The EC’s Google Search (Shopping) precedent
Self-preferencing as leveraging abuse.
In its Google Search (Shopping) decision of 27 June 2017 (the Decision), the Commission found that Google had abused its dominant position on the national markets for general search services by favoring, in its general search results pages, Google’s own comparison shopping service (CSS) as compared to more relevant CSSs. Such self-preferencing was seen as an anti-competitive expansion of dominance from the primary markets for general search services to the secondary markets for comparison shopping services because it conveyed Google’s CSS competitive advantages that did not reflect the quality of the product.
Google’s comparison shopping service - shopping units and standalone website.
Google offered its own CSS, called Google Shopping, through two separate consumer-facing frontends. In all countries in which an infringement was found, Google provided a variety of so-called ‘Shopping Units’; boxes with product images and additional product information. Whenever a user entered any product-related query in the toolbar of Google’s general search service, Google compiled a Shopping Unit with product results that match the query and display it at the top of Google’s general search results pages. Any click on a link in such a box led the user directly to the website of merchant where the respective product could be bought. Clicks on such links stood for the vast majority (more than 80%) of traffic to Google Shopping and made up the core of the case. In seven of the thirteen countries in which an infringement was found, in addition to providing Shopping Units, Google operated a standalone Google Shopping website, where users could enter a query and compare the products and prices separately. However, such websites attracted only few users and less than 20% of the traffic that the condemned conduct converted to Google went to such sites.
The results in both frontends, the Shopping Units and the standalone website, were powered by a common Google infrastructure that constituted Google Shopping. Such infrastructure consisted of a specialized search technology (product search algorithms, product catalogue, index etc.) and a merchant interface to onboard and validate product offers uploaded by merchants as structured data. Merchants could bid for their product offers to appear in Google’s Shopping Units. They could do that themselves by setting set up accounts, uploading their product feeds and bidding directly. Or they could use specialized marketing intermediaries that manage Google Shopping advertising campaigns on the merchant’s behalf. For such technical support, over the years a separate market with hundreds of agencies, affiliate networks, and ad tech companies had developed. Since such players only provide (one-sided) support services to merchants, they operate on a market separate from that of (two-sided) CSSs.
Display of shopping units as abusive favoring of Google’s CSS.
Reflecting the two separate frontends of Google’s CSS, the Decision defined the abusive favoring as “the more favorable positioning and display of links to Google’s own comparison shopping service [..] and/or parts or all of Google’s own comparison shopping services.” The favored “links” to Google’s CSS meant headlines above Shopping Units that linked to the standalone Google Shopping website. The favored “parts or all” of Google’s CSS means the Shopping Units themselves which compare products and prices directly in general search results pages. In countries where Google operated a standalone website, Shopping Units constituted only a “part” of” Google’s own CSS (beside the website). In countries where Google only provided Shopping Units they constituted “all of” Google’s own CSS. In addition to the favorable positioning and display of links to the standalone website or the display of the Shopping Units, Google favored its own CSS by applying ranking mechanisms that pushed rival CSSs further down the results page but were not applied to Google’s own CSS, even though this service exhibited the same characteristics that made rivals prone to be demoted. This practice was referred to as ‘demotion.’
Because the Decision’s central definition of ‘favoring’ covered both a favorable ranking and display of links to a standalone website and of Shopping Units powered by Google Shopping, the Decision consequently found an infringement also in those seven countries where Google did not operate any standalone Google Shopping website at all but only provided Shopping Units. In these countries, the abuse was the fact that Google displayed the frontend of its own CSS, the Shopping Unit, with visually enriched product results from the CSS’s own index and based on the CSS’s own algorithms directly within general search results pages, while all rival CSSs were only displayed as blue links above or below such boxes. Google’s inclusion of Shopping Units in general results pages was considered as an abuse despite the fact that no links in such Shopping Units led the user to any separate standalone Google Shopping website. Neither did it preclude the abuse that intermediaries, including rival CSSs, had the opportunity to place product ads within Google’s Shopping Units on behalf of their merchant customers.
Cease-and-desist remedy: equal treatment within general search results pages.
Due to strong network effects on the market for CSSs, the abusive preferencing of Google’s service had already caused damage to the structure of that market. In order to enable formerly disadvantaged competitors to regain strength, and to thus effectively eliminate the consequences of the infringement, the Commission could have imposed restorative remedies that actively re-instate the competitive process. However, in the interest of Google, the Commission limited itself to a ‘cease-and-desist’ order: Google was given 90 days to “bring effectively to an end the infringement” and refrain from repeating any equivalent act or any act “having the same or equivalent object or effect.” In line with this lenient approach, the Commission left it to Google to choose between the various technical ways in which the search engine could cease the infringement by treating downstream rivals equally. Any such measure should, however, “ensure that Google treats competing shopping services no less favorably than its own comparison shopping service within its general results pages.”
Google’s chosen compliance mechanism
While prior to the Decision, Google Shopping encompassed both the provision of Shopping Units and of a standalone website, in order to implement the remedy, Google split up these two activities into separate business units. Google now refers to the standalone website for product and price comparisons as Google Shopping Europe (GSE) and presents it as its “Google CSS.” However, the service has no impact anymore on the compilation, design, and content of Shopping Units which made up the heart of Google Shopping. Instead, the provision of Shopping Units is now run by a separate Google business unit, which Google has not given a separate name to hide its existence. This new business unit now exclusively operates the infrastructure required to provide Shopping Units within general search results pages.
Google carried out this internal restructuring with a view to its interpretation of the Commission Decision. Google interpreted the imposed remedy as meaning that “when Google shows a Shopping Unit, Google must give aggregators the same access to the Shopping Unit as it gives the Google CSS, using the same mechanisms (processes and methods) to allocate access.” In other words, Google interpreted the Decision as (merely) meaning that (i) Google may continue displaying its main CSS, Shopping Units that compare products and prices based on its specialized product search infrastructure, prominently in general search results pages, as long as (ii) competing CSSs are able to ‘access’ these boxes in the same way as Google’s minor CSS, the separate business unit Google Shopping Europe operating the standalone product comparison website.
Based upon this interpretation of the remedy, Google has neither removed any of the demotion algorithms that are applied to rival CSSs, nor altered the design, the positioning, or the display of Shopping Units. Instead, Google has made two changes. First, while previously merchants only had the option to use intermediaries to upload product inventory to Google’s index, and to bid for product ads to appear in Shopping Units on the merchants’ behalf, merchants must now formally engage a ’CSS’ as defined by Google for their campaigns. Second, while previously the name of such intermediary bidding on behalf of a merchant did not appear in Shopping Units, its brand is now displayed below every product as a link “By CSS” that, if clicked, leads the user to the website of the intermediary. All other links in Shopping Units continue to lead the user directly to the website of the merchant on whose behalf a product ad has been served.
CSSs already had the same option to bid for product results to appear in Shopping Units prior to the Decision, if they bid on behalf of merchants, i.e., also acted as intermediaries on a market separate to that of CSSs. Google’s Compliance Mechanism (CM) did not provide any new business opportunity to them. Therefore, following its introduction, the uptake of the CM amongst genuine CSSs was very low. To counter this, Google defined a ‘CSS’ that could take part in the CM so minimally that not just genuine CSSs but all sorts of online intermediaries, in particular marketing agencies, affiliate networks, and other ad tech companies, could fulfil such criteria without any significant investments within just a few days. All they needed to do was to set up a website and to place one offer from just fifty merchants onto it. In the real world, outside of Shopping Units, such ‘offers’ websites would never suffice to provide a genuine comparison shopping service that attracts any users. However, these low requirements were designed to allow all of the hundreds of intermediaries that had been managing Google Shopping campaigns for merchants for many years (during the infringement) to quickly turn themselves into CSSs as defined by Google (yet not the Decision), in order to take part in the CM as apparent CSSs.
Having set up the scheme, Google did its best to then incentivize merchants to use such fake CSSs instead of Google Shopping Europe to bid on their behalf. Amongst several measures, Google granted merchants that switch from its service to such CSSs a discount of up to 30%. All of this was aimed at making it as easy and lucrative as possible for advertising intermediaries to fulfil the low criteria to qualify as a CSS under the CM, in order that Google could allege to the Commission that more than 600 CSSs  would have entered the market as a success of the CM.
Study on economic impact of Google’s compliance mechanism
From the first day of Google’s roll-out of the CM, (genuine) CSSs had consistently pointed out that the mechanism is not improving anything as it simply upholds the abuse. However, while the Commission has never approved the CM, it did not launch an investigation into Google’s non-compliance. Frustrated by the Commission’s hesitation to timely enforce its Decision, 25 CSSs from across Europe, including the market leaders in several countries, joined forces to prove Google’s non-compliance. They combined their (anonymized) data relating to the CM, such as on consumer traffic, conversions, revenues, and customer reactions, and commissioned a study on the economic and legal implications of Google’s chosen CM.
The author’s study analyzed a representative total of 10.5 billion clicks on Google, relating to over 1 billion in total revenues. It found that the CM only benefits one CSS: Google’s Shopping Units. The data revealed, inter alia, that Google’s CM did not lead to any increase in generic search traffic to competing CSSs, even though the Decision is all about Google diverting generic search traffic. Under the CM, the profitability of generic search traffic even further dropped by 51.4%.
Competing CSSs are unable to outweigh the reduced quantity and profitability of the deprived generic search traffic (which Google’s CM causes) by buying product ads in Google’s Shopping Units on behalf of merchants. That is because such product ads do no lead the user to the website of the CSS but directly to the merchants. Accordingly, more than 99% of any clicks on Google’s Shopping Units lead the user to a merchant’s website, completely bypassing any rival CSS’s offering. Less than 1% of all users click on the “By CSS” link that lead to the website of a CSS. During their entire customer journey, searchers only ever engage with one CSS: Google’s Shopping Units.
Moreover, since CSSs have to bid in an auction to get merchants’ product results into Shopping Units, relying on such results is not a feasible business model. Due to increasing auction payments to Google, the ‘profits’ from commissions for clicks on such products results (i) have decreased by 46.5% since the launch of the CM; and (ii) are currently half of that of clicks on generic search results. A CSS would need to attain two clicks on an ad for each click on a generic result that it loses because of Google’s self-preferencing, which is impossible.
The study further found that the vast majority of the allegedly 600 companies that participate in the CM are no genuine CSSs that compete with Google on the merits. They are advertising firms that only set up ‘hollow’ websites to qualify as a ‘CSS,’ as broadly defined by Google. Without any own backend (product catalogues, product indexes and specialized search algorithms), such ‘fake CSSs’ only exist in Google’s Shopping Units. They gained no market shares on the national markets for CSSs. Moreover, the low profit margin that the CM inherently leaves bidders precludes genuine CSSs from financing or expanding their services, and disallows ‘fake CSSs’ to ever progress into becoming viable CSSs. Instead, the mechanism creates a ‘prisoners’ dilemma’, where (even quality) CSSs feel choiceless but to bid for a product ad in Google’s Shopping Units (to remain findable at all), but ultimately all CSSs are worse off by doing so because the auctioneer – Google – expropriates the CSS’s surplus. Overall, the empirical data showed that Google’s CM did not improve the situation for any competing CSS. The only economic beneficiary is Google.
Legal assessment of Google’s compliance mechanism
The author’s study of Google’s CM further found that the fact that Google’s CM had no economic impact is no coincidence. Its failure was the inevitable consequence of Google’s non-compliance with the equal treatment remedy imposed. Amongst the various options that were available to cease the infringement, Google designed, tested, implemented, and maintained a mechanism that neither brought the anti-competitive conduct nor its effects to an end but had the same object and effect as the identified abuse.
The reason for this ignorance of the imposed remedy likely stems from the legal theory that Google relied upon throughout the entirety of the Commission’s investigation. Google has always argued that the case solely concerned “access” of competitors to the Shopping Units. Google hoped that this would bring the case under the realm of the Bronner-case law for refusals to deal, which has the highest threshold for finding an abuse of dominance under Article 102 of the European Treaty. The Commission had consistently rejected this notion. The Decision explicitly clarifies that “the Conduct does not concern a passive refusal by Google to give competing comparison shopping services access to a proportion of its general search results pages, but active behavior relating to the more favorable positioning and display by Google, in its general search results pages, of its own comparison shopping services compared to [competing] comparisons shopping services.” Accordingly, there is no discussion of any “access” to any Google input in the Decision at all.
However, when setting up the CM, Google ignored this part of the Decision and simply adhered to its own (but rejected) theory: “The remedy chosen by Google is therefore framed as an ‘access remedy’ in that, through the auction mechanism, Google is giving rival comparison shopping services access to the Shopping Unit.” Google tries to focus all attention on the alleged equal conditions for “access to the Shopping Unit” to suggest that an equal right of CSSs to bid for product ads was required and sufficient to comply. Yet, equality within a Shopping Unit is far less than equality within Google’s general search results pages, which the Decision’s remedy actually demands. The relevant legal question is not whether Google grants CSSs any access to Shopping Units on equal terms. The question is whether it “treats competing [CSSs] no less favorably than its own [CSS] within its general search results pages,” of which Shopping units are just one element. Under the CM, equal treatment within general results pages blatantly does not exist because Google remains the only CSS entitled to display rich boxes that satisfy the consumer demand for comparison services directly within such pages. Only Google’s CSS (using Shopping Units as relevant interface) may match a query with results from its own index, based upon its own specialized algorithms. All competing CSS are reduced to serve as an acquisition tool for Google that brings more merchants so that they serve more ads in Google’s Shopping Units.
Lessons for antitrust enforcement in digital markets
Google Search (Shopping) is not the only Commission decision in digital markets which, thus far, failed to have an impact because the Commission has not enforced the remedy it imposed. Another example is the Android case. In both cases, the Commission has hesitated to enforce the decision prior to a final court ruling on the appeal against the decision. Yet, in both cases this hesitation risks undermining the value of the entire investigation of many years in terms of ensuring competition. A victory of the Commission in court helps little if by then all competition is dead because the remedy has not been applied in the meantime.
To avoid similar pitfalls, antitrust enforcement authorities could consider some of the following ten measures when defining and supervising remedies for fast-moving digital targets:
Measures to identify and define suitable remedies.
- Do not make an investigation or the finding of an infringement dependent on a clear-cut remedy, but consider suitable remedies from the very beginning and discuss them with affected parties prior to imposing them: Designing a suitable remedy can be challenging. However, a competition authority should not hesitate to launch an investigation and to find an infringement just because the remedy is not clear-cut. “Not trying to fix something is only an option when it is not broken". Instead, the authority should consider and discuss suitable remedies with both the company under investigation and, more importantly, its affected rivals from the very beginning of the investigation. They should be given the opportunity to comment on the set of remedies envisaged by the authority prior to imposing them on the defendant.
- Close the technical information asymmetry gap to the undertaking under investigation: One of the main reasons for poor remedies is an insufficient understanding of the technicalities of the products and services in question and their interplay with the market. Competition authorities may be afraid to impose far-reaching remedies in fear that it may be technically impossible for the defendant to implement them. Such fear may encourage the authority to define the remedy very vaguely (e.g. a pure cease-and-desist) and to leave the selection of a technically feasible implementation to the defendant. However, such strategy may quickly backfire because it encourages the defendant to choose the least effective implementation amongst the variety of technically feasible solutions. Instead, the authority should (i) request the defendant to explain its service, (ii) engage with rivals and (iii) consult external experts to close the information asymmetry so as to be able to define precise remedies that are technically sound.
- When drafting the remedy, combine unambiguous general objectives with a non-exhaustive list of specific obligations arising from the general principles. The remedy sets the legal framework for the subsequent compliance mechanism. Avoiding loopholes is crucial. An infringement is typically the result of a combination of several factors that render a conduct anti-competitive. Limiting a remedy to an obligation to cease the infringement risks that the defendant just mitigates one factor, so as to fall out of the definition of the infringement, while intensifying other factors, so that the anti-competitive effects remain. Accordingly, a remedy needs to be defined as precisely as possible and as broadly as necessary to avoid anti-competitive ways to work around it to achieve the same ends as the prohibited conduct. To avoid loopholes, the authority should set out general objectives that the undertaking needs to achieve and specify these objectives with a non-exhaustive list of specific prohibitions or obligations arising from the general principle. Combining a principle-based remedy with precisely defined examples makes the remedy less susceptible to strategic neutralization, minimization, or evasion by defendants. This is particularly relevant for behavioral remedies, for which there is a high risk that the defendant seeks to frustrate its effectiveness by endlessly litigating its terms or finding ways to work around it, which can require extensive oversight.
- Consider pro-active remedies, at least to flank or back-up cease-and desist orders: Where a market with strong network effects has already ‘tipped’ in favor of the defendant as a result of its infringement, a mere cease-and-desist is incapable of re-establishing competition. In such scenarios the remedy to stop the prohibited conduct needs to be flanked with obligations that pro-actively initiate competition. This may include, for instance, the active favoring of rivals, the sharing of data or the granting of interoperability with competitors, so as to neutralize the advantages and network effects that the defendant has acquired through the infringement.
- Consider a staggered remedy where a more lenient behavioral remedy is replaced with a more stringent behavioral or structural remedy in case the former fails. Competition authorities are faced with two conflicting objectives. They need to bring the infringement to an effective end. Yet, they also need to limit their measures to what is necessary. To address this conflict, an authority may consider a staggered remedy. As a first step, it may impose a more lenient behavioral remedy that leaves a broader discretion to the defendant as regards its implementation. However, to ensure that the defendant effectively implements this original remedy, the authority may simultaneously impose a more stringent (e.g. pro-active or structural remedy) that applies automatically in case the infringement has not ended within a certain time limit.
Measures to effectively assess and monitor the defendant’s compliance mechanism.
- Staff and incentivize the monitoring team wisely: The Google Shopping case showed how crucial it is for the assessment of a compliance mechanism to fully understand the content and scope of the original decision. The team monitoring a compliance mechanism therefore needs to encompass the most knowledgeable case handler from the preceding investigation. The team also needs to be sufficiently staffed to master the legal, technical and economic complexity of a non-compliance case. If an authority takes recourse to external advisers, it needs to ensure that they have no financial incentive to drag out the monitoring process rather than to address identified failures straight away. This would be the case, in particular, if the advisers are paid more the longer the monitoring goes on.
- Supervise the undertaking’s internal testing of any envisaged compliance mechanism prior to its implementation. When authorities leave the selection of a compliance mechanism to the digital platform, as in Google Shopping, they can be certain that every such proposal will be the result of a rigorous internal experimentation process to test its impact. Authorities can take advantage of this by requiring the defendant to involve the authority in such testings, and to grant access to the same information as the undertaking relied upon selecting a specific mechanism. To further reduce the information asymmetries, the authority could also impose on obligation that allows requesting the defendant to carry out any A/B tests or similar experiments to verify the effects of any alternative solutions to the compliance mechanism chosen by the defendant. This will allow the authority to cross-check any efficiency arguments raised in defense of such mechanism and reveal the feasibility of other technical methods.
- Thoroughly market test a compliance mechanism in due course. One of the reasons for the Commission’s long hesitation to address Google’s non-compliance with its Shopping and Android decisions may have been a one-sided communication and flow of information. The Commission only received information on the compliance mechanism from the company under investigation, Google. Competitors did not get any access to that information to verify the data or point out any misrepresentations. That is regrettable because competitors typically have the best market insights to evaluate the effects of a compliance mechanism. The one-sided (and presumably selective) briefing of the Commission by Google led to significant misunderstandings as to the functioning and effects of its compliance mechanism. To avoid any such shortcomings, the authority should share the defendant’s explanation of the compliance mechanism as well as any subsequent reports on its alleged progress with affected parties, in particular the complainants, and allow them to comment.
- Consider a trialogue hearing on the merits of the compliance mechanism with both the defendant and affected parties present. To assess a company’s compliance with a complex decision can be challenging. An effective means to determine all pros and cons can be an open discussion between the company under investigation and affected competitors. To this end, if competitors plausibly allege a failure to comply, the authority should organize and supervise a trialogue hearing where both sides can present their views and react to the other party’s claims. Such a hearing can be much more effective for all parties involved as compared to several individual talks behind closed doors between the authority and the numerous parties potentially involved.
- If shortcomings of the chosen compliance mechanism emerge, intervene resolutely and immediately, irrespective of whether there is still a pending appeal against the decision. It is understandable that competition authorities are reluctant to enforce a decision that is still under appeal. If the original decision is annulled, any non-compliance decision will be void. However, if undertakings learn that an authority does not enforce any prohibition decision prior to the final court ruling on its legitimacy several years later, infringers have a strong incentive to both appeal each decision and to not comply with it until the final court ruling. In fast moving markets where the winner takes all, this is likely to be the case even if non-compliance triggers an automatic penalty. The long-term gains of tipping a market by being non-compliant may be far higher than the one-off penalty for non-compliance. Accordingly, to make the enforcement of a remedy dependent on the outcome of any appeal against the prohibition decision significantly reduces the deterrent effect and hence the effectiveness of public competition enforcement. Instead, when after a long investigation, an infringement has finally been found and a remedy has been imposed, ensuring compliance with this remedy should be an absolute enforcement priority for any authority. Otherwise the authority jeopardizes the entire value of the investigation to promote competition and ultimately its own credibility.