3(1)(b) of Regulation (EC) No 139-2004 of the undertaking Double Click Inc. ("Double Click", USA) is hereby declared compatible with the common market and the EEA Agreement.

Article 2

This Decision is addressed to

GOOGLE INC.

1600 Amphitheatre Parkway

CA 94043 Mountain View

United States of America

Notes:

1 OJ L 24, 29.1.2004, p. 1.

2 On 19 October 2007 the notifying party provided commitments. In accordance with Article 10(1) of the Merger Regulation, the original first phase deadline was extended by 10 working days until 13 November 2007. In its decision under Article 6(1)(c) of the Merger Regulation the Commission considered that these commitments did not address its concerns regarding the elimination of potential competitive constraints and about non-horizontal issues. On 8 February 2008, the notifying party informed the Commission that (i) the commitments offered on 19 October 2007 were explicitly made pursuant to Article 6(2) of the Merger Regulation to enable the Commission to declare the acquisition of Double Click by Google compatible with the common market and the EEA Agreement under Article 6(1)(b) of the Merger Regulation and (ii) as the Commission did not declare the transaction compatible under Article 6(1)(b), the commitments were automatically withdrawn upon the adoption by the Commission of the decision initiating proceedings under Article 6(1)(c) of the Merger Regulation on 13 November 2007.

3 See Form CO annex 5.4(ii). A SEM is a company which provides search engine marketing optimization services; Performics also operates an affiliate marketing network, an internet marketing method through which companies reward one or more "affiliate" websites for each visitor or customer gained through the affiliate's marketing efforts.

4 Web publishers provide content or services on their internet pages (e.g. from large publishers such as CNN to very small publishers such as an individual blog). In the same way that the printed press sells advertising space on their pages, any internet publisher can monetize empty space on its internet pages by selling online advertising space.

5 Source: Zenith Optimedia Report March 2007. Also for the EEA, IAB Europe, for example, forecasted that the online advertising spend within the IAB network in Europe during 2007 would reach EUR 11.5 billion, up from EUR 8 billion in 2006; see the press release of 12 December 2007 at http://www.iab.nl/data/documenten/persberichten/pb071212.pdf.

6 Some market participants also refer to intermediation as the indirect channel.

7 For instance a hotel in Brussels advertising on the search page after the user has typed a query for hotels in Brussels. Search ads appear on the web pages of search engines (such as Google or Yahoo!) as well as on publisher websites that have embedded a search "box" on their web pages For example, the website of the Le Monde newspaper has a Yahoo! search tool box on its homepage or the website of the El Pais newspaper has a Google search tool box on its homepage. The mechanism that chooses which ad to place takes into account the price that each advertiser bids for each given word.

8 In this case, a software reads the content of the page and, consequently, the ads match the words contained in the web page, as is the case of an advertisement for a golf magazine appearing on a web page containing an article on golf.

9 Offline classified ads are traditionally grouped entirely in a distinct section of the newspaper/magazine and are generally connected with the geographic area of distribution of the given newspaper/magazine. Online classified advertising allows for specific geographic targeting through geographic specifications that the Internet user can enter in the relevant website (e.g. yell.com; pagesjaunes.fr).

10 The competitive assessment carried out below will focus on the two main categories, search and nonsearch advertising. The "classified" category is relevant for the purpose of identifying the correct market size as explained in footnotes 15, 20, 62 and 70 below.

11 In the case of contextual targeting, the user's preferences or interests are revealed by the content of the Internet page he/she is accessing.

12 A crucial requirement for the improved sophistication and effectiveness of behavioural targeting relates to the information available on web surfing by a given user as well as any other information provided during his/her web sessions, as well as the capability to process, clean and organise that information so that it can be used in an optimal way.

13 The terms "display" and "graphical" can be used interchangeably. Note that the online ad terminology is not particularly precise and "display" is sometimes used as a synonym for non-search ads (e.g. a complainant uses display and non-search interchangeably).

14 Estimate of total EEA market size based on a combination of different sources, including IAB, Jupiter, Enders, OVK, Zenith, Gemius (Europe) and PwC, taking for each country the highest market size estimate available from any of these reports, with corrections for those countries in which Google's 2006 search revenues exceed the size of the search segment as reported by Jupiter.

15 The notifying party's estimate was actually made for the total non-search market segment, not only including contextual and display ads, but also classifieds and e-mail. The notifying party estimated this broader non-search market segment to have a size of EUR 5.6 billion. For the division of the total online advertising market into search advertising and non-search advertising, the notifying party referred to Jupiter data, according to which search advertising accounts for 41%, classifieds for 14% and display ads for 42% of this market. Therefore, when excluding classified ads, the parties' estimate of the non-search segment results in a total market size of EUR 4 billion.

16 The market reports available to the notifying party are those produced by Jupiter, Zenith, Enders, IAB search, PwC, OVK, Nielsen, Forrester, Adex, Gemius and Emarketer.

17 Jupiter report. Other market reports do not distinguish between search and non-search advertising.

18 "European Internet Advertising Expenditure Report 2006", A report prepared by PwC for IAB Europe.

19 According to the IAB report, the search advertising segment accounts for 45% of the total European online advertising market. However, with regard to Google, IAB could not separate search and nonsearch revenues because, when passing on revenues to advertiser members of AdSense, Google does not distinguish between revenues from sales through AFS (intermediated search) and revenues from sales through AFC (intermediated non-search or contextual ads). The percentage of 45% for the search segment reported by IAB thus includes Google's (intermediated) sales of contextual ads. The correct size of the search segment according to IAB is therefore 40% of the total online advertising market of EUR 8 billion (i.e. 45% of EUR 8 billion minus Google's non-search advertising revenues, which amount to EUR [...]* million according to the notifying party.

20 According to the IAB report, the non-search advertising segment (excluding classifieds and e-mail) accounts for 31% of the total European online advertising market. However, for the reasons given in footnote 19, the correct size of the non-search segment according to IAB is 36% of the total online advertising market of EUR 8 billion (i.e. 31% of EUR 8 billion plus Google's non-search advertising revenues of EUR [...]* million).

21 There are in fact several categorizations of online inventory: premium, reserved, non-reserved, remnant or unsold. For the purpose of this decision only the most frequently used categorization (Premium/remnant) will be retained.

22 The distinction between premium and remnant inventory is a matter of the publisher's audience profile and its capacity to carry advertising (i.e. generate demand by advertisers). Larger publishers (i.e. publishers with a larger "reach") are not necessarily those whose inventory carries more advertising, although scale may be relevant.

23 In more detail, the steps leading to an ad being served on a specific web page visited by a user are as follows: when a user visits an Internet page, the publisher's content server delivers the content page with the ad tag (the ad tag is a code contained on the page that makes a program call to the ad server). Then, the ad tag calls the publisher ad server or an ad network's server, following which the publisher ad server enters into communication with the advertiser ad server selected through an ad arbitration algorithm, the parameters of which are determined by the publisher. Subsequently, the advertiser ad server delivers (serves) the ad onto the web page and the user sees the relevant ad on the web page. All these operations usually take less than a second.

24 Double Click's ad serving tools are (a) DART For Publishers, its hosted product for publishers (hereafter DFP); (b) DART For Enterprises, its non-hosted product (hereafter, DE) and (c) DART For Advertisers, its hosted product for advertisers (hereafter, DFA).

25 In parallel, advertisers purchase advertiser side ad serving services from third party providers.

26 These combinations were also called "integrated" or "non-integrated" solutions by third-parties during the investigation. The parties' economic consultants dispute the terminology of "integrated" vs. "nonintegrated" solutions on grounds that even the "non-integrated" solutions offered by intermediaries (ad networks or ad exchanges) always comprise some basic ad serving technology.

27 E.g. Zanox.

28 E.g. AdLink.

29 Ad pepper provides search advertising through the search tool provided by Google.

30 An informal survey carried out by Google showed that [a significant proportion]* of its "online" publishers (i.e. smaller publishers having a standard online contract) and [75-100%]* of "direct" publishers (i.e. larger publishers with directly negotiated contracts) operate their own direct advertising sales teams. The notifying party accordingly submits that these direct advertising sales teams compete directly with Google's AdSense offering.

31 There are no industry reports that split online advertising sales between direct and intermediated for search and non-search. The notifying party provided estimates based on Google's own gross revenue split of direct search sales as opposed to intermediated search sales. Google's total gross revenues in 2006 in the EEA for search ads are EUR [...]* thereof EUR [...]* are sold through its AdSense network, representing [...%]* of search ads revenues sales.

32 Source: party's estimate and Online Advertising Networks 2007 - A Buyer's Guide (e-consultancy). In the United Kingdom, direct sales by publishers are estimated to account for around 75% of overall online display advertising spending (25% for ad networks and exchanges). This is also consistent with the 30% estimate for intermediation provided by a third party's economic consultant. Yet while the value of intermediated sales seems relatively low, it is worth noting that directly sold inventory is usually premium and therefore commands higher prices. Thus, the share of ad impressions sold through intermediaries is likely to be much higher (estimated at more than 50% by Microsoft's economic consultant).

33 According to the parties, intermediation only represented 14% of total non-search ads in 2004.

34 See Zenith Optimedia Report March 2007.

35 The term has been used by a competitor of Google in its answer to the question regarding the differences between online and offline advertising.

36 See Commission decision in Case IV/JV.1 - Telia/Telenor/Schibstedt, 27 May 1998, Case IV/M.1439 - Telia/Telenor, 3 October 1999 and Case IV/M.0048 - Vodafone/Vivendi/Canal Plus, 20 July 2000.

37 For instance, the fact that Google does not track the user beyond the initial click on the ad (with one exception discussed below), derives from Google's disclosed privacy policy whereby it is not allowed to accept ad tags from third parties (e.g. third parties websites to which the user navigates after having clicked on the ad and landed on the advertiser's website). The exception is when a user installs the "Google toolbar" as an add-on to its own internet browser. In that situation, Google is able to track the user's behaviour and produce better metrics, more comparable to those produced by using "cookies", like those placed by Double Click's tools.

38 Regulation concerning privacy protection and advertising is weaker in the field of online media than in the field of more traditional offline media. Its enforcement would also be much more challenging because of the limitations of local rules in addressing the global character of internet.

39 They are not complementary in the "economic" sense that volumes of search advertising would increase if the price of non-search decreased (and vice-versa).

40 * Parts of this text have been edited to ensure that confidential information is not disclosed; those parts are enclosed in square brackets and marked with an asterisk.

41 As explained in greater detail in paragraph 93, Google offers its network of publishers two main products: AdSense for Content (AFC) and AdSense for Search (AFS). AFC places contextual ads onto the publisher's web pages while AFS delivers ads next to the search results of queries initiated from the search box embedded on the publisher's website.

42 Source: Google.

43 See, on the customers' side, AT&T's reply to questions 20 and 21 of the "customer questionnaire 1" and WPP's submission of 8 October 2007, at para 2.21. On the competitor' side, see Microsoft's reply to question 23 of competitors questionnaire; Yahoo!'s reply to question 24 of the competitors questionnaire; True effect's reply to question 23 of the competitors' questionnaire; Linkshare's reply to question 24 of the competitors' questionnaire. This would suggest that "remnant" inventory is not a matter of quality of the publisher but just a matter of its scale, not justifying direct sales because of the cost of having its own sales force.

44 [...%]* of advertising spending at Google in 2006 by advertisers with an EU billing address occurred outside the Member State in which the advertiser was based.

45 AdSense can serve display ads as well but in view of Google's refusal to serve third-party ad tags and to send cookies, advertisers have tended to serve display ads through other networks. Google estimates that its share in the display online ad segment is about [<5%]*.

46 As is the case with the 2006 data reported by Jupiter for 6 EEA Member States, that is to say Germany, Ireland, Luxembourg, the Netherlands, Spain and the United Kingdom, where Google's search revenues exceed Jupiter's estimate of the size of the search segment by up to [EUR [several hundred]* million]*.

47 Market size as reported by Zenith.

48 Estimated as the sum of the total EEA-wide market size (based on a combination of different sources, including Jupiter, IAB, Zenith, PWC, Enders, OVK, Gemius), the total Australian market size (Google Australia/IDC), the total Brazilian market size (Gemius/IBOPE - netratings) and the total United States market size (estimate of search segment from Enders, estimate of display segment from Enders, estimate of "other" from Morgan Stanley).

49 Lower bound estimate of total EEA-wide market size based on Jupiter, where available, and on Zenith for all other countries for which no Jupiter data was available, with corrections for those countries in which Google's 2006 search revenues exceed the size of the search segment as reported by Jupiter.

50 Upper bound estimate of total EEA-wide market size based on a combination of different sources, including IAB, Jupiter, Enders, OVK, Zenith, Gemius (Europe) and PWC, taking for each country the highest market size estimate available from any of these reports, with corrections for those countries in which Google's 2006 search revenues exceed the size of the search segment as reported by Jupiter.

51 The differences in market shares for one given EEA Member State result from the use of different sources of data (Jupiter where available for the upper bound estimate, other marketing companies including Zenith, Emarketer, PWC, Forrester and IAB for lower bound estimate).

52 Lower bound market size calculated by multiplying the notifying party's best estimate of the proportion of search advertising in online advertising in Europe (i.e. 42%, equal to the best estimate of the search segment in Europe, EUR 4.138 billion, divided by the best estimate of the total size of online advertising in Europe, EUR 9.8 billion) by Zenith's estimate of worldwide advertising (EUR 19.4 billion).

53 Upper bound market size calculated by multiplying the notifying party's best estimate of the proportion of search advertising in online advertising in Europe (i.e. 42%, equal to the best estimate of the search segment in Europe, EUR 4.138 billion, divided by the best estimate of the total size of online advertising in Europe, EUR 9.8 billion) by the notifying parties' upper bound estimate of worldwide advertising (EUR 24.1 billion).

54 Austria, Belgium, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Luxembourg, the Netherlands, Portugal, Spain, Sweden and the United Kingdom.

55 Lower bound estimate of total EEA- wide market size based on Jupiter with corrections for those countries in which Google's 2006 search revenues exceed the size of the search segment as reported by Jupiter.

56 Upper bound estimate of total EEA-wide market size based on a combination of different sources, including IAB, Jupiter, Enders, Zenith and PWC, taking for each country the highest market size estimate available from any of these reports. Where the notifying party based itself on a report that did not break down the online advertising market into search, the latter calculated the size of the search segment by applying Jupiter's estimate of the share of the search segment in the total online market to the total market size in the relevant report.

57 Only including those Member States for which Jupiter data is available, see footnote 54.

58 The differences in market shares for one given country come from the use of different sources of data (Jupiter where available for the upper bound estimate, other marketing companies including Zenith, Emarketer, PWC, Forrester and IAB for lower bound). There is no information concerning the total market size of online advertising related to search available for all the Member States.

59 This data takes into account search advertising revenues generated on all websites, such as revenues generated on Google.com and revenues generated on third party websites that are part of the Google network. Whereas revenues generated on Google.com are attributed entirely to Google, around [>65%]* (or more) of the search advertising revenues generated on third-party websites (AFS partners) are passed on by Google to the third-party website publisher. Therefore, the total market is not only shared amongst the search engines, such as Google, MSN and Yahoo!, but a significant part is attributed to the third-party website publishers.

60 Market division as reported by Jupiter, endorsed by the notifying party in its submissions containing market size estimates.

61 In the total non-search advertising market, publishers generate revenues from (relatively more profitable) direct sales and both publishers and intermediators generate revenues from intermediated sales. As Google is active in this market only as an intermediator, its market position has to be assessed only on the basis of its intermediation revenues and not on the basis of the total price of the ad paid by the advertiser (that is to say the gross revenues including the Traffic Acquisition Costs that Google has to pass on to the relevant publishers). For these reasons, Google's market position in the total non-search market is fairly limited.

62 See paragraph 16; search intermediation accounts for roughly [15-25%]* of Google's total search advertising revenues; non-search intermediation accounts for roughly 25% of the total non-search advertising market (excluding classifieds and e-mail); the latter has been confirmed by the Commission's market investigation for the total non-search intermediation market.

63 The value of the overall intermediation sales at EEA level in 2006 has been calculated on the basis of the split between search and non-search advertising revenues of the notifying party's estimate of the overall market value of EUR 9.8 billion. When considering that intermediation is 20% of the search online advertising sector (with a market value of EUR 4.1 billion, see paragraph 16) and that intermediation constitutes 25% of the non-search online advertising sector (with a market value of EUR 4 billion, see paragraph 16), the value of the intermediation segment in the overall online advertising sector is EUR 1.8 billion. Google's turnover of [...]* is [40-50%]* of the total EUR 1.8 billion intermediation sales value. Also note that the market share of Google represents the value of ads sold through its network (that is to say it is based on gross revenues which includes the share of revenues redistributed to publishers, also called Traffic Acquisition Costs).

64 As explained in footnote 61, it should be noted that the market size and market shares for the intermediation market segments have been based on gross revenues (i.e. including the share of revenues that is redistributed to publishers, also called Traffic Acquisition Costs).

65 "European Internet Advertising Expenditure Report 2006", A report prepared by PwC for IAB Europe.

66 Austria, Belgium, Denmark, Finland, France, Germany, Greece, Italy, the Netherlands, Slovenia, Spain and the United Kingdom. According to the notifying party, the remaining EEA Member States account for an additional EUR 540 million in online advertising spending.

67 According to the notifying party, the IAB report underestimates both the French and the German market by about EUR 500 million and the United Kingdom market by about EUR 300 million.

68 Note that the market size and market shares for the intermediation market segments have been based on gross revenues (that it to say including the share of revenues that is redistributed to publishers, also called Traffic Acquisition Costs).

69 According to the IAB report, the search advertising segment accounts for 45% of the total European online advertising market. However, with regard to Google, IAB could not separate search and nonsearch revenues because Google sells all advertising space through AdSense. The percentage of 45% for the search segment reported by IAB thus includes Google's (intermediated) sales of contextual ads. The correct size of the search segment according to IAB is therefore 40% of the total online advertising market of EUR 8 billion (that is to say 45% of EUR 8 billion less Google's non-search advertising revenues, which amount to EUR [...]* million according to the notifying party's submission).

70 According to the IAB report, the non-search advertising segment (excluding classifieds and e-mail) accounts for 31% of the total European online advertising market. However, for the reasons given in footnote 69, the correct size of the non-search segment according to IAB is therefore 36% of the total online advertising market of EUR 8 billion (that is to say 31% of EUR 8 billion plus Google's non-search advertising revenues of EUR [...]* million).

71 Range of possible market size based on results of the market investigation for search intermediation and, for non-search intermediation, a combination of the results of the market investigation and the IAB report (assuming that 25-40% of non-search sales are intermediated, as confirmed by the market investigation).

72 Range of the possible market size based on the results of the market investigation.

73 Lower and upper bound calculated on the basis of the IAB report, assuming that between 25% and 40% of non-search sales are intermediated (as confirmed by the market investigation); the results of the Commission's market "reconstruction" are situated within the indicated range of the market size.

74 Market share estimates based on revenue figures provided by the relevant parties during the market investigation.

75 According to a survey that Google conducted to measure the use of search engines in various countries, [40-50%]* of the 725 respondents in the Czech Republic named Seznam.cz as their primary search engine (compared to [10-20%]* for Google.com and Google.cz).

76 Market share estimates based on revenue figures provided by the relevant parties during the market investigation.

77 As indicated earlier, Double Click's DFA and DFP products are primarily residing on Double Click's systems, which host the overall DART infrastructure, and are accessed by the customers through a web browser. However, software providers may provide "non-hosted" versions of their tools, as is the case for Double Click's non-hosted version of DFP called DE, which runs on the publisher's own server. Double Click does not provide a non-hosted version of its advertiser's tool DFA.

78 This includes BlueStreak/Aegis, Mediaplex and Newtention active on the advertiser side, as well as Weborama, Smart Ad Server, Eyeblaster and TradeDoubler active on both the advertiser and the publisher side.

79 It can be observed that these market share estimates vary from the estimates in Table 6 resulting from the market investigation.

80 For example, one of the several EEA customers that Double Click lost in 2007 represented independently [...%]* of Double Click's publisher-side revenues in the EEA in 2006.

81 Between January and October 2007, Double Click also lost [...]* existing customers and [...]* of its prospective customers in bidding contests.

82 Source: AdTech's website http://www.adtech.info/en/index.html.

83 See: "The ad serving market and competitors", Double Click internal document provided to the FTC.

84 For instance, when assisting [...]*(formerly [...]*) in replacing tags on twelve [...]* portal sites that serve various European countries, the use by Double Click of the [...]* content management system allowed replacing the tags on all portals in less than four hours.

85 See FTC doc. 35.

86 See FTC doc. 36 and 37.

87 Churn rate, as applied to a customer base, refers to the proportion of contractual customers or subscribers who leave a supplier during a given time period.

88 Note that of the [<100]* customers that switched in 2006, [57%]* were Falk customers (Double Click acquired Falk in March 2006). 89 Note that these numbers exclude the revenues generated by the Falk customers in 2005. The loss in revenues is significant on the publisher side as the 2005 revenues from the customers that switched represented [...]* of Double Click's publisher side revenues.

90 The loss in revenues is significant on the publisher side as the revenues from the customers that switched represented [5-10%]* of Double Click's publisher side revenues.

91 See press release of 24 September 2007 from http://www.atlassolutions.com/news-20070924NC.aspx.

92 http://www.e-consultancy.com/news-blog/361053/accipiter-offers-special-ad-serving-migration-packageto- falk-customers.html?keywords=accipiter%2Doffers%2Dspecial%2Dad%2Dserving.

93 Accipiter Offers DART Alternative, Adotas, Mar. 27, 2006, http://www.adotas.com/2006/03/accipiteroffers- dart-alternative/.

94 "You migrate to ADTECH without any effort or risk within a short period of time. ADTECH has a long lasting experience in system migrations and offers professional project management in addition to a well engineered service package which is tailored to your needs. . . . Migration is so easy! . . . The migration occurs with the least effort and best possible result for you! On all levels ADTECH meets the high requirements of an efficient migration such as maintaining and smoothly transferring your data." ADTECH, http://www.adtech.de/en/go_adtech.html.

95 As indicated earlier, the customers that Double Click lost in 2006 accounted for around [<10%]* of its EEA revenues in 2005. However this percentage does not include former Falk customers, amongst which the largest Falk's customers; this share is therefore the lowest bound. In 2007, the customers that Double Click lost accounted for [<10%]* of its 2006 EEA revenues.

96 For example, [...]* uses DFP for some volume but also has an in-house solution (for remnant inventory) and swaps volume to and from DFP on a regular basis, creating large swings in its DFP volume. In a recent illustration, [...]* DFP volume increased by approximately 24% between November 2006 and January 2007, from 1.5 billion impressions to over 2 billion. The next month DFP volume fell to 583 million, a month-over-month decrease of over 71%. Volume has since risen over 44%, to 1.05 billion impressions in November 2007. This example illustrates how an ad network can "switch" on a regular basis between ad serving providers (including in-house).

97 DART Enterprise customers, such as [...]*, have also switched from Double Click (in these cases, to OAS 24/7).

98 The notifying party has provided many more examples such as [...]* that migrated from 24/7 Real Media to Double Click in 74 business days (and is likely to join Yahoo! in the coming months) and [...]* (a major wedding media and services company with 3.2 million unique visitors and 2.4 million ad impressions per month) migrated three of its main websites in 53 business days. [...]* are two of Double Click's larger publisher-side customers by revenue (both are United States customers).

99 A joint press release states that Atlas "will enable Viacom to better monetize its U.S. Web properties, such as MTV, VH1 and Comedy Central, by serving graphical and video ads on those sites, providing an ad serving solution, optimization features, inventory forecasting, real-time reporting and precise audience advertising targeting." Press Release, Microsoft Corp. & Viacom Inc., Viacom and Microsoft Announce Long-Term Digital Content and Advertising Partnership (19 December 2007), http://www.microsoft.com/presspass/press/2007/dec07/12-19ViacomContentPR.mspx.

100 [...]*, one of these websites, receives 14.4 million unique visitors each month.

101 Atlas declared in its press release that the transaction provides evidence that "[l]eading Publishers are switching away from older solutions that haven't kept up with evolving market needs," and that "Atlas AdManager is the top choice among Publishers for ad serving." Press Release, IAC, IAC Advertising Solutions Taps Atlas to Power Ad Serving Across IAC's Network of Sites (6 August 2007), available at http://www.iacadvertising.com/pressitem/32.

102 Source: The notifying party's economic advisers.

103 Source: The notifying party's economic advisers.

104 To prove its point, the same complainant claimed that margins (EBIDTA) have in fact been stable over time but the stability of Double Click's margins was not confirmed by the information gathered by the Commission.

105 [4 customers]* were amongst Double Click's top 10 publisher-side customers in the EU in terms of 2006 revenues.

106 See, for instance, replies to questions 57 and 58 of the questionnaire to customers of 10 September 2007 by eBay, or a major online travel company, or a major French TV operator or a leading US media company.

107 See FTC doc. 15, 16, 17 and 19.

108 [...]*, a direct marketer, left its agency and then its ad serving provider, Double Click, to move its campaign management and ad serving completely in-house between the end of 2005 and the beginning of 2006.

109 The amount of CPI data of Double Click's competitors is deemed more limited because they have less customers and customers of a lesser "quality".

110 See the various steps leading to an ad being served in footnote 23.

111 Indeed, Google provides primarily text ad serving to publishers participating in the Google network AdSense (all ads purchased through the Google network are served by Google's own servers to the websites of those publishers) as an ancillary service. Google does not charge separately for this ad serving function and does not provide this service to third parties independently of the provision of ad space

112 Note that there is a very minimal actual overlap in the market for intermediation of online advertising space, due to the launch in June 2007 by Double Click of its ad exchange which has not yet achieved full commercialization. It is therefore unlikely to exert presently any actual significant competitive pressure on Google's intermediation activities. It could not be excluded, however, that absent the merger, such an exchange would have become a significant competitor in intermediation of online advertising space. In view of this, this issue will be examined below in assessing potential competition concerns. (see paragraphs 222 et seqq.).

113 Publishers will compare the net profit realized when selling ad space through the various channels while advertisers will compare the total cost of placing an ad on an ad space (relative to the reach/targeting achieved).

114 This range covers percentages given by the majority of respondents to the Commission's market investigation. A few respondents indicated that the ad serving costs could be up to 10% of total advertising costs. As regards ad networks and ad exchanges, the market investigation has indicated that ad serving costs broadly account for about 10-15% of intermediation revenues.

115 Using an example provided by a complainant, suppose that a publisher paid 5 cents per thousand impressions in ad serving fees and generated USD 2 per thousand impressions on the sale of its ad space, netting USD 1.95. If ad serving fees increased by 10% to 5.5 cents, the publisher's net revenue would drop from USD 1.95 to USD 1.945, that is to say. a drop of 0.3%. Suppose an advertiser paid 7.5 cents per thousand impressions in ad serving fees and paid USD 2 per thousand impressions to the publisher for the purchase of the ad space, for a total cost of USD 2.075. If ad serving fees increased by 10% to 8.25 cents, the total cost would increase to USD 2.0825, that is to say an increase in price of 0.4%.

116 In the example, the complainant assumes that, in case of an online ad which is sold at USD 2 per thousand impressions (i.e. this the price paid by the advertiser), the cost of intermediation (i.e. paid by the publisher to the ad network) would be around 40 cents, the cost of the publisher ad server would be 5 cents and the cost of the advertiser ad server would be 7.5 cents. Hence, the total cost of intermediation is 52.5 cents (40+5+7.5), of which publisher ad serving represents 9.5% (5/52.5) and advertiser ad serving represents 14.3% (7.5/52.5). See also footnote 114 above on the relation between ad serving costs and intermediation revenues.

117 See footnote 116.

118 From the point of view of the publisher, the cost of ad serving represents 2.5% (5/200) of the price of the ad, 11.1% of the cost of serving the ad (5/45) and 3.2% (5/155) of the publisher's net profit from selling the ad. This last percentage is the most relevant one for the publisher's choice of sales channels.

119 If serving ads through AdSense led to higher net profits to begin with, the publisher would most likely have already chosen to serve through AdSense (in which case a price rise of ad serving on other networks would not alter the choice). If serving through AdSense led to lower net profits (more than 0.3%), it will remain less profitable (in which case a price rise of ad serving on other networks would not alter the choice either). It is only in the particular situation where the AdSense net profit was initially very close but slightly lower (by 0.3%) compared with the net profit of serving on another network that a change in the ad serving price might induce a switch from the unbundled solution to AdSense. This simplistic example also assumes that the publisher does not use DFP to serve ads on AdSense and that AdSense and the other networks are close substitutes, which they may not necessarily be (for instance, AdSense offers mainly contextual, text ads while many other networks offer primarily graphic display ads).

120 See footnote 116.

121 From the point of view of the advertiser, the cost of ad serving represents 3.8% (7.5/200) of the price of the ad.

122 See footnote 114.

123 In the United States, GFP is in beta testing with [>50]* testers. In the EU, GFP was being tested at one site, namely, the United Kingdom website of [...]*, a United States-based company with Internet traffic and sales generated mainly in the United Kingdom. The company initially switched most of its traffic to GFP for testing, but has since switched most of its traffic away from GFP and at the same time concluded an agreement with Yahoo! for ad sales and ad serving in the United Kingdom and Ireland, which will possibly evolve into a worldwide agreement.

124 According to some respondents to the market investigation, this type of "opacity" characterises Google's AdSense network.

125 Indeed, the fact that publisher A uses AdSense to serve ads on the top of its web page while publisher B uses Double Click on the top of its web page does not necessarily mean that for each of these publishers AdSense and Double Click are close substitutes (i.e. publisher A may not consider Double Click served display ads to be substitutes for the top of its web page while publisher B may not consider AdSense to be suitable for its top web page). Moreover, the survey of websites by the complainant did indicate some tendencies (while not systematic). Considering the web pages where ads where served by either Double Click or AdSense, the survey indicated that 76% of the combined Double Click and AdSense ads located on the top right of the page were served by Double Click (and 24% by AdSense). On the contrary, of the combined Double Click and AdSense ads located at the bottom of the surveyed web pages, 33% were served by Double Click and 67% by AdSense.

126 Guidelines on the assessment of horizontal mergers under the Council Regulation on the control of concentrations between undertakings (OJ C 31, 5.2.2004, p. 5).

127 The guidelines actually state that "there must not be a sufficient number of other potential competitors", but in the light of a contextual interpretation, it is apparent that the second condition would neither be met if a sufficient number of actual competitors are left in the market.

128 The revenues anticipated by Double Click in 2008 would represent respectively less than [<1%]* of the EEA and less than [<1%]* of the worldwide total intermediation net revenues achieved in 2006.

129 "The combination of BlueLithium's network and Yahoo!'s engaged audience will provide an unprecedented buying opportunity across what we believe will be one of the world's leading online display ad networks," said Susan Decker, president of Yahoo!. "And BlueLithium's expertise in network management will better enable Yahoo! to manage supply and demand across our network, by balancing advertiser goals with publisher value. Building on our acquisition of Right Media, this transaction demonstrates our commitment to increasing our investments in areas which can most contribute to Yahoo!'s long-term success." See http://yhoo.client.shareholder.com/releasedetail.cfm?ReleaseID=262635.

130 "We believe the addition of AdECN to the Microsoft portfolio is a perfect fit and will create more efficiency for the industry by forming a more robust marketplace between advertisers and publishers, aggregating more supply and demand. This is good for the whole advertising industry." said Kevin Johnson, president of the Platforms and Services Division at Microsoft. "Joining forces with Microsoft will provide the capital and resources to enable AdECN to scale the exchange at a much faster pace, making it more attractive to the advertising networks and other traffic aggregators looking to better serve their advertisers and publishers," said William Urschel, founder and chief executive officer, AdECN. See http://www.microsoft.com/Presspass/press/2007/jul07/07-26AdECNPR.mspx in July 2007 when the deal was announced.

131 For those customers, Double Click does not, and cannot, record impression volumes because the ad does not run through Double Click's servers. Therefore, in calculating percentages of total ad impressions served by Double Click, the percentage will be overstated when ad impressions flowing through a publisher using DE are included.

132 Nielsen/NetRatings and comScore measure internet traffic and are considered by the online advertising industry as providing the main available data.

133 See submission of the notifying party's economic advisers of 13 September 2007, page 4.

134 [...]*.

135 This is indicated by the fear, clearly voiced in the market investigation, that a merged entity would use data in precisely such a way.

136 It must be noted, however, that until now Google has a self-imposed policy not to accept third parties' ad tags and, thereby, not to track a user's behaviour after the first click on the ad.

137 Moreover, this consideration also shows that Google did not have an incentive to acquire Double Click in order use its strong position in the market for search ads to coerce Double Click's customers to enter into less constraining contracts because this would only give it the ability to do something that Google could already do pre-merger.

138 See Bobby White: "Watching What You See on the Web. New Gear Lets ISPs Track Users and Sell Targeted Ads; More Players, Privacy Fears", The Wall Street Journal, 6 December 2007, p. B1.

139 This update service automatically determines content and configuration of Personal Computers running the Windows operating system and proposes updates for download from Microsoft's centralised servers. This gives Microsoft access to information that is potentially very valuable for advertising. For example, it would be very interesting for advertisers to know whether the users they target have an up-to-date modern system (indicating above-average buyer power in view of the required more expensive hardware) or whether they use certain software products that may indicate their profession.

140 "Facebook presents a big opportunity for online advertising, in part because it collects detailed information about its users -- such as their hobbies, favourite music, location, age, and gender -- that can be used to place highly targeted ads" Microsoft Bets On Facebook Stake And Web Ad Boom, by Robert A. Guth, Vauhini Vara and Kevin J. Delaney, The Wall Street Journal Online, 25 October 2007.

141 Double Click's and Google's databases on customer online behaviour would confer the merged entity with an asset that could not be replicated by competitors.

142 To illustrate possible anti-competitive effects, a competitor has suggested that the merged entity might increase the price of DFP to serve all ads but deemed the strategy was "probably unlikely". The same competitor suggested that Google-Double Click might refuse to allow DFP to serve ads from competing ad networks (as an extreme form of raising rivals' costs).

143 The quality of Double Click's ad serving tools when used with competing networks could be degraded by the use of means such as making the Double Click tools technically incompatible with other networks, by failing to develop codes to allow interoperability with new ad networks/exchanges, by serving ads more slowly on competing ad networks or by refusing to offer certain services.

144 A number of respondents have suggested that Google might offer Double Click tools for free when used on AdSense in return for exclusivity agreements or the right of "first-look" to remnant inventory. Under a pure bundling strategy, the merged entity might only sell Double Click tools if used alongside AdSense. Under a mixed bundling strategy, the price of Double Click tools would be lower (even free) for publishers committing to market all or a certain share of their advertisements or their "remnant" inventory via AdSense.

145 The ad arbitration mechanism is the process by which ads are selected to be served on a given web page based on the rules determined by the publisher. Currently, Double Click's ad arbitration is considered to be neutral with regard to competing intermediation networks in the sense that it does not give any preference to a network but simply follows the instructions inputted by the publisher.

146 Or degrade the quality.

147 We focus on the strategies relating to DFP. A similar analysis can be undertaken for DFA.

148 In this case, a "liquid" intermediation platform (ad network or ad exchange) would allow publishers to quickly and easily find advertisers to sell their inventory to and advertisers to quickly and easily find the appropriate inventory for the placement of their ads.

149 Guidelines on the assessment of non-horizontal mergers under the Council Regulation on the control of concentrations between undertakings, adopted on 28 November 2007, http://ec.europa.eu/comm/competition/mergers/legislation/nonhorizontalguidelines.pdf

150 For example, Double Click has lost a few customers to Adition in Germany and Smart Ad Server in France.

151 See submission of the notifying party's economic advisers of, 26 September 2007, page 10.

152 See submission of the notifying party's economic advisers of 17 January 2008.

153 As indicated above, most publishers (and advertisers) responding to the market investigation indicated that the cost of ad serving amounts to 2-5% of their ad revenues/costs (see para. 197 above).

154 As indicated above, ad serving costs account for approximately 10-15% of the intermediation revenues of ad networks. All ad networks and ad exchange need some form of ad serving technology to operate.

155 Better targeting is attractive for both publishers and advertisers as it increases the monetization potential for publishers and achieves greater targeting for advertisers.

156 See submission of the notifying party's economic advisers of 13 September 2007, page 4.

157 Source: expert opinion commissioned by the parties, submitted on 29 October 2007, Exhibit 6.

158 While this data relates to the United States, it illustrates that participation to ad networks or ad exchanges is not exclusive.

159 In particular, the economic advisors of two complainants indicated that Google's intermediation margins on AdSense are significantly higher (about 8 times) than those realised by Double Click on ad serving.

160 Mixed bundling covers the strategies aimed at rendering DFP (or DFA) more expensive (or less performant) when used to serve ads on competing networks.

161 Google's AdSense partners that use DFP generate average revenues of about EUR [...]* on AdSense while they generate average revenues of about EUR [11% of this figure]* for ad serving. However, the bulk of the AdSense margins are generated in search intermediation (AFS) where average margins are EUR [...]* million whereas non-search intermediation (AFC) generates average margins of EUR [0.23% of this figure]* i.e. the same as Double Click's margins. Search intermediation (AFS) does not require DFP to serve ads (as this is done by Google) and therefore, any mixed bundling strategy involving DFP would not affect search intermediation.

162 The same argument could be made on the advertiser side as the new entity may have incentives to sell DFA in combination with AdWords (participation in AdWords by advertisers gives access to the inventory of publishers participating in AdSense). As the sales margins derived from the sale of ad space far exceed the sale margins on ad serving tools, offering intermediation and tools as a bundle would be profitable if it attracted more advertisers towards the Google network. While our analysis suggests that switching is likely to be limited, in the case of advertisers, an additional benefit might arise from increasing the cost of serving ad on competing ad networks as advertisers may consider search and nonsearch ads to be substitutes. In this context, some substitution towards Google's search advertising might potentially slightly increase the profitability of the strategy relative to a similar strategy applied to publishers.

163 This strategy has similar effects as refusing to serve ads from competing networks with DFP.

164 For instance, if the Financial Times, a publisher, has requested that a guaranteed ad be served into the single banner-sized ad slot on its homepage, DFP will first look at the priority level fixed by the publisher. Among all qualifying ads at the same priority level, DFP will determine which ad to deliver based upon the number of impressions that it must serve of each ad and the number of impressions it already has served of each ad. If a Nike ad is scheduled to deliver one million banner-sized impressions to the Financial Times homepage and a BMW ad is scheduled to deliver two million banner-sized impressions to the homepage or the auto section, and the BMW ad is behind schedule due to low traffic to the auto section, DFP will serve the BMW ad to the homepage first before the Nike ad.

165 For instance, a software update recently introduced a bug into Double Click's ad decision algorithms, causing a change in the serving ratios of time-based pre-emptible ads. [>100]* customers noticed the change in Double Click's ad decision algorithms and filed numerous high-urgency support requests both in North America and Europe. Double Click customers similarly would notice a revenue reducing change that favoured AdSense and would respond in a comparable time frame.

166 In fact, the notifying party has indicated that Microsoft has already started to offer bundles including the use of its intermediation services and ad serving tools.

167 In September 2007, AOL announced the launch of "Platform A", and ad buying platform and network intending to integrate AOL's ad network with ADTECH, Tacoda as well as other ad networks the company currently purchased (ThirdScreenMedia - a mobile phone ad network - and Lightningcast - a rich media solution provider).

168 Given Google's high market shares in the overall intermediation market and each of its two possible sub segments (i.e. search intermediation and non-search intermediation), this section considers bundling strategies whereby either of the following three types of intermediation services serves as the bundling service: (1) Google's intermediation services as whole (search & non-search); (2) Google's search intermediation services only; and (3) Google's non-search intermediation services only.

169 See paragraph 99 of the Non-Horizontal Merger Guidelines.

170 This is evidenced by the history of internet search. In the mid-1990s, Excite was the leading search provider. In the course of the 1990s, Excite was passed by Alta Vista. At the end of the 1990s, Lycos became the most popular search engine. Shortly thereafter, Google entered the market and quickly turned into the most-used search engine.

171 It should be noted, however, that Ask.com's search ads in Europe are sold by Google under a syndication agreement.

172 Google estimates that of its top 250 advertisers, at least 72% also advertise on Yahoo!. Apart from that, the notifying party also claims that churn rates are high, not only for competitors, but also for Google's own network, both on the advertiser and on the publisher side. According to Google, both the churn rate of publishers selling ad space via the Google network in January 2006 that did not sell ad space in December 2006 and the churn rate of advertisers using AdSense for Search (AFS) calculated as the percentage of advertisers using AFS in January 2006 that did not use AFS in December 2006, are around [20-40%]* for the EEA. Similar churn rates apply for individual EU Member States, including France, Germany, Italy, Spain and the United Kingdom. It should be noted, however, that these churn rates are based on numbers of advertisers and publishers. The corresponding churn rates in terms of revenues are lower, namely approximately [<10%]* for publishers and approximately [<20%]* for advertisers. This seems to suggest that Google lost mainly smaller publisher and advertiser customers. Apart from that, it is unclear whether these lost customers actually switched to other providers of search advertising or simply stopped using search ads.

173 According to a survey that Google conducted to measure the use of search engines in various countries, [40-50%]* of the 725 respondents in the Czech Republic named Seznam.cz as their primary search engine (compared to [10-20%]* for Google.com and Google.cz); [20-30%]* of the 284 respondents in Russia cited Yandex as their most used search engine, [20-30%]* cited Rambler and [<10%]* Google; in Ukraine, [10-20%]* cited Yandex, [10-20%]* cited Google, [10-20%]* cited Rambler and [10-20%]* AlltheWeb.com; in China, [60-70%]* of the 4 500 respondents named Baidu as their most used search engine (compared to [20-30%]* for Google).

174 If search intermediation and non-search intermediation were to belong to the same product market, Google's search ad offering would still hold such importance for many advertisers and publishers that it would not be possible to exclude that, even in this wider intermediation market, Google would have a sufficient degree of market power to foreclose rivals in the ad serving market.

175 See, in particular, the Non-Horizontal Merger Guidelines, paragraph 98.

176 Google's [...]* direct AFS partners accounted for [>80%]* of its European AFS revenues in 2006.

177 In the light of Double Click's market shares of 40-50%, it seems unlikely that the total number of customers which use both Double Click's or any competing ad serving technology and Google's search ad (intermediation) services would be much more than twice the overlap between Google's and Double Click's customers on the advertiser side and on the publisher side.

178 Microsoft owns a portal (MSN), an ad network, a publisher-side display ad server, an advertiser-side ad serving business (Atlas), an ad agency (Avenue A), a publisher-side ad serving business (Atlas AdManager) an ad network (DRIVEpm) and an ad exchange (AdECN). It has recently also launched its new Content Ads service (in beta format) to all Microsoft adCenter advertisers, which will automatically distribute Microsoft ads to both search results pages and content pages. Finally, it has launched an entirely new technology known as "AdMarket", which is designed to allow publishers to sell premium inventory without using a direct sales force, an ad network or ad exchange. Also Yahoo! owns a portal, a search engine and an ad network, has begun marketing a publisher-side display ad serving solution and has recently acquired a leading ad exchange (RightMedia) and a large ad network (BlueLithium). Finally, AOL operates a web portal, sells online advertising through its Advertising.com network, owns a provider of ad serving technology (ADTECH), owns a rich media solution (Lightningcast), has acquired a mobile phone advertising network (Third Screen Media) and an ad network offering behavioural targeting services (Tacoda). In addition, in September 2007, AOL formally announced the formation of "Platform A", an ad buying platform and network intended to integrate AOL's display ad network, Advertising.com, with its ADTECH, Tacoda, Lightningcast, and Third Screen Media businesses and technologies. See: http://press.aol.com/article_display.cfm?article_id=1297&section_id=14.

179 Google's search intermediation services offered to advertisers through AdWords.

180 A similar argument can be made for AFC (Google's intermediation services for the sale of contextual ads) as also AFC accounts for only [<10%]* of advertiser spending on AdWords and most advertisers currently also choose not to opt out of AFC because AFC is available at no significant extra cost.

181 The market investigation indicated that the commission that intermediators receive for their (non-search) intermediation services on average amounts up to approximately 25% of the total price of the relevant display or contextual) advertisement.

182 A "user" in this context means a specific browser on a specific computer. Generally this is a very good proxy for a specific person.

183 See Bobby White: "Watching What You See on the Web. New Gear Lets ISPs Track Users and Sell Targeted Ads; More Players, Privacy Fears", The Wall Street Journal, 06/12/2007, p. B1.

184 OJ C 364, 12.12.2000, p. 1.

185 OJ L 281, 23.11.1995, p. 31.

186 OJ L 201, 31.7.2002, p. 37.