Is so much of a stir justified?
The General Court’s ruling in Google Search (Shopping)[1] is perhaps the most eagerly-awaited judgement in the last years in the field of abuse of a dominance under Article 102 of the Treaty on the Functioning of the European Union (the “TFEU”) and even in the domain of competition law. The reason is not only the size of the interests at stake but also that it was expected to re-define the limits of competition rules by establishing whether a dominant company’s refusal to share its own competitive advantage can only be regarded as abusive if the qualified standard of “essential facilities” is met. Traditionally, this doctrine, instituted by United State case law to regulate the use of irreplicable physical facilities (e.g. railways or ports) by operators competing with the facility owner in a related market (e.g. freight transport), determined that the latter’s refusal to allow access by the former was only abusive if such access could be seen as indispensable for rivals to compete in the related market and, consequently, refusal would lead to elimination of competition in this market. However, its applicability has declined especially in digital economy where it is difficult to predicate the indispensability of a platform given the multiplicity of alternative channels (competition is “one click away” as Google put it).
Back in March, the Court of Justice already clarified in its Telekom Slovak ruling[2] that the essential facilities doctrine is not applicable to cases where the dominant company’s conduct does not qualify as “competition on the merits” – i.e. conduct that is not explained by normal economic logic, but only by the dominant company’s artificial exploitation of its position of advantage (which the Spanish Supreme Court calls “objectively unlawful conduct”). In such case, the ordinary standard for exclusionary abuses applies instead, which is significantly less stringent than the essential facilities test in that it only requires proof of (i) conduct either not qualifying as competition on the merits (objectively unlawful exploitation of the dominant company’s artificial advantage) or specifically aimed at foreclosing competitors; and (ii) potential (rather than actual) foreclosure effect causally linked to the conduct (i.e. mere ability to foreclose).
Therefore, despite the uproar in the competition community, neither the Slovak Telekom nor the Google Search (Shopping) judgements introduce any novel legal question, but merely restate the raison d’être of the essential facilities doctrine: establishing the exceptional circumstances in which, where the dominant company competes on the merits (i.e. carries on no “wrongful”, objectively unlawful, abnormal, artificial or uneconomic conduct), an administrative intervention interfering with its freedom to conduct business so intense as to force it to share its advantage with rivals is still justified by the legal interest in competition. This very same postulate was already affirmed in the seminal ruling on essential facilities, the Magill judgement.[3]
What is truly new about the Google Search (Shopping) ruling is that it takes a decisive philosophical step in establishing when the dominant company’s conduct is to be considered wrongful or outside competition on the merits – a question that lies more in the realms of economics or ethics (Latin fas and nefas) than in the legal order. Indeed, in the Slovak Telekom ruling, the reproachable nature of the dominant company’s conduct was quite clear – i.e. preventing competitors’ access to the local loop through clearly artificial mechanisms in the sense of lacking an explanation in normal economic logic (such as margin squeeze). Contrariwise, in the Magill judgment, it was evident that there was nothing artificial or uneconomic in the television channels’ refusal to share their intellectual property rights with the publisher of a rival programme guide. However, the conduct that the European Commission found fault with in the Google Search (Shopping) decision[4] lies in a rather grey zone, in that the allegedly dominant tech giant merely designed its online general search services in such a way that results from Google’s own specialised search services were displayed more visibly than those from its rivals’ specialised search services. Does this make economic sense? Is it wrongful?
Factual background – this part can be skipped if the reader is familiar with the facts
Google is active in the market for online general search services, consisting of displaying on Google’s general results pages, in response to internet user queries, (i) general or generic results – selected by Google’s search engine through internet crawling and indexed according to relevance criteria; and (ii) specialised results – selected from Google’s specialised search database which is fed by merchants interested in advertising their products and displayed along with generic results on the general results page. Alongside generic and specialised results (both called “natural” results in that their display is independent of payment), Google’s general results page also displays sponsored results or ads, the indexing of which depends on payment by advertisers (to which their webpages ads are linked) through an auction mechanism based on pay per click.
The market for specialised search services is therefore related or adjacent to the market for online general search and may refer to news, information on services (e.g. as transports or restaurants) or product and price comparison – in this latter specialised search market Google competes with third-party comparators, such as Bestlist, Nextag, IdealPrice, Twenga, Kelkoo or Prix.net. In this regard, the Commission’s Google Search (Shopping) decision relies on (i) national markets for online general internet search – where Google would have been dominant since 2008 given its high market shares consistently above 70 % compared to its competitors Bing and Yahoo in all Member States; and (ii) national markets for comparison shopping services – excluding merchant platforms like Amazon and direct online sales channels.
Google displayed on its general results page the specialised results of its own comparison shopping services in a particularly visible way using different formats since its inception in 2002. Firstly, Google comparison shopping services were a separate page (“Froogle”) from the general results page and listed results from a merchant-fed database. From 2005, Google’s specialised results were also included in a box (“Product OneBox”, later called “Product Universals”) immediately below and left to ads (sponsored results) and above generic results and, over time, both natural specialised results from Google’s comparator grouped in “Product Universals” and sponsored results were enriched with product images and links to both merchants’ pages and related specialised searches. In 2013, Google removed “Product Universals” and maintained on the general results page (along with generic results) only sponsored results, grouped in rich-format boxes (“Shopping Units” at the top) or as normal links labelled as “ads” or “sponsored” (immediately below). Both types of sponsored results were linked only to the advertisers’ pages, so that, if users wanted to access further related ads in Google’s specialised search service, they had to click on the “Shopping” tab (“Google Shopping”), which equally no longer listed natural specialised results but only sponsored results that did not make it into the “Shopping Units” according to the combined relevance and auction criteria.
The Commission’s efforts in the decision concentrate on demonstrating how the way in which Google ranks the results of rival shopping comparators as compared to the way it ranks its own comparison results is not competition on the merits – in order to escape the essential facilities test (which would be applicable if the advantage derived by Google from this conduct qualified as competition on the merits). To this end, the Commission criticises the combination of (i) the display of results from rival comparators on Goggle’s general result page always as regular generic results, which by definition do not have access to the most visible boxes with images and rich format; (ii) the application of an algorithm that made results from rival comparators systematically more prone to demotion among generic results due to inherent characteristics of those results (e.g. lack of original content); (iii) the combined fact that the demotion algorithm did not affect Google’s comparison results and that only these had access to the most visible and richly formatted boxes with images and dynamic information. The Commission clarifies that it does not object to the more favourable display of Google’s comparison results or to its allegedly relevance-based ranking criteria, but to the fact that it systematically excluded results from rival comparators – thus linking the reproach to discrimination.
Having established the wrongfulness of Google’s conduct, the Commission analyses the second part of the general exclusionary abuse test (i.e. the ability to foreclose) by (i) showing the essentiality of traffic as a competitive parameter for comparators; (ii) proving that Google’s conduct resulted in an actual increase of traffic from Google’s general results page to its comparator and a concomitant actual decrease of this traffic to rival comparators, which they could not replace by other traffic sources such as AdWords, mobile apps or direct traffic (alternatives that would, however, militate against the indispensability of traffic from Google’s general results page and make the essential facilities doctrine difficult to apply); and (iii) inferring the potential anti-competitive effect in the form of potential rival comparators’ disappearance (to the detriment of innovation and consumer choice), the extension of Google’s dominance in general search to shopping comparison and the feedback effect of strengthening dominance in general search because of the enhanced resources obtained in shopping comparison. Google’s claims of objective justification and efficiencies were rejected.
Summary of the General Court’s analysis – no new thing under the Sun
The General Court begins by analysing Google’s plea that the Commission failed to identify which elements of its conduct, which Google describes as improving the quality of its online search services, departed from competition on the merits, and hence that watchdog would therefore have had to rely on the essential facilities doctrine in order to mandate Google to share its advantage. Specifically, Google argued that improvements in search page design are a key parameter for competition in online markets and should therefore be considered “normal” or on the merits (in an economic sense); consequently, any foreclosure effect that they may cause is irrelevant. Naturally, the General Court accepts this premise, since it is for the Commission, as a necessary condition for finding an exclusionary abuse, to identify how the dominant company used its dominant position to deviate from the methods that constitute normal competition based on the merits.
Significantly, the General Court clarifies that the practice named “leveraging” is defined as that identified in one market which has an impact on another market and may take various forms (tying, margin squeeze, loyalty rebates or refusal to provide interoperability information). Leveraging can also be lawful depending on all specific circumstances, which need to be looked into to deem it abusive. In the case at stake, the Commission found Google’s favouring its own business in the related market for comparison shopping services to be abusive based on the following factual findings:
(i) the importance of traffic from general results pages for comparators – allowing for increased attractiveness of the service for advertisers that generates a network effect virtuous circle: revenues from ads, information on user behaviour improving the relevance of results and the possibility to experiment and recommend other products for users feeding back into enhanced attractiveness;
(ii) user search behaviour – providing evidence that a high percentage of users only see the first three to five results irrespective of their relevance; and
(iii) the actual diversion of traffic from rival comparators to Google’s comparator not replaced by traffic from other sources.
The General Court then feints to impose on Google a qualified special responsibility (as opposed to the regular special responsibility on any dominant company) as a super-dominant or “ultra-dominant” (a new-fangled term) company controlling a “gateway” (note how the term “gatekeeper”, which is alien to abuse case law, is conveniently avoided). This attempted qualified special responsibility seems to involve some sort of reversal of the burden of proof whereby super-dominant companies would be obliged to justify any difference in treatment against their competitors. This obiter dictum is a nod to the popular debate on the guarantee of a “level playing field” on those digital platforms that fly under the radar of essential facilities (including references to the operation of network effects as barriers to entry and even to the equal treatment obligation introduced by Regulation (EU) 2015/2120 of the European Parliament and of the Council of 25 November 2015 laying down measures in relation to access to an open internet). However, the General Court does not venture to establish a legal rule and ends up linking the level playing field to competition on the merits by stating that differences in treatment are “abnormal” (in the sense of artificial, uneconomic or wrongful) when implemented by platforms with a “universal vocation” and, above all, where such discrimination did not occur at the stage of dominance of the primary market (in casu general search) and starts to be used at the stage of expansion to the related market (in casu specialised search).
Then, the General Court goes on to question Google’s reasoning that the practical result of the Commission’s decision is an obligation to deal (to allow rival comparators’ access to Google’s “technologies and designs”), which in Google’s view is only possible if the essential facilities requirements are met – in particular the indispensability of access to traffic from Google’s general results page to compete in the market for shopping comparison services. In that regard, the General Court acknowledges that the subject of the dispute is indeed rival comparators’ access to Google’s general search results pages, which would have similar characteristics to those of an essential facility in that the traffic from such facility is not “effectively substitutable”. However, in order to avoid the application of the essential facilities doctrine the General Court resorts to a reasoning that is questionable as excessively formalistic: since there is no explicit refusal to deal at the causal origin of foreclosure, but an instance of discrimination, the essential facilities doctrine is not applicable.
It comes across as surprising that the General Court uses this clunky subterfuge when the Slovak Telekom judgment would have allowed it to avoid the essential facilities doctrine simply by insisting on the conclusion already drawn that Google’s conduct was not competition on the merits (be it an explicit refusal to supply, some discriminatory conduct or any other practice that may be considered wrongful or abnormal in the sense of drawing an artificial advantage from a dominant position). A different matter is that a refusal to supply may sooner look like an irreproachable business decision (e.g. if there has been no prior dealing or legal obligation to deal) while discrimination can more intuitively labelled as wrongful. In fact, interestingly, the General Court itself seems to return to this logic by stating that the applicable legal test does not depend on the remedy – i.e. not because halting a reproachable conduct entails an access obligation does the essential facilities doctrine apply, something which, on the other hand, was already settled in the Slovak Telekom judgment.
The next plea that the General Court examines concerns Google’s contention that, by grouping its own comparison results most visibly on general results pages, it intended to improve the quality of its general search service rather than to divert traffic to its offer comparison service – as its sticking to relevance criteria to list results would show. In response, the General Court recalled that intention is not a necessary element for proving or excluding the existence of abuse (this being an objective concept). Hence, the Commission was right to apply the general exclusionary abuse test by establishing the elements that made Google’s combined conduct (more visible display of its comparison results and relegation of its rivals’) objectively outside competition on the merits and the resulting potential forclosure effect, while any possible pro-competitive effect was to be analysed at the stage of efficiencies and justification.
On the other hand, Google argues that neither the differentiated treatment of its natural specialised results by displaying them in “Product Universals” nor, once these were removed in 2013, that of sponsored results in “Shopping Units” should be considered discriminatory given that they are in a different situation from the (necessarily generic) results coming from rival comparators. Concerning its own specialised results as opposed to those from rival comparators, Google argues that the latter are always generic results (as Google can only access them by crawling the web), whereas for the former it had direct access to the merchant-fed database. The General Court replies that what comes in for criticism is not the application of different relevance criteria to the one and the other, but the combination of results from rival comparators (i) not being eligible (as generic results) for access to the most visible display, reserved for Google’s specialised and sponsored results, and (ii) being systematically relegated among generic results by an algorithm that penalised their inherent characteristics, such as the lack of original content.
The same applies to sponsored results (the only ones with access to the “Shopping Units” after Google’s decision to remove specialised results in 2013). This is not contradicted by the positioning of “Shopping Units” either depending on relevance (which meant that they could appear below generic results) or being linked to payments from advertisers (which was found irrelevant since both ads and comparison shopping services are provided to internet users for free and are therefore substitutable from the point of view of demand). Rather, what the General Court regarded as relevant is the fact that a rival comparator could never access the “Shopping Units” even in exchange for payment, unless it changed its business model (either to merchant including a shopping functionality or to ads intermediary) and thus ceased to be a rival comparator.
Besides, Google also argued that “Shopping Units” do not generate any profit for Google’s comparison services identified with Google Shopping (beyond pay-per-click revenues from result sponsorship by advertisers which is unrelated Google Shopping). As a response, the General Court contends that the “comparison shopping service” which Google favours comprises both the “Shopping Units” on the general results page and the separate Google Shopping page (both containing since 2013 only sponsored results paid by advertisers, who, by the way, when paying, do not know whether those will be displayed in “Shopping Units” or on Google Shopping), as well as discontinued “Froogle” and “Product Universals”.
As regards the second part of the exclusionary abuse test (i.e. the ability to foreclose), Google also denied that anti-competitive effects resulted from its conduct by questioning that the decline in traffic from its general result pages to competing comparators was causally linked to its conduct in the absence of a valid counterfactual analysis by the Commission. Google, on the other hand, provided a counterfactual where traffic evolves in a similar way in countries where “Product Universals” and “Shopping Units” were introduced and in countries where they were not, and does not change either in the scenario where those products are removed. Nevertheless, the General Court makes clear that the Commission did not criticise either the demotion of rival comparators’ results among generic results or the more visible display of Google’s comparison results, but the combined effect of the two practices. In this light, the counterfactual proposed by Google in which only one of the two limbs of the combined conduct (more visible display) is factored in cannot be valid.
For this reason, Google cannot benefit from the jurisprudence on the burden of proof enshrined in the Intel judgment[5], according to which competition authorities must rebut the arguments that investigated companies in sanctioning proceedings raise to contradict the analyses presented against them. This holding is interpreted narrowly by the General Court as requiring the investigated companies to provide a full set of arguments rebutting all the points raised by competition authorities to demonstrate causation between conduct and potential effect. Similarly, in relation to Google’s criticism of the increase in traffic to its comparator shown by the Commission, the General Court considers that Google should have proved that such increase would not have benefited rival comparators in the absence of the conduct complained of. As concerns specifically the criticism of the causal link to the conduct, Google should have rebutted the correlations that the Commission established between positioning and clicks instead of defending the improved relevance of its comparator’s results.
In addition, Google suggested that the anti-competitive effect claimed by the Commission was speculative, as it should have materialised given the lengthy period of the conduct complained of. To this the General Court responds again by nuancing the Intel judgment (which picked up on the so-called “more economic approach” by placing on competition authorities the burden of examining all relevant circumstances if the defendant provides evidence that the conduct complained of is not restrictive). Indeed, the General Court clarifies that this case law cannot go so far as to require the demonstration of an actual effect. Therefore, if was enough for the Commission to prove the reduction of traffic to rival comparators, the consequent reduction of their market share and the conclusion that alternative sources of traffic were not substitutable, and to infer that those competitors would end up being foreclosed as a result even if they had not exited the market yet. However, the General Court did consider speculative the Commission’s assertion that the anti-competitive effects would have fed back into the market for general search services through the exceptional revenues secured in comparison shopping services due to the conduct, and annulled this point of the decision.
Still as part of the assessment of effects, the General Court examined Google’s objection to the exclusion of merchant platforms from the relevant market for comparison shopping services. Google argued that competitive pressure from merchant platforms would have defeated anti-competitive effects on the market where these were supposedly felt (i.e. comparison shopping services), had this market been defined correctly as including those platforms. The General Court considers that the Commission’s analysis was complete and, in particular, took into account both sides of the merchant platforms (i.e. both users and sellers) and Google failed to rebut the differences (supported by Google’s own internal documents), especially in terms of substitutability on the user side arising from the purchase functionality which justified the definition of separate markets in the decision.
The General Court also ruled out the possibility that the anti-competitive effects were gainsaid by the competitive pressure which, in Google’s view, the merchant platforms would exert even if they were seen to belong to a separate market. It found that the Commission would not have been obliged to expressly take this into account in its market share analysis, especially after having excluded merchant platforms from the market. Finally, Google’s arguments concerning the Commission’s obligation to analyse traffic from other sources and the existence of barriers to entry or to use the equally efficient competitor test (which would be reserved in the General Court’s view for price abuses) were rejected.
Regarding objective justifications and efficiencies attempted by Google, the General Court upheld the Commission’s rejection. Those included (i) the pro-competitive effects of both demotion algorithms (as increasing the relevance of generic results) and the deployment of “Product Universals” and “Shopping Units” (as increasing the quality of comparison) – which would not have been sufficiently demonstrated in view of the Commission’s argument that the exclusion of rival comparators would have led precisely to higher prices for retailers and thus for users and, ultimately, less choice and innovation; (ii) the need to treat rival comparators differently in order to meet users’ expectations (which were not found to be such), and to monetise the space on its general results page in case (which was done by other means anyway); and (iii) the technical impossibility of indexing rivals’ comparison results alongside those of Google’s own comparator in a consistent way because it did not have access to the databases that fed the former – which was found hard to believe given that results from rival comparators are generic results.
As regards the fine, the General Court dismisses Google’s application for annulment based on the alleged novelty of the infringement and the negotiations of commitments which were ultimately abandoned by the Commission. In so doing, the General Court recalls that the Commission is not precluded from sanctioning novel practices (that fact being relevant only for the purposes of intention or negligence), and has discretion as to whether or not to adopt a commitments decision. Likewise, the General Court, in the exercise of its unlimited jurisdiction for fine calculation, validated the amount of the fine on the grounds that the gravity of the conduct (based on the economic importance of the market concerned, the duration, Google’s market shares and intentionality) justified the addition of the additional 10 % imposed in the decision for deterrence.
Conclusion – less (law) is more (discretion)
The Google Search (Shopping) judgment does not introduce a new legal test, but avails of a borderline case to extend the scope of business decisions that can be considered outside competition on the merit so that (if they also have a potential exclusionary effect) they can be considered abusive without the need to meet the essential facilities standard (reserved for irreproachable business decisions). Meanwhile, Google’s efforts to defend the applicability of the essential facilities doctrine based on the intense remedy that finding fault with its conduct leads to (allowing rivals to access its advantage on equal terms) are barren after the Slovak Telekom ruling. Indeed, this judgement clarified that the absence of competition on the merits is what matters and, once this is established, the remedy is to start competing on the merits, which can certainly entail an obligation to deal with competitors (especially if the dominant firm was already dealing with such competitors before or had a legal obligation to contract).
Now, and this is the one real breakthrough, the General Court leaves discrimination in favour of the dominant company’s own business or “self-preferencing” (inherent in the right to property) outside competition on the merits. However, it still requires that the favouring of the dominant operator’s own business in a market related to the dominated one feature an element of economic abnormality or wrongfulness – e.g. Google’s using mechanisms contrary to the normal economic logic of generic results relevance (which is assumed to drive general search services) to divert traffic to its comparison shopping services. Therefore, the logic does not seem to deviate from that of the Slovak Telekom judgement: freedom to conduct business is not affected because there is a prior decision to deal (in casu to derive traffic to rivals by indexing generic results following a relevance criterion) but then such dealing is artificially carried out on worse terms for rivals (in casu by deviating from the relevance criterion underlying general search only when this favours its business in a separate market).
In other words, the General Court takes up the gauntlet thrown by the Court of Justice in the Slovak Telekom judgment and opts for the simple and obvious, one might say minimalist, solution and (to some extent forced by Google’s pleas formulation) keeps the debate at the level of the wrongfulness of Google’s conduct, thereby avoiding the essential facilities test. However, it cannot help hinting at a principle of equal treatment on super-dominant platforms (a hot debate in the digital field where it is difficult to speak of essential facilities due to the lack of indispensability) based on some sort of reversal of the burden of proof where the dominant company would be obliged to justify any differences in treatment that it implemented. In any case, the General Court refuses to consecrate a legal rule in this regard, but rather makes discrimination a mere factual element for assessing the deviation from competition on the merits – a full-fledged level playing field obligation would have been difficult to found on case law (perhaps in the cases on Article 106 TFEU such as the DEI saga) in order to withstand cassation, although it could have reduced the wide discretion that competition authorities are left with as judges of the wrongfulness of dominant companies’ business models.
The question remains whether conduct outside competition on the merits could be extended even further to instances where the dominant firm has never dealt with competitors and has no legal obligation to do so (think of Apple’s policy of not licensing its operating system) on the grounds that it is artificial or economically abnormal and has an exclusionary effect on rivals even if it is not indispensable for them to compete. Apart from the legal certainty problem that this raises, the freedom to conduct a business could be compromised if the need for public intervention to protect competition is not pondered over against it as is the case under the essential facilities doctrine (henceforth restricted to no-dealing policies that are irreproachable in light of e.g. their consistency over time and equal application to all rivals). The solution could be the Digital Markets Act legislative initiative[6], if eventually passed, as it would undoubtedly provide a more appropriate framework for preventive and dialogue-driven analysis of the effects that super-dominant digital platforms may have on competition and proportionality of the remedies to counter these.
[1] See judgment of the General Court of 10 November 2021 in Case T-612/17 Google LLC and Alphabet, Inc. v European Commission, ECLI:EU:T:2021: 763.
[2] See judgment of the Court of Justice of 25 March 2021 in Case C 165/19 P Slovak Telekom, a.s. v. European Commission, ECLI:EU:C:2021:239.
[3] See judgment of the Court of Justice of 6 April 1995 in Joined Cases C-241/91 P and C-242/91 P RTE and ITP v Commission of the European Communities, ECLI:EU:C:1995:98
[4] See decision of the European Commission of 27 June 2017 in Case AT.39740 Google Search (Shopping).
[5] See judgment of the Court of Justice of 6 September 2017 in Case C-413/14 P Intel v Commission, ECLI:EU:C:2017:632.
[6] See proposal for a regulation of the European Parliament and of the Council on contestable and fair markets in the digital sector (COM/2020/842 final).
