Relevant market: The General Court of the European Union upholds the Commission’s decision imposing a record fine on the dominant operator for strengthening its dominant position in the markets for general internet search services and intelligent mobile operating systems (Google / Android)

*This article is an automatic translation of the original article, provided here for your convenience. Read the original article.

In a judgment of September 22, 2022, the General Court of the European Union largely confirmed the Commission’s decision that Google had imposed illegal restrictions on Android mobile device manufacturers and mobile network operators in order to consolidate the dominant position of its search engine. To better reflect the gravity and duration of the infringement, the Court fines Google EUR 4.125 billion, using a different reasoning than the Commission in some respects.

Recall that the Commission’s decision, dated July 18, 2018 intervened in a context of political confrontation between the American and European authorities. It had also prompted a strong reaction from former President Donald Trump (who had declared on Twitter "I told you so! The European Union just slapped a Five Billion Dollar fine on one of our great companies, Google. They truly have taken advantage of the U.S., but not for long! ").

In particular, the Commission accused Google of imposing anti-competitive contractual restrictions on original equipment manufacturers [hereinafter "OEMs"] and mobile network operators [hereinafter "MNOs"] in order to protect and consolidate its dominant position on the national markets for general search services. It also attached behavioral injunctions to its decision, requiring it to end the practices in question, namely the pre-installation of applications such as Google Search, Chrome or Play Store, and its restrictions on the development of mobile operating systems [hereinafter SEM] (the anti-fragmentation clauses).

These practices were part of the development of the mobile internet. Google had put in place a global strategy to occupy space on the mobile Internet and to direct Internet traffic from Google Android devices to the Google search engine. Within this framework, three sets of contractual restrictions were identified. Firstly, restrictions inserted in the mobile application distribution agreements [hereafter "ADAM"] by which Google required the FEO to pre-install its general search(Google Search) and navigation(Chrome) applications, before being able to obtain a license for its application store(Play Store). Secondly, restrictions inserted in the anti-fragmentation agreements [hereafter "AAF"] under which OEFs wishing to pre-install Google applications could not sell devices running on non-Google approved versions ofAndroid. Finally, there were restrictions in the revenue sharing agreements [("RSAs")] under which Google gave the OFEs and MNOs a percentage of its advertising revenue provided they agreed not to pre-install a competing general search service on any of the devices in a mutually agreed portfolio ("Portfolio RSA").

Google’ s appeal was essentially rejected by the Court of First Instance, which upheld only the third and first parts of the fifth plea. But the judgment is particularly rich in lessons. The first lessons concern the framework for analyzing practices implemented in a "digital ecosystem". In particular, there are developments on the context of competition between "ecosystems", questioning the competitive links between the Google and Apple systems, on the method of delimiting the relevant market and on the assessment of the dominant position. The other contributions of the judgment concern the analysis of the abusive nature of certain essential points of the open mobile operating system model, both from a technical point of view, with the anti-fragmentation clauses, and from an economic point of view, with the pre-installation of applications on the one hand, and the revenue sharing clauses on the other. The balance of effects underlying these assessments is particularly revealing in terms of the room for maneuver that an ecosystem’s hub may have in the face of possible complaints of hindrance to inter-ecosystem competition or bias in intra-ecosystem competition. While the Commission’s assessments were validated by the Court of First Instance with regard to the effects of pre-installations and fragmentation clauses, they were rejected with regard to the foreclosure effects resulting from revenue-sharing agreements, attesting once again to the vulnerability of the use of the competitor test, which is so effective in the context of judicial review.

Our commentary is structured as follows. A first section focuses on the discussion of dominance. A second section discusses the restrictions on pre-installation of certain applications. A third section focuses on the effect of revenue sharing clauses. A fourth section deals with the analysis of anti-fragmentation clauses. A fifth section is devoted to the dimensions of the rights of defence and the setting of the quantum of the fine.

Dominance in the context of competition between ecosystems

It is therefore, first of all, on the question of competition between ecosystems that the judgment provides elements of analysis. At the time of the alleged facts, the competitive context of mobile operating systems was taking shape with, for the main ones,Apple’s iOS, BlackBerry OS, Android, Windows 10 Mobile, and other often "open source" operating systems, among which the "Android fork", which are based on theAndroid source code.

In this competitive landscape, Google criticized the Commission for having qualified the competition exerted by Apple and iOS as only an "indirect constraint", insufficient to challenge the dominant positions held by Google, whereas it should have taken into account the competition between ecosystems.

On this issue, the court takes care to recall the contours of competition between ecosystems. Without proposing an exhaustive definition, the judgment states that an ecosystem "brings together and causes several categories of suppliers, customers and consumers to interact within a platform; the products or services that form part of the relevant markets that make up this ecosystem may overlap or be connected to each other in view of their horizontal or vertical complementarity" (pt 116 of the judgment). Each relevant market is, in short, a component of a system of global dimension, which is itself subject to internal competitive constraints or constraints from other systems. The examination of the market power of an undertaking must therefore be carried out on several levels, and involves measuring not only the degree of dominance, but also its extent.

It is in this context of "distinct but interconnected" markets (pt. 129) that the Court then examines the objections made to the Commission concerning the delimitation of the relevant market.

After recalling the traditional methods of defining the relevant market, the judge concedes that their implementation sometimes requires a more detailed examination, going beyond this market segmentation alone, in order to better assess the competitive constraints prevailing on these markets and the economic position held by the company concerned (pt. 114). This is particularly the case for markets that are, as in this case, part of the digital economy (page 115). In fact, it is known that in this type of market, competitive parameters such as prices or market shares may be insufficient in the face of other variables, including innovation, access to data or user behavior.

This was therefore the first part of the plea. In the contested decision, the Commission considered that unlicensed OSes were not part of the same market as licensed OSes and that the competitive constraints arising fromApple ’s and BlackBerry ’s unlicensed OSes were insufficient. The Court reiterates this analysis, finding that the evidence relied on by Google does not show thatApple exercises a competitive constraint such as to prevent it from behaving to an appreciable extent independently of its competitors, customers and consumers.

To delimit the market for licensed OSes, the Commission took into consideration the fact, not contested by Google, that OFEs did not have access to unlicensed OSes, which includeApple’s iOS. In short, it differentiated between two distinct market strategies: closed vertical integration(Apple) and licensing(Google).

However, Google argued that Apple was constraining it in terms of progress and innovation: according to the company, the Commission had wrongly dismissed the investments made by Google to develop its Android system and the regularity of innovations. But the evidence provided by Google was not enough. Although the documents provided by Google did indeed show a "perception" of competition, they only attested to the existence of a competitive relationship between Google and Apple, without making it possible to assess its importance or to establish its significant nature with regard to the power held by Google on the licensed OS market. The arguments based on the constraint exerted by the race to innovate are also rejected. According to the Court, Google ’s investments in the development ofAndroid cannot, on their own, be explained by the importance of Apple ’s competition with Google (pt 149).

It was also the method of defining the relevant market that was questioned. In its decision, the Commission had considered that unlicensed OSes did not belong to the same market as licensed operating systems because of an indirect and insufficient constraint with regard to users and application developers. However, in the case of platforms, the question of the method of defining the relevant market may be questionable, in particular because the implementation of the hypothetical monopolist test may be called into question by the structuring of prices on several markets (or sides of a market). To overcome this difficulty, the Commission had implemented the quality-based test, the "Small but Significant and Non Transitory Decrease in Quality" (SSNDQ) test, suggested in particular by the OECD for so-called "zero price" markets. The test consists of examining the hypothesis of a small but significant and non-transitory deterioration inAndroid’s quality.

The Court of First Instance validated the Commission’s approach and, by the same token, the use of this test. It considers that, in the case of a product that could not easily give rise to the classic hypothetical monopolist test aimed at verifying the market’s response to a slight but significant and non-provisional increase in the price of a good (SSNIP), the SSNDQ test, which considers the deterioration in the quality of the product in question, was indeed a relevant indicator for defining the relevant market. Competition between firms can take place not only in terms of price, but also in terms of quality and innovation.

As for the information required to perform the test, the court is conciliatory. It considers that the definition of a precise quantitative standard of quality degradation of the targeted product cannot be a prerequisite for the implementation of the SSNDQ test. What matters is that the quality degradation remains slight, but significant and not temporary (pt. 180). Therefore, the Commission was right to consider the quality degradation of Android using the SSNDQ test.

With respect to Google’s dominance, the second part of the plea concerned Google’ s dominant position on the app store market. In this respect, the Commission had also defined this market, including all application stores for Google Android devices and other devices running on Android. This is reminiscent of the Microsoft case, except that there is no substitute for Google Play, with theApp Store being reserved foriPhone users to provide credible competition.

On this market, the Commission considered that Google held a dominant position with the Play Store. Several criteria were used: market share, the number and popularity of downloadable applications, accessibility of updates, the obligation to use the Play Store to benefit from Google Play services, the existence of barriers to entry, and the absence of countervailing buyer power for OEFs. The Commission also considered that the competitive constraint of the app stores for unlicensed mobile OS was insufficient.

This analysis has been challenged by Google, on two counts. First, Android and the Play Store are interdependent and simultaneously competitive: the dominance of one cannot be dissociated from that of the other. On the other hand, by dissociating the Play Store fromAndroid, the Commission would not have taken into account the competition from Apple. But this was without taking into account the fact that, unlike Android, iOS had only one application store and could not, for this reason alone, be separated from it. In this sense, the Play Store and theApp Store were both competing with each other through the system to which they belonged, Android and iOS (pt 246).

In the end, Google ’s argument suffered from a contradiction, since, in support of the first and second limbs of the first plea, the company challenged the definition in isolation, as well as its position on the markets for licensed operating systems and Android application stores, while at the same time requesting that competition between ecosystems be taken into account (pt 269). The Commission also recognized that bothiOS andApple ’sApp Store could exert a certain degree of constraint on Google. But this did not allow for a presumption ofApple’s competitive constraint. On the contrary, Apple is not a priori likely to influence Google’s dominant position in the national markets for general search services. During the infringement period, Apple benefited from a revenue-sharing agreement that was conditional on the default setting of Google Search on its mobile Internet browser, Safari. And, as a result of this agreement, Apple had no incentive to enter these markets to compete with Google Search (pt 272). For these reasons, the pleas are rejected.

Discussion on the abusive nature of pre-installations

The question of the effect of the pre-installation of certain applications is all the more interesting to deal with since it is part of a very contemporary issue relating todark patterns of choice. Numerous research articles and reports from competition authorities show that it is possible to manipulate individual choices and thus distort competition by playing on consumers’ cognitive biases. Among the biases that can be used for strategic purposes are the status quo biases that are central to this aspect of the Android decision.

In the present case, the Court of First Instance must rule on the anti-competitive effect of the mobile application distribution agreements [hereinafter ADAM] by which Google required equipment manufacturers [hereinafter "FEO"] to pre-install Google Search and Google Chrome, i.e., Google ’s search engine and internet browser, in order to be able to access Google Play, the application store, which is considered a must-have for consumers.

The competitive issue is the characterization of the foreclosure effects of a tying practice. While the foreclosure of a competitor is not necessarily anti-competitive, as long as it is based on the merits and results in the exit from the market (or marginalization) of a competitor that is less efficient than the dominant operator, it may become so if the latter uses its market position to drive out a competitor that is as efficient as it is. Its behavior can then be qualified as anti-competitive whether its effects occur on the dominated market or on a related market. In this respect, tying practices can be qualified as anti-competitive.

In this case, it was necessary to determine whether consumers still had the incentive to download competing search and browsing applications (although the ability to do so was not in question). The characterization of the foreclosure effect in this case must be not only "potential" but also "concrete", given that the practices in question took place from 2011 to 2018. The estimation of this capacity is based on an alternative past and not on an alternative future insofar as the usable counterfactual is that of a market in which the clauses in question would not have been imposed.

In its analysis of the Commission’s decision, the Court of First Instance must therefore successively consider the effects of the coupling between, first, Google Search and Google Play and, second, between Google Chrome and Google Play on the one hand and Google Search on the other, in order to assess the restriction of competition. For the Commission, the first coupling made it more difficult for competitors to enter the market and reduced the incentives for the development of competing search engines. The second coupling gave Google an advantage that its ’non-vertically integrated’ competitors could not replicate.

Google argues that the Commission has not provided evidence of the restrictive effect of its practices. For the applicant, the contractual clause only required that the relevant icons appear on the home screen without preventing the installation of competing applications or requiring less visibility. The requirement was only that competing search and navigation applications not be exclusively pre-installed on devices that had to access the Play Store. In addition, it was only a pre-installation and not a default or privileged installation.

The Court of First Instance followed the Commission in its assessment of the competitive advantage provided by pre-installation. It also confirms its position that the contractual clause reduces the incentives of equipment manufacturers to install in parallel to the applications pre-installed by Google. Furthermore, it confirms the Commission’s analysis that the foreclosure effect of pre-installation (the ADAM distribution agreements) was mutually reinforcing with the revenue sharing clauses (the "RPAs"). As a reminder, the RPAs to which we will return in our next section were revenue sharing clauses leading Google to pay a percentage of its advertising revenues to cell phone manufacturers that agree not to pre-install a search service competing with Google Search both on a device-by-device basis (device RPA) or on a product portfolio basis (portfolio RPA). The Court of First Instance confirms the Commission’s position that these clauses exclude competing browsers and search services from the incentives of the device manufacturers. Exclusive pre-installation of the latter would in fact deprive consumers of access to the Play Store.

Despite the Court’s acceptance of some of the arguments put forward by Google to show that the pre-installation of third-party applications was possible, it nevertheless validates the Commission’s position: the possibility for consumers to download competing applications and change the default settings or for manufacturers to pre-install other applications did not have a sufficient effect to counteract the foreclosure effects of the competitive situation. Furthermore, Google has not, in the Court’s view, provided sufficient evidence to explain the status quo on the basis of arguments relating to its qualitative superiority or on the basis of specific legal or economic circumstances.

For example, the Court did not take up the arguments developed by the applicant in its defense on the basis of efficiency. Pre-installation could indeed be justified within the logic of a two-sided business model through the capture of data flows allowing the implementation of a cross-subsidy mechanism for the benefit of users. Pre-installation would then be a way to amortize investments through the monetization of data flows. While the argument may be admissible in economic terms, in the European competition procedure, it is up to the respondent company to justify its practice. In other words, it must show that its behavior is objectively necessary and that the resulting foreclosure effects are more than offset by the resulting efficiencies, which must also benefit consumers. In the present case, the Court considers that the company has only provided vague and general elements which do not satisfy the burden of proof incumbent on the defendant.

Assessment of the unfairness of revenue sharing clauses

While in matters of efficiency defense, the burden of proof is on the respondent company, in matters of characterization of the theory of harm, it is on the authority responsible for implementing the competition rules. In this case, while the Commission had established in its July 2018 decision that the payment mechanisms conditioned on the pre-installation of Google ’s search service on a contractually predefined set of mobile devices was likely to give rise to anticompetitive foreclosure, the Court departs from its analysis and considers that it has not met the required standard.

The Commission considered that these portfolio RPA mechanisms (and not the device RPAs) involved exclusivity payments. It considered that these clauses covered a large segment of terminals marketed and that they had the capacity to foreclose competitors that were at least as efficient as the dominant operator. However, this had to be proven in the specific circumstances of the case. In accordance with the principles laid down by the Court of Justice in its judgment in Intel of September 6, 2017(Intel v. Commission, C-413/14P, EU: C: 2017: 632, pts 137 and 138), the Commission is required to analyze the market coverage of the practice in question, the conditions and modalities for triggering the payments, their duration, their amount and the existence of a strategy aimed at foreclosing competitors that are at least as efficient.

It is in the perspective of this demonstration that the Commission may have to carry out an "efficient competitor" test as defined in its February 2009 guidelines on its priorities for the application of Article 82 (currently 102) to exclusionary abuses (OJEC 2009, C45, p.7). This test aims in this case to ensure that a competitor as efficient as Google would have been able to match or outbid the payments made. It is therefore a question of ensuring that a competitor with the same costs as Google could have had an economic interest in winning the disputed share of search services covered by the portfolio RPA contracts. This test is by no means a prerequisite for demonstrating the possibility of such foreclosure, but if it is carried out by the Commission in the context of its decision, the Court of First Instance must, in its judicial review, verify whether it has been conducted in a sufficiently rigorous manner to satisfy the burden of proof that rests with the Commission.

While the Court recognizes that these mechanisms amount to the implementation of exclusivity payments and may be part of a foreclosure strategy, it departs from the Commission’s position in assessing their capacity to foreclose competition. First, it does not follow the Commission in its assessment of the share of demand covered by the exclusive dealing mechanism. The level of coverage of the RPAs by portfolio does not seem significant. This low rate of coverage is the first element that calls into question the Commission’s analysis, the second is due to errors in the methodology used to evaluate the costs of the competitor who is as efficient as it is. In this case, the Commission has overestimated the costs retained. It therefore reduces the margin of a competitor as efficient and thus misjudges its ability to replicate Google ’s incentive offers in terms of revenue sharing. It did not take into account the marginal costs of the challenged operator, but rather the costs taken from a document provided by a third party. It was therefore at fault in not seeking to corroborate these costs by means of an information request. In addition to this error concerning the cost benchmark, the Commission committed a second error by basing itself, in an unfounded manner, on the premise that an equally efficient competitor could only contest 12% of the request.

Thus, once again after the Intel and Qualcomm judgments, the performance of such an effective test leads to the annulment - here partial - of a Commission decision. This method based on a cost benchmark thus induces once again a real fragility in the application of competition rules to price-based predatory abuses (see in particular Caffara C., (2022), "The EU General Court confirms Android abuse of dominance through tying, with the real legacy of the case extending far beyond (Google Android)", e-Competition, n°108801). At least, this test demonstrates the existence of an effective judicial remedy capable of correcting material errors in the conduct of the tests and preventing the foreclosure of a competitor as efficient as the dominant operator, if a new entrant can be as efficient as the dominant operator, regardless of the nature of the market in question...

Confirmation of the crowding-out effect of anti-fragmenting clauses

However, the Court confirms the Commission’s assessment of the restrictive nature of the anti-fragmenting agreements (AAF), which prevented device manufacturers who wanted to access Google ’s applications from selling devices based onAndroid forks not approved by Google. Forks are new versions ofAndroid (available as open source) that can diverge as Google ’s version develops and potentially give rise to an alternative ecosystem.

While Google does not prohibit the use of these forks, it does subject contracts with OEMs to a process of technical certification of the forks (a compatibility test - hereafter STC). A fork that has not undergone this certification is considered incompatible. The arguments put forward by Google are part of an efficiency defense in that the induced restriction of competition is subject to an efficiency defense based in particular on the guarantee for the consumer and the other members of the ecosystem of reliability, security, interoperability and facilitation of developments. Moreover, the restrictive effects are considered speculative or erroneous, as Google considers that the now-defunct Amazon Fire OS and Alibaba Aliyun OS could not fall under such an anti-competitive foreclosure scenario.

The Court of First Instance, on the other hand, accepted the Commission’s arguments. On the one hand, it found that there was anticompetitive intent, considering that the contractual clauses were indeed intended to hinder the development of non-compatible forks. On the other hand, it considers that the foreclosure effect can be retained in view of the consequences for the two forks mentioned above of the unavailability of access to the Google Play Store. It also rejects the objective justifications put forward by Google for this restriction.

Rights of defence and amount of the fine

While the anti-competitive nature of these practices is confirmed, the methodology used by the Commission has been questioned on one point. By qualifying one of the targeted practices (that of having conditioned, in the context of certain RPAs, revenue sharing with OFEs and MNOs, to the exclusive pre-installation of Google Search on a predefined portfolio of devices), as a fourth abuse of a dominant position without having demonstrated its abusive nature in an autonomous manner, the Court of First Instance reformed Article 2 of the decision (pt 1029 et seq). This partial annulment does not, however, affect the overall validity of the infringement finding.

It is therefore by exercising its full jurisdiction, recognized in Article 31 of Regulation 1/2003, that the judge of the Union has reformed the Commission’s decision on this point. It took into account, in particular, the deliberate nature of the practices at issue and the value of the relevant sales made by Google during the last year of its full participation in the infringement, as well as the evolution over time of the various aspects of the conduct complained of. The fine is therefore reduced to EUR 4.125 billion.

Note Chronicles



Marie Cartapanis, Frédéric Marty, Relevant market: The General Court of the European Union upholds the Commission’s decision imposing a record fine on the dominant operator for strengthening its dominant position in the markets for general internet search services and intelligent mobile operating systems (Google / Android), 14 September 2022, Concurrences N° 4-2022, Art. N° 109569, pp. 71-76

Visites 504

All reviews