The long-awaited Google Shopping judgment is out (see the press release here and the full decision here). The General Court has dismissed Google’s action almost in its entirety, upholding the fine of € 2.42 billion that the EU Commission issued in 2017 on the company for abusing its dominance as a search engine by favouring its comparison shopping service. This post will give a brief overview of the Court’s main arguments and discuss some of the most important aspects of the decision.
A brief reminder of the facts
According to the Commission, Google systematically gave prominent placement to the Google Shopping Service and demoted rival comparison shopping services in the Google Search (generic search) results. While results from other comparison shopping services could only be found in the generic results (the typical blue links), results from Google Shopping would show up at the very top of the general search results page, displayed separately and in a particularly rich format (e.g., including pictures) in the “Product Universal Box” or the “Google Shopping Unit” (which contained paid advertising) and were not subject to the general relevance-based ranking of the generic search results. At the same time, competing comparison shopping services were demoted in the search results due to a change to Google’s algorithm (the ‘Panda’ algorithm). Google’s own offer was hence much more visible to consumers, who tend to only click on the first couple of search results. Google thus “exploited” consumers’ reliance on defaults (here presented by the top search results) with the effect that consumers stuck with Google Shopping rather than looking out for a competing comparison-shopping website. This would then lead to a vicious circle: Google Shopping would generate more and more traffic, putting itself in a better position to convince merchants to provide information about their products, generate more revenue and collect more information on users. This would then allow Google to show even more relevant results to users and thus attract more and more of them, eventually forcing competitors to exit the market.
The Commission, after a seven-year-long investigation, fined Google for a violation of the EU antitrust rules based on a theory of self-preferencing (without relying on the essential facilities doctrine) as abusive conduct. In the aftermath and already during the investigation, this sparked a debate on whether or not self-preferencing can fall under Art. 102 TFEU. Opponents argued that a prohibition of self-preferencing contradicts the strict requirements for access to a dominant undertaking’s infrastructure that have been developed under the essential facilities doctrine and that, ultimately, consumers would benefit from Google’s allegedly anticompetitive practices.
What was disputed?
First of all, what was not disputed: Google’s dominant position. Admittedly with market shares of around 90 % in most relevant geographic markets, it is hard to think of any argument that Google could have come up with.
What was disputed? Everything else. Google’s main argument was that its conduct was, in fact, a quality improvement in its online search service and, therefore, competition on the merits (or at least justified) and that the Commission had imposed on it a duty to supply in violation of the Bronner case law. The counter-argument from the Commission was that even if Google’s argument that the conduct in question had improved its service was true, this would not preclude an abuse of dominance. Besides, the improvement as such was disputed: how could there be an improvement in the service if Google users were encouraged to click on results other than the most relevant for their search query?
Why was the essential facilities doctrine not applicable?
Google compared itself to Mediaprint, a newspaper publisher that had refused to deliver competing journals through its distribution network, which gave rise to the Bronner judgment. There, the CJEU established the so-called essential facilities doctrine under which even dominant undertakings have to grant access to their facilities only under strict criteria. As it relied on a different theory of harm, the Commission did not establish that these conditions were fulfilled.
The Court clarified that what the Commission accused Google of was a refusal to grant (equal) access (para. 222) – something the Commission had been shy to admit. The Court emphasized that Google’s general results page “has characteristics akin to those of an essential facility” as there is no substitute available and competing shopping services are generally dependent on traffic from Google (para. 224). Nevertheless, the Court distinguished Google from Bronner. Unlike traditional infrastructures, whose value lies in the owner’s ability to exclude others, Google’s search engine’s value lies in “its capacity to be open to results from external (third-party) sources” (para. 178). The promotion of one type of specialized result “involves a certain form of abnormality” (para. 176). Google’s preferential treatment towards its shopping service in the general search result was not competition on the merits because it was “not consistent with the intended purpose of a general search service” (para. 184). Moreover, just because the remedy against Google’s conduct can be similar to the one applied in essential facilities cases, this does not mean that the abuse has to be of the same nature (para. 246).
Did Google just improve the quality of its general search results?
Dismissing Google’s argument of quality improvement, the Court emphasized that product or service improvements as such do not exclude that a conduct has anticompetitive effects. Such considerations are reserved for the question of whether the conduct is justified (para. 188). The Court also held that while the Commission can assess the dominant undertaking’s intent in evaluating whether a certain business strategy is anticompetitive, the existence of anticompetitive intent constitutes only one of several factors which may be taken into account. Here the Court is very clear: “the existence of an intention to compete on the merits, even if it was established, could not prove the absence of abuse” (para. 257).
Comparing apples with pears – discrimination or not?
Closely related to the argument of product improvement, Google claimed that the different treatment of its own shopping comparison results with competing shopping websites was not discriminatory. While it did not dispute that it treated them differently, it argued that it did so for legitimate reasons (para. 272): generic results were based on “crawled” data and on the relevance derived from this data, whereas product results were based on data feeds directly provided by the merchants and on product-specific relevance signals. Google thus applied different technologies to different situations with the legitimate goal of improving the quality of its results.
Here, too, Google did not convince the Court, which emphasized that the discrimination did not lie in a different treatment based on the nature of the results, i.e., product-related or general, but on the differential treatment between the origin of the results: those coming from Google were treated better than those coming from competitors (para. 284). As the Product Universal Box was not open to competitors, their results – even if they happened to be more relevant for a search query – could never appear in there. Effectively, the same was true for the Google Shopping Unit. While Google argued that the Shopping Unit was open for competing shopping search engines, this was only possible when they directly offered users to purchase products – thus acting either as seller or intermediary. In essence, this would have required them to change their business model from a comparison website to a retailer or intermediary platform, such as Amazon. Such a change would have required competing shopping comparison websites to become Google’s customers rather than its competitors (para. 351).
Google moreover argued that generic search results were not comparable to the product-specific paid results that were shown in the Shopping Unit as they were marked as “sponsored” and easily distinguishable for the user. Like the Product Universal Box, however, the Shopping Unit could never be demoted and thus, according to the Court, did not face the same treatment as competing comparison shopping services whose results could never appear in the Shopping Unit (para. 310). Although containing paid results, the Shopping Unit would only show results from Google’s comparison shopping service paid for by the seller. The label “sponsored” did not ensure that users understand that the results are ranked according to different mechanisms than the general search results (para. 313).
The anticompetitive effects
The Commission argued that Google’s self-preferencing had led to a reduction in traffic from Google (general search) to competing comparison shopping services. Google here disputed the causal link. The Court held that the Commission for demonstrating an infringement of Art. 102 TFEU does not necessarily have to provide a counterfactual analysis of how the market would have developed absent the conduct in question, particularly since it is sufficient to establish potential effects. The Commission had sufficiently shown such potential effects by showing a correlation between the practices and the reduction in traffic, to shift the burden to cast doubt on this evidence on Google (para. 382).
Google failed to provide a convincing counterfactual. It argued that the competing comparison websites’ reduction in traffic was attributable solely to the change of the Google general search algorithm’s ranking parameters. The Court did not accept this as evidence of lack of causality since it was not the complete counterfactual. Since the conduct in question consisted of two elements, Google could not just leave one of those two elements out of the counterfactual (para. 390). Google also argued that the traffic reduction was linked to broader industry developments and shifting user preferences towards merchant platforms instead of comparison shopping searches. Here, the Court saw a classical chicken-egg problem: did users shift their preferences toward merchant platforms just because they were ranked better on Google than comparison websites or the other way around (para. 391 f.)? This was hence not enough to cast doubt on the Commission’s evidence.
Google also tried to argue that Google Shopping did not benefit from the practices in question (para. 396 ff.). This was quite an uphill battle: while Google’s comparison shopping service had initially been largely unsuccessful, from the moment it started implementing the practices in question, the situation dramatically changed with traffic multiplying in a short time. Google, not disputing the increase in traffic as such (although there was some debate about the correct amount), argued that this was due to a greater relevance of the results within Product Universals and Shopping Units for users rather than because of their better positioning. In the Court’s view, however, this could not have explained why traffic only increased after Google started to position its comparison shopping results better and to demote competitors in the search results: if Google’s shopping results were so much more relevant than those of competitors, Google would not have had to engage in its anticompetitive practices.
Google also argued, more generally, that the Commission had not established anticompetitive effects, leading to higher prices and less innovation (para. 421 ff.). The Court here reminded Google of the burden of proof: while the Commission has to demonstrate at least potential anticompetitive effects, taking into account all the relevant circumstances, it did not have to identify actual exclusionary effects (para. 436) – an abuse remains an abuse even it was unsuccessful.
While the Commission had found potential anticompetitive effects in the markets for specialized comparison shopping search services and in the markets for general search services, the Court followed the Commission only with regard to the former. For specialized search, the Commission had shown quite a detailed analysis on how traffic and market shares of competing specialized search websites declined after Google implemented its practices, potentially reducing their (and also Google’s) incentive to innovate and reducing the ability of consumers to access the best-performing comparison shopping services (para. 451).
In the market for general search, however, the Commission had only argued that Google, through its anticompetitive practices, was protecting the revenue generated by specialized search, which it, in turn, could use to finance the general search service and thus reinforce its dominant position in this market. This essentially comes down to the allegation that Google was earning money with its practices. This was not enough for the Court, which annulled the decision in so far as the Commission found an abuse on the basis of anticompetitive effects in the markets for general search (para. 456 ff.) – Google’s only small victory in the judgment.
Who is more efficient?
EU competition law does not seek to protect inefficient competitors. The CCIA (Computer & Communications Industry Associations of which Google but also Amazon, Facebook and many other tech companies are members), an intervener on Google’s side, argued that this was exactly what the Commission sought to do. It argued that the competing comparison shopping websites were “not particularly innovative and had not taken appropriate measures to generate traffic from sources other than Google” (para. 514). The Court rejected this, holding that the as-efficient-competitor test is warranted only in the case of pricing practices (e.g., predatory pricing or margin squeeze) and was thus irrelevant here (para. 538). It recalled the purpose of this test: to show that even an as efficient competitor as the undertaking establishing the pricing practice would not be able to withstand it by adopting the same pricing. Hence, the test does not aim to assess actual market participants’ efficiency but constitutes a theoretical exercise not warranted in cases that do not involve pricing issues.
As to the objective justification, Google once again played the card of the improvement of its services’ quality. This was essentially argued on every single point of critic on the Commission’s assessment. The Court rejected it every single time. So it did for the objective justification. The Court could not see any reason why any improvements required the self-preferencing of results from Google shopping over those from other sources (para. 567). The demotion of competing comparison shopping services as service improvement was rejected as such because Google did not provide sufficient evidence on the positive effects of this conduct. The Court concluded that even assuming that it was technically impossible to avoid the unequal treatment in the context of Product Universal or the Shopping Unit, as there are no (sufficient) efficiency gains, “it is irrelevant that what was done to achieve them could not be implemented technically otherwise” (para. 572). Moreover, while Google argued that it could not have applied similar processes and methods to rank results from competing comparison shopping sites, it did not show that it was prevented from applying different methods to those that led to similar results (para. 576).
The 2.42 billion Euro question
The fine of € 2.42 billion that Google was ordered to pay compared to the subsequent fine in the Google Android case (€ 4.34 billion) might seem almost mild. Furthermore, although Google is likely able to pay it out of its cash reserves, the company was not very happy with it. Google and the CCIA argued that the “stratospheric” fine was unjustified because the Commission, for the first time, had “classified conduct aimed at improving quality as abusive” and developed a novel theory of harm (para. 598). Moreover, it saw the fact that the Commission initially had undertaken to deal with the case under the commitments procedure as an indicator that a fine was not appropriate (para. 600).
The Court once more did not follow Google’s line of arguments. First, it held that nothing in the law prevents the Commission from penalizing conduct contrary to EU competition rules that is found for the first time (para. 628). As long as the infringement was committed intentionally or negligently, a fine may be imposed. Second, turning to the Commission’s initial treatment of the case under the commitment procedure, it held that the Commission has a margin of discretion in the choice between the commitment and the infringement procedure and may revert to the latter procedure after having embarked on a commitment procedure and issue a fine without breaching the principle of legitimate expectations (para. 635 ff.). Google also claimed to be treated unequally to other cases where the Commission had refrained from imposing fines and thus violated the principle of equal treatment. This the Court rejected with the uniqueness of the Google case – turning against Google its own argument that the case was a “first” (para. 624).
As regards the amount of the fine, Google, questioning the Commission’s calculation, had argued that the fine should in any case not have been higher than EUR 91 million (para. 650). The Commission had calculated the fine based only on the value of advertising revenue related to the markets for specialized search services, so the partial annulment of the decision could remain without consequence for the fine. The Court accepted the Commission’s fine calculation, considering that Google’s conduct constituted a particularly serious infringement and the intended deterrence effect. Exceptionally large undertakings can require exceptionally large penalties.
In my view, the two most important aspects of the judgments are 1. how the Court dealt with the essential facilities doctrine and 2. how it rejected Google’s argument that the company was merely seeking to improve its services.
- While the Court has emphasized to quite some extent the importance of Google’s general search infrastructure (even stating that it has “characteristics akin to those of an essential facility”, para. 224), in my opinion, the judgment’s key point is why the essential facilities doctrine was not applicable. Unlike traditional infrastructures, whose value lies in the owner’s ability to exclude others, Google’s search engine’s value lies in “its capacity to be open to results from external (third-party) sources” (para. 178). If it is Google’s business model to show results from other parties and therefore use them in its infrastructure, it must be treated differently than companies whose business model it is to invest in their own infrastructure to use it exclusively. Google cannot have it both – be a neutral search engine using results from diverse sources and then discriminate between them. Colomo, in one of the first contributions to the debate on the judgment (a very recommended read!), has raised the point that self-preferencing is widespread also in other industries with more competitive market structures, such as supermarkets. I think that these scenarios still differ (apart from the obvious point that the undertakings there are not dominant), considering that Google’s business model, at least from the users’ perspective, is more directly aimed at providing consumers with choices. Although consumers seem also to prefer supermarkets with a wide range of choices, it is at least theoretically possible to imagine a supermarket offering only one product per category (one cheese, one bread etc.) all coming from the same producer. On the other hand, Google’s general search engine would completely collapse if it would show only results from the same source. In this light, while the Court’s formulation (“abnormality”) might have been chosen more carefully, it remains that giving up neutrality completely for the most visible part of the search result page is hardly explainable by pro-competitive motives. Google’s proximity to and influence over consumer choices (which is at the core of its business model) are the delineating factors compared to the essential facilities cases.
- Why did the Court not buy Google’s argument that everything Google was doing is improving services for consumers? A product improvement is nothing else than an efficiency gain, so the Court’s treatment of this issue under the question of justification makes sense. Thus, it was for Google to raise the plea and support it with evidence (para. 554). However, Google’s arguments that Products Universal and the Shopping Unit improved the quality of its general search could not explain why it did not include competing comparison search services. After all, this would have most likely made the improvement even better from the consumers’ perspective. While it is difficult to establish when results are actually more relevant in general, it seems logical that results are likely to be more relevant when they are chosen among a greater number of sources (see: para. 562). Hence, Google failed to explain why both parts of the anticompetitive conducts – the change in the general search service design and the unequal treatment – were necessary for any product improvement. The mere fact that users actually seemed to click more on the particularly visible Google shopping results does not prove the alleged improvement because the link between consumer choices and market success is broken: Google can basically divert consumers’ attention to wherever it wants (see: para. 566). Turning to Google’s argument that it was technically impossible to treat equal results from competing shopping services, the Court made the point that equal treatment does not require that Google applies exactly the same processes and methods but rather equal treatment in the results. Considering that Google has quite some experience in ranking results from third-party services, which, after all, is its core business model, it seems indeed far-fetched that Google could not have found a way to do so. The Court could perhaps have said a few more words on why the demotion of product results in the general search results was no product improvement. However, the key point here seems that Google was contradicting itself: while denoting product results in the general results, it showed them – on the prime spot – in Products Universal and the Shopping Unit.
The judgment will certainly make it more difficult for Google to deviate from the traditional Google search engine and further develop its search engine by proactively deciding what to show to the consumer and where to steer her clicks. Every change that Google makes will be very carefully looked at by competition authorities (and competitors for that matter), and it will be tough for Google to justify any behaviour which is considered anticompetitive on the basis of “product improvements”. Some months ago, for example, a court in Munich put an end to Google showing health-related information from only one source in an info-box on top of the search results, and the Bundeskartellamt is already investigating the Google News Showcase (see for an overview of the Bundeskartellamt’s big tech investigations here). While some will certainly argue that this strict scrutiny harms innovation, the point can be made that it gives back some power to consumers to decide for themselves what they want to see and click on.