Self-Preferencing in Digital Markets

This is an Insight article, written by a selected partner as part of GCR's co-published content. Read more on Insight


Since the European Commission (EC)’s milestone Google Shopping decision,[2] self-preferencing theories of harm have taken a key role in European competition enforcement in digital markets. The recent judgment by the General Court (GC) upholding the EC’s decision confirmed that self-preferencing by a dominant firm can be a stand-alone abuse in certain circumstances.

Which types of conduct exactly constitute self-preferencing is still a hotly debated topic. Enforcement activity in Europe so far has cast a relatively wide net, with various national competition authorities and the EC focusing their investigative efforts on cases where dominant vertically integrated platforms have given more favourable treatment to their own products or services to the detriment of non-affiliated rivals.[3]

Recent regulatory efforts at national and EU levels have also targeted various forms of self-preferencing by dominant digital platforms. For example, Germany revamped its competition rules at the beginning of 2021, banning ‘Undertakings of Paramount Significance’ in multi-sided markets from presenting their own offers more favourably than those of rivals, and from pre-installing their own offers on devices when providing access to supply and sales markets.[4] Similarly, the UK is moving towards a regulatory regime where a Digital Markets Unit (DMU), which is part of the Competition Markets Authority (CMA), can designate certain undertakings as having ‘Strategic Market Status’. The DMU can then specify bespoke codes of conduct for each of these undertakings, which could include requirements to not engage in undue self-preferencing of its own services and, if necessary, impose ‘pro-competitive interventions’ such as imposing functional separation remedies to remove self-preferencing incentives.[5] At the EU level, the Digital Markets Act (DMA) that entered into force in November 2022 prohibits digital platforms designated as a ‘gatekeeper’ from ranking their own products and services more favourably than those of third parties.[6]

This chapter gives an overview of the key issues relating to self-preferencing theories of harm in digital markets. We begin by providing a brief summary of the key points in the EC’s Google Shopping decision and the GC’s subsequent upholding of it.[7] We then present a brief overview of recent competition enforcement cases that have considered self-preferencing. In particular, we summarise the main findings from three recent self-preferencing investigations that were concluded at the national level but are ongoing at the EU level with respect to the conducts considered. Next, we explain the broad economic theory of harm associated with self-preferencing. Finally, we discuss the regulatory approach adopted by the DMA with respect to self-preferencing.

Self-preferencing as an abuse of dominant position

Following a seven-year investigation, the EC fined Google €2.4 billion for infringing Article 102 TFEU. The investigation was opened in 2010 and reinvigorated in 2016 with a refreshed focus on Google’s comparison-shopping service (CSS) after a failed market test of commitments proposed by Google.[8] While Article 102 TFEU prohibits dominant firms from ‘applying dissimilar conditions to equivalent transactions with other parties, thereby placing them at a competitive disadvantage’, it was the EC’s 2017 Google Shopping decision that put self-preferencing on the radar of European competition enforcement as a standalone abuse of dominance.[9]

Having established Google’s dominance in the relevant national markets for general search services and having identified distinct national product markets for CSSs,[10] the EC found that Google abused its dominance in general search services in each of the 13 national EEA markets under consideration, by positioning and displaying its own CSS on its general search engine results page (SERP) more favourably compared to rival CSSs.

First, the EC found that web pages of rival CSSs could only appear as text-based results in Google’s SERP, and their SERP ranking was prone to demotion by Google’s algorithms.

Second, the EC found that Google’s own CSS was prominently positioned at the top of the SERP, displayed in richer format and was not subject to demotion by its own algorithm.[11]

The EC concluded that Google’s conduct was capable of extending its dominant position in the national markets for general search services into the national markets for CSSs,[12] and capable of having anticompetitive effects in both sets of markets. In addition to the fine, the EC ordered Google to comply with the principle of equal treatment by implementing a measure of its own choosing that would subject its CSS to the same process of determining the positioning and display on the Google SERP as rival CSSs.[13]

The EC’s decision was almost entirely upheld by the GC in its judgment dismissing Google’s appeal in November 2021.[14] In doing so, the GC made two points confirming that self-preferencing by a dominant firm could be considered, in certain circumstances, as a stand-alone abuse of dominance.

The GC clarified that self-preferencing is one of the many ways in which leveraging can manifest,[15] and found that abusive self-preferencing as implemented by Google is distinct from an express refusal to supply.[16]

The GC confirmed that leveraging is not necessarily abusive and certain conditions must be met for it to be considered as an abuse of dominance.[17]

The GC has also found that Google’s conduct did not amount to an express refusal to supply and ruled that the EC was therefore not required to establish that the criteria of the essential facilities doctrine (as laid out by the Court of Justice in the Bronner judgment, which include the requirement of indispensability) were met to show that an abuse has taken place.[18],[19]

The GC’s judgment highlights two conditions for self-preferencing to constitute an abuse of dominance:

  • the conduct must have actual or potential anticompetitive effects;[20] and
  • the conduct must depart from what would be expected under normal competition.[21]

The GC ruled that the EC’s investigation ticked both of these boxes (with the first condition relating to the national markets for CSSs). In doing so, the GC also clarified that a causal link relating to the first condition can be established by showing that there is a correlation between the conduct and market outcomes and there is additional corroborating information (such as assessments of market participants).[22], [23] Further, in relation to the second condition (‘abnormality’), the GC noted that a ‘product improvement’ defence should be considered only at the stage where objective justification and potential efficiencies are examined, when the conduct is capable of having anticompetitive effects.[24]

Competition enforcement involving tech giants is increasingly focusing on self-preferencing

Since the EC’s Google Shopping decision, there has been an increasing number of investigations in Europe that rest on abusive leveraging theories of harm, as well as enforcement of self-preferencing (or closely related derivatives). The EC initiated an investigation in November 2020 into Amazon’s practices of self-­preferencing its logistics services, which is currently at the stage of market-testing the commitments proposed by Amazon.[25] Further, in June 2021 the EC also started investigating whether Google abused its dominant position by favouring its own ad tech services.[26] These types of conduct, or closely related ones, were also the subject of three recent decisions by different European national competition authorities, which we summarise below.

In the first of these decisions, the French Competition Authority (FCA) concluded a two-year investigation into Google’s practices relating to its ad tech business in June 2021, following a complaint filed by several media groups. The investigation related to the allocation of display ads. Typically, when a user visits a website or an app, slots for display ads on the website (the ‘publisher’) are allocated to advertisers via a series of almost-instantaneous auctions that are run by ad tech intermediaries.[27] The investigation focused on two of the several interrelated markets that make up the supply-side (i.e., publisher side) of this intermediation chain:[28] first, the market for publisher ad servers where the FCA found that Google held a dominant position with its DoubleClick for Publishers (DFP) service; and second, the market for ad space sales platforms (also called a supply-side platform (SSP)) where the FCA found that Google’s Ad Exchange (ADX) service held at least a pre-eminent position (without establishing dominance).[29] The FCA then established two distinct but closely related abusive practices by Google.

First, the FCA found that Google’s publisher ad server DFP self-preferenced by consistently giving more favourable conditions to its SSP ADX than those given to rival SSPs, which allowed ADX to consistently outbid its rivals.[30],[31] The FCA saw this as an act of abusive leveraging, extending Google’s market power from the dominated market for publisher ad servers into the market for SSPs.

Second, the FCA found that Google’s ADX self-preferenced its DFP service by withholding full interoperability with rival SSPs, which led to publishers preferring Google’s DFP over other publisher ad servers. The FCA viewed this conduct as strengthening Google’s dominance in the market for publisher ad servers.

Based on the above, the FCA concluded that Google infringed Article 102 TFEU by abusing its dominant position in the market for publisher ad servers, imposed a fine of €220 million on Google, and accepted binding commitments from it.

In November 2021, shortly after the GC’s upholding of the EC’s Google Shopping decision, the Italian Competition Authority (ICA) concluded its investigation into Amazon’s practices of self-preferencing its e-commerce logistics services that it offers to merchants selling on its Amazon Marketplace platform. The ICA first established the ‘super-dominance’ of Amazon in the narrowly defined Italian national market for intermediation services on online marketplaces (i.e., a market including e-commerce platforms such as where third-party merchants can sell their products).[32] The ICA then established an abusive conduct that rests on two legs.

First, the ICA found that Amazon tied the allocation of the Prime badge to third-party merchants to the use of Amazon’s own logistics service Fulfilment by Amazon (FBA) for a large part of the review period.[33] Obtaining the Prime badge is important to merchants because it makes it easier to access a large number of high-spending Prime consumers and it allows the merchant to participate in special events promoted by Amazon (such as Black Friday sales).[34]

Second, the ICA also established that the algorithm Amazon uses for selecting featured offers that appear in the ‘BuyBox’ (a box positioned at a highly visible place on Amazon’s product search results page) discriminates against merchants who are not using FBA as their logistics provider for products that they sell on Amazon’s marketplace.[35]

The ICA found that Amazon’s conduct had a dual effect by extending Amazon’s dominance in the market for intermediation services on online marketplaces to the market for e-commerce logistics, and also strengthening Amazon’s dominant position by discouraging merchants from selling their goods in other online marketplaces.[36] Based on this, the ICA concluded that Amazon infringed Article 102 TFEU by abusing its dominant position, and imposed a fine of €1.1 billion on Amazon as well as behavioural remedies intended to restore competitive conditions in the relevant markets.[37]

The CMA recently concluded an investigation into Google’s ‘Privacy Sandbox’ proposals. The CMA had previously expressed concerns about these proposals in its online platforms and digital advertising market study.[38],[39] Six months after publishing the study, and following the complaints from various stakeholders who alleged that Google’s proposals amounted to an abuse of dominance, the CMA’s investigation was opened on 10 January 2021. Unusually, the investigation concluded before Google’s proposals were implemented. Google’s proposals concerned replacing its cross-site tracking of users on Chrome (Google’s web browser) via third-party cookies and other methods with alternative tools that would provide similar functionalities.[40] Two of the CMA’s main concerns in regard to Google’s proposals related to self-preferencing.[41]

The CMA highlighted the risk that these proposals would limit the information collection and targeting abilities of rival publishers and ad tech providers. Meanwhile Google’s own abilities would be unaffected as it would maintain advantageous access to first-party data.[42]

The CMA pointed out that some of the functionalities currently performed by ad tech providers would move to Google Chrome under the proposals. The CMA argued that this would give Google the opportunity to leverage its ‘likely dominant’ position in the market for web browsers into relevant markets relating to open market advertising,[43] by self-preferencing its own ad tech services or the ad inventories that it manages.

The CMA found that the Privacy Sandbox proposals, if adopted, would likely amount to an abuse of dominant position without sufficient regulatory oversight. Furthermore, the CMA found that the announcement of the proposals itself as well as the preliminary steps that Google has taken to implement them likely constituted an abuse in the specific circumstances of the case.[44] The CMA decided to close the investigation after agreeing to the final set of commitments offered by Google.[45]

Recent investigations focusing on self-preferencing are not limited to Amazon and Google. Apple was fined by the Dutch Competition Authority in 2021 for prohibiting dating app developers from using third-party payment systems in their iOS apps, and is facing an EC investigation for the same conduct vis-à-vis music streaming app developers.[46] German and Polish competition authorities recently started to look into whether Apple’s App Tracking Transparency (ATT) Framework, which obliges third-party apps to ask for their users’ content, constitutes abusive self-preferencing on the basis that it does not seem to affect Apple’s ability to use and combine user data from its own ecosystem.[47] Apple also faces an EC investigation for preventing mobile wallet app developers from accessing software and hardware components necessary for implementing contactless payments, to the benefit of its own mobile wallet Apple Pay.[48] Similarly, Meta faces an investigation by the EC relating to its use of the data it collects from online classified ads providers who offer their services via Facebook, as well as its integration of Facebook Marketplace into its social network platform, which could constitute a form of tying.[49] While the EC did not explicitly label Meta’s conduct as self-preferencing, its concerns relate to the same underlying consideration of a platform using its dominance in the market where its core platform services operate (the social network) to potentially foreclose third-party suppliers in an adjacent market within its ecosystem (online classified ad services). The precise conduct via which leveraging takes place, or the label chosen to describe it, is becoming of secondary importance.[50] Overall, competition authorities are increasingly eager to challenge the core strategies of the ‘big tech’ companies, including those that may have helped them build their ecosystems in the first place, with overarching concerns about potentially abusive leveraging and distorting competition in adjacent markets, as well as strengthening of existing dominant positions.

Theories of harm in self-preferencing cases

The theories of harm in cases concerning self-preferencing or leveraging have two central elements: foreclosure and consumer or user harm. The foreclosure part of the story typically follows a similar pattern. First, a vertically-integrated platform operator implements a discriminatory mechanism that can generate input or customer foreclosure effects, depending on whether the platform is upstream or downstream vis-à-vis its ancillary service within the supply chain.[51] This may then lead to the partial or full exclusion of current (and potentially future) rivals in the market where the ancillary service is provided. Exclusion may occur via a combination of reduced access, increased costs or diminished incentive to innovate, depending on the severity of the discrimination strategy and whether rivals can substitute to inputs or customers from other sources (which in turn depends on the extent of the platform’s dominance). As a consequence, competition is distorted at the upstream or downstream market to the benefit of the platform operator’s vertical affiliate (leveraging). Distortion of competition in the upstream or downstream market can lead to indirect foreclosure effects in the platform market, further strengthening the platform operator’s dominance. In all the recent cases that are summarised above, the foreclosure parts of the theories of harm fit into this framework.

Google Search is an upstream supplier of ‘traffic’, which is a key input for CSSs including Google Shopping. Google’s practice of demoting the organic ranking of rival CSSs in the Google SERP and only allowing Google Shopping to benefit from a position at the top of the SERP and within a box displaying product links in richer format could be interpreted as a case of input foreclosure. The EC’s assessment indicates that Google’s conduct foreclosed rivals both directly – by reducing their supply of free clicks from the SERP and increasing the supply of free clicks to Google Shopping – and indirectly by increasing their costs since non-affiliated CSSs had to rely more on paid clicks by purchasing search ads.

Amazon Marketplace is a platform enabling third-party merchants to make sales to consumers. Third-party merchants purchase services from logistics providers, which include Amazon’s FBA service. The ICA found that Amazon treated third-party merchants who used its FBA service more favourably by providing them with enhanced access to consumers (compared to merchants who used alternative logistics providers’ services) and that this created a strong incentive for third-party merchants to choose FBA over the services of alternative logistics providers. The ICA also found that Amazon’s conduct had exclusionary effects in the market for e-commerce logistics by hindering integrated logistics operators’ ability to innovate,[52] and by preventing the ability of non-integrated delivery operators’ capability to improve their product offerings by reaching a sufficient scale of deliveries.[53] This could be interpreted as a form of customer foreclosure, since the practices in question were found to restrict rival upstream logistics providers’ demand.

Further, the ICA also assessed that foreclosure in the e-commerce logistics market was capable of distorting competition in the market for intermediation services on online marketplaces by discouraging third-party merchants on the Amazon Marketplace from selling on other e-commerce platforms, as that would either require replicating logistics costs, or purchasing Amazon FBA’s expensive multi-channel management services.[54]

Google’s publisher ad server DFP sells the ad inventory it manages to the highest-bidding advertisers. It determines the highest bids by procuring auction intermediation services from publisher-side SSPs, including its own ADX. More specifically, different SSPs submit offers (which reflect the underlying bids by advertisers and platform commissions) to Google’s DFP, and DFP picks the highest offer. As explained above, the FCA identified that Google implemented various mechanisms that grant an informational advantage to its own SSP ADX vis-à-vis its rivals, allowing ADX to consistently outbid competing SSPs in auctions organised by Google’s DFP. Given the dominant position of Google’s DFP, the FCA found that this led to the foreclosure of ADX’s competitors. This could be interpreted as a case of customer foreclosure since the mechanisms implemented by Google were found by the FCA to limit DFP’s purchasing of intermediation services from non-affiliated SSPs.

In addition, the FCA also established that Google ADX granted full interoperability only to Google’s DFP so that the latter was the only publisher ad server that had full access to ADX. This could be considered as an input foreclosure finding given the role of ADX as an upstream supplier to publisher ad servers.[55] The FCA found that this hindered the ability of rival publisher ad servers to compete against Google by forcing them to use Google’s DFP as their primary ad server, since ADX’s offers are generated from a larger pool of advertiser bids compared to competing SSPs.[56]

A self-preferencing (and leveraging) theory of harm would be incomplete from an economics standpoint without a description of how the actual harm to consumers materialises. In the recent decisions that are summarised above, competition authorities have offered varying degrees of detail when analysing the mechanisms through which consumer harm arises.

In its Google Shopping decision, the EC found that Google’s conduct could lead to higher fees for merchants by eliminating competition and increasing Google Shopping’s market power, as well as to higher consumer prices if merchants reflected the higher fees in their own prices.[57] The EC also stated that the conduct was likely to reduce CSSs’ incentives to innovate, indirectly harming consumers through reduced quality or relevance.[58] In addition, the EC found that the conduct would likely diminish consumers’ ability to access the most relevant comparison shopping service, explaining that Google’s CSS did not always show the most relevant results to users and that some consumers in some periods may not have been aware of this.[59]

In its Amazon FBA decision, the ICA highlighted that Article 102 TFEU also covers indirect consumer harm through distortion of the competitive process,[60] and emphasised how Amazon’s conduct was capable of distorting competition in two separate markets: e-commerce logistics and intermediation services on online marketplaces. While the ICA noted that this would negatively affect consumers, it did not explicitly explain how.[61]

The ICA focused on how the conduct deprived third-party merchants of the freedom to choose the logistics operator best suited for their business needs,[62] and pointed to evidence that seems to indicate that this may have hindered their ability and incentives to minimise their logistics costs.[63],[64] In addition, the ICA highlighted that Amazon FBA increased its share of deliveries in the Amazon Marketplace despite significantly increasing its storage and delivery fees in 2018 and 2019.[65] In conjunction with the ICA’s finding that Amazon’s conduct hindered innovation by logistics operators, it may be inferred that one of the ICA’s concerns regarding consumer harm could have been that higher logistics costs were passed on to consumers shopping on Amazon.[66]

The ICA also focused on the impact on rival e-commerce platforms and presented evidence that Amazon Marketplace was able to increase its market share substantially between 2016 and 2019. While the ICA did not elaborate on how this might have led to consumer harm, it argued that the market share trends reflected third-party merchants’ increasing propensity to single-home on Amazon Marketplace, because of their reliance on Amazon FBA.[67] With this in mind, another of the ICA’s concerns may be increasing prices of goods sold by third-party merchants due to diminished inter-platform competition. That is, more merchants selling only on Amazon may reduce the competitive pressure to lower merchant fees, which in turn could exert upward pressure on merchants’ downstream prices.

In its Google AdTech decision, the FCA focused on the harm to Google’s ad tech rivals and publishers. First, the FCA found that Google’s DFP was able to increase its market share while the market shares of rival publisher ad servers declined and some exited the market.[68] Similarly, the FCA found that Google’s ADX was able to sustain higher prices than its competitors without slowing its growth relative to rivals.[69] Second, the FCA found that publishers, particularly press groups, were deprived of higher competitive prices from SSPs and thus lost revenues that they could have earned by selling their ad inventories.[70] While the FCA did not discuss potential consumer harm that could result from this, the CMA notes that publishers’ incentives and ability to invest in content would likely decrease if they received a lower share of advertising revenues than they should, which would harm consumers who value this content.[71]

A first necessary condition for conduct to harm consumers is that it forecloses a substantial part of the market.[72] This depends on the importance of the platform as a supplier (input foreclosure) or buyer (customer foreclosure), as well as on the lack of availability of sufficiently large alternative routes to market that would allow foreclosed rivals to mitigate the loss of access to inputs and/or customers. A second necessary condition is the absence of offsetting efficiencies from the conduct.

In the cases referred to above, the competition authorities tended to focus their evidence collection efforts on the existence and significance of foreclosure rather than actual consumer harm.[73] This may be because they found it harder to show harm to consumers in the context of digital platforms, which may be for a variety of reasons, including that many platforms offer their services to users for free (i.e., the absence of a price parameter to measure), or the indirect and/or dynamic nature of the harm.[74]

Regarding efficiencies, we mentioned above that the GC considered in its Google Shopping judgment that efficiency arguments, including those relating to product improvement, should be brought in at the objective justification and efficiencies stage. In the investigations summarised above, both Google and Amazon proposed various efficiency and objective justification arguments to defend their conduct. These were not accepted by the competition authorities, which considered that the arguments were not sufficiently substantiated or that the efficiencies were not demonstrably linked to the conduct under investigation.[75]

However, it should not always be assumed that this would be the case. A relevant counterpoint is the dispute between Streetmap.EU and Google in the English High Court related to Google’s exclusive positioning of its own mapping service Google Maps into a graphical display box (Maps OneBox) prominently placed on top of its SERP, which can be considered to be an example of self-preferencing case, although it was not argued in that way, and indeed the case pre-dated the EC’s Google Shopping decision. In that case, Mr Justice Roth found that Google’s conduct corresponded to a technical improvement to the benefit of consumers, and it was objectively justified because there was no alternative, less restrictive way of achieving similar improvements without adopting a less practical design or incurring disproportional costs.[76]

With this in mind, efficiency and objective justification arguments could still play a role in the future enforcement activity concerning self-preferencing. However, as we discuss below, the agreed final text of the DMA may mean that such considerations are no longer relevant in relation to digital markets.

Self-preferencing in the Digital Markets Act

The DMA shifts the focus of European authorities’ enforcement activities towards ex ante regulation of dominant vertically integrated digital platforms.[77] The text of the DMA specifies a list of behavioural obligations that platforms designated as ‘gatekeepers’ must comply with.[78]

The list of behavioural obligations provided in Articles (5)-(7) of the DMA includes per se prohibitions of various types of self-preferencing and leveraging conduct. Several examples are provided below:

  • Article 6(5) prohibits gatekeepers from ranking their own products and services more favourably than those of third parties.
  • Article 5(7) prohibits gatekeepers from requiring that business users use the payment service of the gatekeeper.
  • Articles 5(9) and 5(10) imposes fee transparency requirements on gatekeepers operating in the ad tech supply chain vis-à-vis advertisers and publishers.[79]
  • Article 6(2) prohibits gatekeepers from using non-public data that are provided or generated by business users (in the context of core platform services or ancillary services provided by the platform operator) in competition with those business users.

The context of these prohibitions is that they relate to behaviour that had been considered in various different investigations by the EC and national competition authorities. These include the three cases considered above, the Dutch Competition Authority’s investigations into Apple’s practice of obliging the usage of its own in-app payment services to third-party app developers, as well as investigations into Amazon’s and Facebook’s use of data generated by their business customers (merchants and advertisers respectively) in their respective platforms.

The DMA’s per se prohibition of self-preferencing by gatekeepers is consistent with the recommendations of the EC’s panel of economic experts,[80] but it goes beyond the recommendations of the ‘Vestager Report’, in which the authors acknowledged that self-preferencing may have pro-competitive rationales and recommended it being subject to an effects test with the burden of proof falling on the dominant platform.[81] It also stands in contrast to the effects-based approach endorsed in the Google Shopping judgment, where the GC ruled that self-­preferencing by a dominant undertaking is not abusive per se, and in case its conduct was shown to be capable of restricting competition by an effects assessment, the undertaking may still show that its conduct generated mitigating efficiencies or convincingly that its conduct was objectively justified.[82] Except in relation to a limited set of behavioural obligations and within pre-defined limits,[83] the text of the DMA does not seem to contain a mechanism by which a gatekeeper can justify its behaviour on objective justification or efficiency grounds.

The recitals of the DMA highlight promoting innovation as a key regulatory goal and justify the behavioural obligations imposed in Articles (5) to (7) as necessary to increase contestability by reducing barriers to entry and expansion. With this reasoning, the DMA takes the view that weak contestability is embedded in the nature of the core platform services in the digital sector due to structural characteristics such as economies of scale, network externalities and learning-by-doing effects enabled by data and algorithms,[84] and that regulatory intervention is necessary to strengthen contestability.

The link between innovation and contestability (in a low-entry barriers sense) in the economics literature is not as straightforward as intuition would suggest. The DMA’s position that ‘weak contestability reduces the incentive to innovate and improve products and services for the gatekeeper, its business users, its challengers and customers’ has in its roots a view first introduced by Arrow who suggested that a monopolist would have a reduced incentive to innovate compared to a competitive firm because the former would lack the latter’s business-stealing motive.[85] Later, this perspective was complemented by the view that if the monopolist is threatened by entry, then it would have an incentive to innovate, sometimes even a stronger incentive than the prospective entrant because it has more to lose.[86] At the other side of this debate is the Schumpeterian view, which stresses the importance of a prospect to achieve higher market power for spurring innovation.[87] The critics of the DMA, accordingly, argue that ex ante regulation can weaken a platforms incentive to invest by reducing its ability to appropriate future rents associated with its innovation. We note that the empirical literature suggests that both the Arrowian and Schumpeterian views may co-exist.[88] We also note that the recent theoretical economics literature is ambiguous on the broader welfare effects of self-preferencing and leveraging in the context of vertically integrated digital platforms.[89],[90],[91]

The case for a per se approach in regulation is that it offers clarity and allows for speedy and effective enforcement. Proponents of the DMA have argued that this is especially important in digital markets since restorative remedies for abusive self-preferencing are particularly difficult to design and enforcement decisions might come too late to provide any effective remedy or to restore competition. In other words, enforcement may be neither effective nor timely enough to prevent competition from disappearing. However, there is a risk that a blanket ban on widely defined categories of self-preferencing may have a negative impact on innovation and prevent outcomes that are in the interests of consumers, such as the development of new products. This may be especially true in the case of the DMA due to its lack of flexibility to consider potential evidence that gatekeepers’ self-preferencing activities have a strong efficiency rationale and adjust its prohibitions accordingly. In contrast, the CMA’s proposed regulatory regime for digital markets allows for a more flexible approach and is expected to allow the regulator to take on board potential efficiencies and impose more targeted remedies.

There is still some flexibility embedded in the text of the DMA, however. The behavioural obligations specified in Article (6) (as opposed to Articles (5) and (7)) are considered ‘susceptible of being further specified’, and Article (8) states that gatekeepers may request a regulatory dialogue (at the EC’s discretion) regarding their implementation. Further, Article (12) allows the EC to update gatekeepers’ obligations. Whether this flexibility will help with the risk of over-enforcement and the impact that the DMA will have on competitive dynamics and innovation in digital markets more generally will be clearer in the years to come.


In this chapter we have reviewed the key developments relating to self-­preferencing conduct. Specifically, we have considered the Google Shopping decision and the subsequent judgment by the GC, as well as the decisions in three recent investigations by the national authorities. We then explained the broader considerations underlying the theories of harm related to self-preferencing. Finally, we discussed the regulatory approach adopted in the DMA with respect to self-preferencing, and contrasted its per se approach against the effects-based approach endorsed by EU competition law.

Self-preferencing went from being a rarely heard concept to a key issue in any discussion concerning competition in digital markets. The GC has developed new case law and opened the door for competition authorities to develop cases with self-preferencing as a stand-alone abuse. The DMA has also placed self-preferencing in the limelight, with potentially highly significant implications for the business models of dominant digital platforms, as well as those of their competitors. We expect that the enforcement and regulation of self-preferencing will remain a hotly debated topic in the years to come, as the EC will develop new cases with wind in its sails from the GC’s Google Shopping judgment, and newly proposed regulation will come into effect and be tested in practice. It remains to be seen how open the authorities will be to considering arguments and evidence regarding the efficiencies associated with self-preferencing behaviour.


1 Matt Hunt is a managing director, Safter Burak Darbaz is a senior vice president and Robert Scherf is a senior vice president at AlixPartners UK LLP.

2 Commission Decision of 27.6.2017 – Case AT.39740 (EC Google Shopping Decision).

3 Not all such conduct was labelled as self-preferencing, however. For example, in a recent decision concerning a dispute between Google and Enel where the conduct under investigation was Google’s refusal to allow Enel’s electric car recharging app (Juicepass) on the Google Android Auto platform, the Italian Competition Authority framed its decision as an outright refusal to supply abuse, even though it noted that Google’s conduct had the consequence of favouring its own Google Maps app. See the ICA press release, case A529 (

4 10th Amendment to the German Act against Restraint on Competition, Section 19(a), Para. 2.1.

5 ‘A new pro-competition regime for digital markets’ (July 2021), Presented to the UK Parliament by the Secretary of State for Digital, Culture, Media & Sport and the Secretary of State for Business, Energy and Industrial Strategy, Para. 106.

6 Article 6(5) of the agreed text of the DMA, dated 11 May 2022.

7 Google appealed the GC’s judgment at the European Court of Justice on January 2022.

8 CSSs are platforms that allow users to search for products to compare their prices and characteristics across different online retailers, and possibly provide links to such retailers. Google’s CSS was named ‘Froogle’, ‘Google Product Search’ and ‘Google Shopping’ at various points in time. See the EC Google Shopping decision, Paras. 28, 31, 192.

9 Note that the EC does not use the term ‘self-preferencing’ explicitly but refers to ‘more favourable positioning and display by Google, in its general search results page, of its own comparison service compared to competing comparison shopping services’. See the EC Google Shopping decision, Para. 2.

10 Importantly, EC rejected Google’s claim that CSSs and online merchant platforms (such as Amazon Marketplace and eBay) were in the same market. See the EC Google Shopping decision, Para. 246.

11 EC Google Shopping Decision, Paras. 344, 390, 395.

12 EC Google Shopping Decision, Para. 342.

13 EC Google Shopping Decision, Para. 700.

14 The only part of the EC’s decision annulled by the GC was the finding of possible anticompetitive effects of Google’s conduct on the national general search services markets. The GC argued that the evidence relied upon by the EC was too imprecise to show that there were even potential anticompetitive effects in general search services markets attributable to the conduct in question. See Judgment of the General Court of 10 November 2021 – T-612/17 (GC Google Shopping judgment), Para. 457.

15 The GC states that leveraging is a generic term describing practices taking place in one market and having an impact on another market (see GC Google Shopping judgment, Para. 163). Also note that the GC does not use the term ‘self-preferencing’, but rather the term ‘internal discrimination’ (see GC Google Shopping judgment, Para. 237).

16 GC Google Shopping judgment, Para. 232.

17 GC Google Shopping judgment, Paras. 163–164.

18 See Deutscher, E (2021) – Google Shopping and the Quest for a Legal Test for Self-Preferencing Under Article 102 TFEU, European Papers, 6(3), pp. 1345–1361.

19 Notwithstanding this, the GC also notes that Google’s general search ‘has characteristics akin to those of an essential facility’. See GC Google Shopping judgment, Para. 224.

20 GC Google Shopping judgment, Para. 438.

21 GC Google Shopping judgment, Paras. 195, 437.

22 GC Google Shopping judgment, Para. 382.

24 GC Google Shopping judgment, Para. 188.

25 Case AT.40703 – Amazon - BuyBox.

26 Case AT.40670 – Google Adtech and Data-related practices.

27 This is the case unless the website has an agreement to sell its ad inventory to advertisers directly. If not, the intermediation works as follows: first, the publisher notifies its ad server about the available ad slots. The publisher’s ad server then requests offers from multiple ad space sales platforms and evaluates them to determine the winner and notifies the winning advertiser’s ad server, which serves the ad to the publisher website. Each ad space sales platform in turn requests, receives, and evaluates offers from multiple demand-side ad space purchase platforms (also called a demand-side platform, or ‘DSP’). These DSPs make offers by evaluating their participant advertisers’ bids contingent on their pre-sets and the available information on the visitor. Platforms at each side of the chain (i.e., SSPs and DSPs) compete against each other on the offers they send (by subtracting a commission from the winning bid).

28 The EC’s analogous investigation into Google’s ad tech and data-related practices (AT.40670) brings the demand-side (i.e., advertiser side) of the intermediation chain in the picture and seems to be emphasising YouTube ads. In its press release, the EC stated that it would be focusing on, inter alia, the obligation to use Google’s DSPs DV360 or Google Ads when purchasing ads in YouTube, the obligation to use Google’s publisher ad server Google Ad Manager (which was called Google DFP prior to 2018) when serving display ads in YouTube, and the apparent favouring of Google’s AdX by its own DSPs and vice versa.

29 See Autorite de la Concurrence Decision 21-D-11 of 7 June 2021 regarding practices implemented in the online advertising sector (FCA Google AdTech Decision), Para. 346.

30 ibid., Paras. 374–381.

31 The way in which DFP did this differed over the review period. For example, the FCA established that DFP only allowed ADX to submit real-time bids, whereas rival SSPs could only participate with estimated bids. FCA also established that Google configured its publisher ad server DFP to create an informational asymmetry favouring AdX, allowing AdX to consistently outbid rival SSPs by having a ‘last look’ advantage until late 2019. See FCA Google AdTech decision, Paras. 105, 180, 194.

32 See L’Autorita Garante della Concorrenza e Del Mercato Decision No. 29925 – Amazon FBA, 30 November 2021 (ICA Amazon FBA decision), Para. 680. This market definition excludes online retailers that directly sell to consumers through their inventories.

33 This condition was changed in 2021 to allow merchants using logistics service providers approved by Amazon as part of the Seller-fulfilled Prime programme (SFP). The ICA, however, found that the SFP programme was not able to end the contested conduct, mainly because the participation of independent logistics service providers to the programme did not depend on predefined, objective and monitorable quality standards, and Amazon interfered excessively in the contractual agreements between merchants and SFP-qualified logistics service providers. See ICA Amazon FBA decision, Paras. 786–789.

34 ICA Amazon FBA decision, Paras. 762, 811.

35 IICA Amazon FBA decision, Paras. 774–778.

36 ICA found that selling in other online marketplaces would require merchants who purchase Amazon’s FBA service to either replicate their logistics cost, or purchase Amazon FBA’s expensive multi-channel management services. See ICA Amazon FBA decision, Para. 702.

37 The ICA obliged Amazon to establish a system where merchants receive sales and visibility benefits according to uniform and non-discriminatory criteria in line with the level of service provided to Prime customers (i.e., not explicitly conditioning on the choice of logistics service providers), mainly by modifying the structure of its SFP programme. The ICA also ordered Amazon to refrain from any form of intermediation of the relationship between merchants and independent logistics service providers. See ICA Amazon FBA decision, Paras. 890–902.

38 CMA Online Platforms and Digital Advertising Market Study (1 July 2020), Paras. 5.322-5.326.

39 The EC’s investigation into Google’s ad tech and data-related practices (AT.40670) is also looking into Google’s Privacy Sandbox proposals.

40 CMA Decision to Accept Commitments Offered by Google in Relation to its Privacy Sandbox Proposal, Case Number 50972, 11 February 2022 (CMA Google Privacy Sandbox decision), fn. 6.

41 The third concern the CMA expressed is the possible imposition of unfair terms to Google Chrome users, if they are deprived of the choice to adjust the level of privacy and ad targeting in line with their preferences. See CMA Google Privacy Sandbox decision, Para. 3.83.

42 CMA Google Privacy Sandbox decision, Para. 3.39.

43 Open market advertising refers to the process of buying and selling ad space in the open market via ad tech intermediaries. The relevant markets in this case are the same markets considered in the FCA investigation.

44 CMA Google Privacy Sandbox decision, Para. 3.39.

45 The accepted commitments include behavioural restrictions to ensure that Google does not engage in self-preferencing and does not gain an advantage over rivals when third-party cookies are removed, as well a commitment to involve the CMA and the Information Commissioner’s Office in the shaping of the proposals. See CMA Google Privacy Sandbox decision, Para. 5.64.

46 Case AT.40437 – Apple – App Store Practices (music streaming).

47 Polish competition authority UOKiK initiated its investigation on 13 December 2021. Germany’s Bundeskartellamt’s investigation was launched more recently on 14 June 2022. In addition, the French authority, in response to complaints by advertisers and a French startup lobby, considered but declined to block Apple’s launch of ATT but stated that it would continue investigating it.

48 Case AT.40452 – Apple – Mobile payments.

49 Case AT.40684 – Facebook leveraging.

50 This is perhaps best highlighted by the example of several cloud providers who have relied on the label self-preferencing in their recent complaints against Microsoft’s practice of bundling its cloud service OneDrive with its operating system Windows.

51 Input and customer foreclosure are competition concerns that are often considered as potential concerns in relation to vertical mergers. Input foreclosure refers to the situation where the upstream entity of the vertically integrated firm restricts access to products or services that it would supply to downstream rivals absent the merger, causing restriction of competition in the downstream market. Customer foreclosure refers to the situation where the downstream entity of the vertically integrated firm restricts the purchase of inputs from rival upstream firms, thereby weakening them and distorting upstream competition.

52 Integrated logistics operators are those providing services at both upstream (e.g., warehouse management) and downstream (e.g. delivery, collection for returns) levels of the logistics supply chain.

53 ICA Amazon FBA decision, Para. 806.

54 ibid., Para. 702. The ICA also highlighted that orders from other platforms fulfilled by Amazon FBA’s multi-channel service are packaged with Amazon branding, which could have the effect of creating consumer confusion and steering consumers ordering from other platforms back to Amazon, reducing the profitability of a multi-homing strategy of listing its products in more than one marketplace. See ICA Amazon FBA decision, Para. 836.

55 ibid., Paras. 218–225.

56 According to the FCA, this is because Google submits most of the bids from its own DSP (Google Ads) exclusively to its own SSP Google AdX, and because most advertisers engage in single-homing behaviour and only use Google Ads when participating in auctions. See FCA Google AdTech decision, Paras. 227–228.

57 EC Google Shopping decision, Paras. 593–594.

58 ibid., Paras. 595–596.

59 ibid., Paras. 597–599.

60 ICA Amazon FBA decision, Para. 707.

61 ibid, Para. 724.

62 ibid., Para. 802.

63 For example, the ICA highlights complaints from third-party merchants regarding high storage costs (in particular for lower turnover goods – ibid., Para. 325) and points out to claims by some third-party logistics operators that they could offer lower storage fees (ibid., Para. 351). The ICA highlights complaints about high and unpredictable inbound shipping costs (ibid., 328), and also points out to how lack of packaging customisation in Amazon FBA leads to duplication of warehouse costs for some merchants (ibid. Para. 327).

64 The ICA points out to a survey showing 77 per cent of current FBA retailers saying that they would keep using Amazon FBA even if there were alternative and cheaper logistics service providers (ibid., Para. 319).

65 Ibid., Paras. 810–811.

66 However, the ICA also notes that Amazon can use its bargaining power vis-à-vis couriers to secure cheaper delivery prices when it is outsourcing FBA delivery (ibid., Para 351), which would be an offsetting effect.

67 Ibid, Paras. 844–846.

68 FCA Google AdTech decision, Paras. 326–328.

69 ibid., Paras. 391–395.

70 ibid., Para. 450.

71 Separately, the CMA highlights the potential harm to broader society as a result of potential deterioration in high-quality and plural news content. See CMA Online Platforms and Digital Advertising Market Study (1 July 2020), Para. 6.39.

72 Foreclosure does not necessarily mean strict exclusion. It could also mean imposing a restriction on the ability to compete effectively, eg, by raising rivals’ costs – see Salop, S C and Scheffman, D T (1983) – Raising Rivals’ Costs: Recent Advances in the Theory of Industrial Structure, American Economic Review.

73 For example, in Google Shopping, the EC conducted a detailed set of analyses on the evolution of traffic from Google SERP to CSSs, as well as whether the traffic lost by independent CSSs as a result of Google’s conduct was not effectively replaceable from other sources (EC Google Shopping decision, Paras. 539–588).

74 Some researchers have suggested to use natural or controlled experiments to test whether self-preferencing increases consumer welfare. See, eg, Edelman and Lai (2016) who use a natural experiment to study whether Google’s prominent placement of its Flight Services above organic search result increased clicks on sponsored search advertising compared to clicks on organic search results; see also Luca et al. (2015) who use randomised controlled trials to identify whether consumers prefer search results from Google’s ‘specialised search’ over those generated by its organic algorithm.

75 For example, in Amazon FBA, the ICA argued that Amazon’s comparison of delivery speed between orders fulfilled via FBA and by all other merchants not using FBA was irrelevant since it reflected a comparison of FBA performance against the average of all other logistics operators. The ICA stated that there was no technical link between the conduct and the claimed efficiencies, and that Amazon allowing other logistics service operators into the newly introduced SFP programme showed the existence of alternative logistics operators that can comply with the standards set by Amazon. See ICA Amazon FBA Decision, paras. 717, 720, 722.

76 The judgment includes a detailed assessment of the various alternative ways proposed by Streetmap.EU’s expert and agrees with Google regarding their inferiority, impracticality, or disproportionality in terms of costs. See England and Wales High Court Google Maps Judgment of 12 February 2016 (EWHC 253), Paras. 151–176.

77 Regulation (EU) 2022/1925 of the European Parliament and of the Council of 14 September 2022 on contestable and fair markets in the digital sector and amending Directives (EU) 2019/1937 and (EU) 2020/1828.

78 Designation of gatekeeper status are reserved to providers of core platform services (defined in Article 2(2) of the DMA) who satisfy three criteria: having a ‘significant impact on the internal market’, constituting an ‘important gateway for business users to reach end users’, and enjoying (or will be enjoying in the near future) ‘an entrenched and durable position’. These criteria are subject to presumptions of satisfaction of quantitative thresholds specified in Article 3(2) of the DMA.

79 CMA Online Platforms and Digital Advertising Market Study (1 July 2020), Paras. 5.334–5.340.

80 Cabral, L. et al. (2021) – The EU Digital Markets Act: A Report from a Panel of Economic Experts, p. 13.

81 Cremer, J, de Montjoye, Y-A., Schweitzer, H. (2019) – Competition Policy for the Digital Era, pp. 66, 69.

82 GC Google Shopping judgment, Para. 577.

83 An example for exception is provided in Article 6(4), which obliges gatekeepers to allow third-party apps and app stores to be installed on its operating system, grants an exception by allowing the gatekeeper to implement proportionate measures (possibly including imposing constraints on installations or limiting inter-operability) if installation would threaten the integrity of the hardware or the operating system.

84 Recital (32) of the DMA.

85 Arrow, K (1962) – Economic Welfare and the Allocation of Resources for Invention, in: The Rate and Direction of Inventive Activity: Economic and Social Factors, pp. 609–625. Princeton, NJ: Princeton University Press.

86 This view is largely based on Gilbert and Newbery (1982) (Gilbert, J R and Newbery, D M G (1982) – Preemptive patenting and the persistence of monopoly, The American Economic Review, 72(3), pp. 514–526).

87 Schumpeter, J A (1942) – Capitalism, Socialism and Democracy, Harper & Brothers (New York).

88 One of the most cited papers in the economics literature on innovation, Aghion et al. (2005), finds an increasing relationship between innovation and competition at low levels of competitive intensity, turning into a decreasing relationship at high levels of competitive intensity (Aghion, P, Bloom, N, Blundell, R, Griffith, R and Howitt, P (2005) – Competition and Innovation: An Inverted-U Relation, Quarterly Journal of Economics, 120(2), pp. 701–728).

89 De Corniere & Taylor (2019) specify a theoretical model where they show that favouring of its own ancillary products or services in rankings by an intermediary providing search services reduces consumer welfare if the ancillary products or services compete in price, but it can increase or decrease consumer welfare if they compete in quality, depending on the trade-off between the preference mismatch induced by the ranking bias and the magnitude of quality differences between products or services (De Corniere, A and Taylor, G (2019) – A Model of Biased Intermediation, RAND Journal of Economics, 50(4), pp. 854–882).

90 Hagiu et al. (2022) study a stylised economic model where an intermediary that runs a digital marketplace for third-party seller can additionally act as a retailer itself, and potentially self-discriminate in two ways: (1) imitating third-party sellers’ superior products without incurring costs (which they argue captures leveraging data advantage); or (2) introducing ranking bias. The authors show that banning imitation will never decrease consumer welfare or total welfare but banning ranking bias may result in the platform reverting to a pure retailer model, leading to a decrease in total welfare without an increase in consumer welfare (Hagiu, A, Teh, T-H., Wright, J (2022) – Should Platforms Be Allowed to Sell on Their Own Marketplaces?, RAND Journal of Economics, 53(2), pp. 297–237).

91 Padilla et al. (2022) study a theoretical model where a device seller can exclude third-party firms’ access to its app store. The authors show that even if the device seller has an inferior app competing with a third-party app, it could have an incentive to exclude the third-party app (to the detriment of consumers and at the expense of reducing the value of its device to consumers) if demand growth for the device is slow. On the other hand, their results also imply that the device seller’s incentive to exclude will be stronger if the quality of its own app is higher (Padilla, J, Perkins, J, Piccolo, S (2022) – Self-Preferencing in Markets with Vertically Integrated Gatekeeper Platforms, The Journal of Industrial Economics, 70(2), pp. 371–395).

Unlock unlimited access to all Global Competition Review content