The rhyme and reason of gatekeeper designation under the Digital Markets Act

Abstract

We present a comprehensive account of the first 34 designation decisions under the Digital Markets Act (DMA). We first explain how the Commission delineates and classifies core platform services (CPSs) for assessment and how it explains the rationale for finding several gatekeepers in certain CPS segments. We then turn to how the Commission determines gatekeeper status and the criteria that may be used to contest gatekeeper status. Having shown how the decisional practice and the first General Court judgment add substance to the DMA text, we discuss what designation and non-designation decisions tell us about the nature of gatekeeper power.

1. INTRODUCTION: DESIGNATION CRITERIA, PROCEDURES, AND OUTCOMES

The Digital Markets Act (DMA) applies to firms that operate a core platform service (CPS) and have been designated with gatekeeper status in the provision of that service.¹ There are 10 CPSs, including online intermediation services (eg app stores), search engines, social networks, operating systems (OSs), and web browsers.² Gatekeeper status is conditioned on three criteria: (a) having a significant impact on the internal market; (b) serving as an important gateway for business users to reach end-users; and (c) enjoying an entrenched and durable position.³ These qualitative criteria are presumed to be fulfilled when a firm meets certain quantitative criteria, respectively:

• achieving European Union (EU) revenue of at least €7.5 billion in each of the last 3 financial years or having an average market capitalization/fair market value of at least €75 billion in the last financial year, and providing the same CPS in at least three Member States⁴;

• providing a CPS that has at least 45 million monthly active end-users and 10,000 yearly active business users in the EU in the last financial year⁵; and

• meeting the threshold of (b) in each of the last 3 financial years.⁶

As the General Court has explained, ‘the DMA has established a specific regulatory framework governing the designation of gatekeepers, which is characterised by features that are particular to it.’⁷ Potential gatekeepers must notify the European Commission within 2 months when one or more of their CPSs meet the above thresholds.⁸ The Commission then proceeds to designate (or not) the CPS-providing firm within 45 working days.⁹ It completed a first round of designations in September 2023, which resulted in six gatekeepers—Alphabet (Google), Amazon, Apple, ByteDance (TikTok), Meta (Facebook), and Microsoft—for a total of 22 CPSs.¹⁰ Apple was found to be a gatekeeper in another CPS in April 2024,¹¹ and a seventh gatekeeper, Booking, was designated in May 2024.¹² (For an overview, see Figure 1.)

The gatekeeper designation process has been called ‘quick but not dirty’: ‘quick’, in view of the short designation timeline; but ‘not dirty’, as CPS-providing firms can rebut the gatekeeper presumption.¹³ Firms do so by showing that, despite meeting the quantitative thresholds, they do not fulfil the qualitative criteria.¹⁴ Some gatekeepers have been successful in not getting their CPSs designated (crossed out in Figure 1). The other way around, the Commission can also designate gatekeepers that meet the qualitative but not the quantitative criteria.¹⁵

The procedures that deviate from a simple reliance on quantitative thresholds are meant to reduce errors. In this context, the designation of a CPS that does not actually constitute an entrenched, important gateway with a significant impact on the internal market is a type I error (false positive); the non-designation of a CPS that does constitute such a gateway is a type II error (false negative). The rebuttal procedure and the purely qualitative designation process can, respectively, avoid type I errors and type II errors. Doing so often requires a closer look, which the DMA organizes through market investigations (shaded grey in Figure 1).¹⁶

Even after designation, the DMA has mechanisms to correct errors or to prevent them from coming into being. First, the Commission reviews whether gatekeepers still meet the qualitative criteria at least every 3 years.¹⁷ Further, the Commission can, at any moment, upon its own initiative or request, amend or repeal its designation decisions when there is a substantial change in the facts on which they were based.¹⁸ This allows a designated gatekeeper to bring up any market developments that call for a reassessment of the earlier designation decision. To avoid errors not in the application of the DMA but in the instrument itself, the Commission can also, after a market investigation, propose that the European legislators amend the list of CPSs.¹⁹ Finally, the Commission’s designation decisions are subject to review by the Court of Justice of the EU.²⁰

The Commission has now largely completed its designation exercise, and the General Court (‘Court’) handed down its first appeal judgment.²¹ Marginal changes notwithstanding, we expect the current list of gatekeepers and CPSs to remain stable over the coming years. This makes it a good time to take stock of how the Commission carried out the designation process and where it has landed. In the remainder of this article, we therefore study the rhyme and reason of gatekeeper designation by reviewing all implementing decisions.²² First, we address preliminary questions about the delineation and classification of CPSs (Section 2). Then, we study the rhyme of designation decisions (Section 3) and non-designation decisions (Section 4), laying bare the internal mechanics of the process. Finally, we turn to the reason for designation, evaluating the procedure and its outcome in light of the DMA’s logic and goals (Section 5).

2. CORE PLATFORM SERVICES

By selecting certain CPSs as the object of regulation, it was thought that one could eschew market definition exercises. This is true up to a point because in some of the decisions the Commission has had to delineate the CPS (‘Core platform service delineation’ section). The Commission also had to classify ambiguous services as either this or that CPS (‘Core platform service classification’ section). And the designation of multiple gatekeepers in a single CPS category throws up questions regarding the essence of the gatekeeper concept (‘Multiple gatekeepers per core platform service’ section).

Core platform service delineation

The concept of CPS is based on definitions provided in the DMA or other secondary legislation referred to there.²³ These concepts are applied using the general principle of technological neutrality. This means that the same CPS may be offered via different means or devices (eg mobile phones or desktop computers).²⁴ This is generally followed in the decisions (eg Google Play Store, Google Shopping, Google Search, Chrome and YouTube were all defined as a single CPS irrespective of platform).²⁵ Apple tried to avoid the application of this principle, suggesting that the App Store for the iPad was a different service than that designed for the iPhone,²⁶ but the Commission demurred: the App Store has the same function across all Apple devices and is a single CPS.²⁷ This was confirmed by finding that Apple engages with business users in the same way irrespective of the platform used (eg developer agreements and guidelines) and the support services for end-users are also the same. Moreover, both users have the same account (Apple ID) that works on all devices and an app downloaded on one device may then be used on other devices.

However, there is one instance where the Commission has departed from technological neutrality and focused on the specifics of the devices on which the CPS operates: when delineating Apple’s OSs, it distinguished iPadOS, macOS, watchOS, and tvOS.²⁸ From a technological perspective, the OS is that which controls a specific device, and Apple optimizes each OS for each device category.²⁹ From a business user perspective, developers adapt their apps depending on the OS, and end-users obtain different services from the devices: some activities are almost exclusively carried out on iOS, while others are almost exclusively in the iPadOS.³⁰ To distinguish the mobile and tablet OSs, ‘Apple presented a non-exhaustive table with 30 functionalities that are only available on either iPadOS or iOS, but not on both.’³¹ Evidence was also available about the different uses that consumers make of iPhones and iPads.³²

Curiously, this approach was not applied in the analysis of Alphabet’s OS, where the CPS is the OS for both tablets and mobile devices because business users ‘develop their apps to fit Google Android as a whole and follow largely the same development process for tablets and smartphones’,³³ and Google uses the same app vetting process for both devices.³⁴ This was also Alphabet’s own submission. Two things are worth noting: first, that the Commission is seemingly content with different delineations across the same CPS category; secondly, that in Alphabet, the Commission placed emphasis on how business users develop the app.³⁵ It did not inquire whether business users considered the two OSs as interchangeable. Nor did the Commission consider the end-user side: what if some apps are more frequently used on a smartphone and others on a tablet? Arguably, these issues matter if one is trying to work out whether the CPS is a gatekeeper, as they would help determine if business users and end-users are willing to switch from one device to another.

Sometimes a platform offers more than one service, and this requires some efforts to disaggregate the CPS from other services. For example, when browsing on Google Search, one might land on services provided by Google Shopping. According to the Commission, Google Shopping results are ‘a distinct or additional online intermediation service from Google Search’.³⁶ The matter is particularly complex for advertising because firms offer a wide range of services to advertisers on one side, inventory holders on another, as well as intermediary services.³⁷ In the case of Alphabet, this includes at least eight different services.³⁸ Two services were contested, but the Commission decided to also include these. Google Analytics (which provides behavioural measurement and data on website traffic) was included to the extent that business users have integrated this into their Alphabet online advertising services because performance measurement is ‘an integral part of online advertising services’.³⁹ Likewise, AdSense for Search (a service that displays paid search results on third-party websites) was seen as part of the advertising services and not part of Google’s search CPS.⁴⁰ In Microsoft, however, LinkedIn Marketing Solutions was treated as a separate advertising service from the others that Microsoft offers because it served a different user base.⁴¹

Another instance of (dis)aggregation is Booking, where the Commission found that Booking.com was the relevant online intermediation service for hotels, flights, and car rental services. These are aggregated within a single CPS. The Commission excluded other services that Booking offers (Agoda, Priceline, Kayak, and OpenTable). The rationale is that these services are marketed and managed separately and are used for different purposes (from meta-travel search to restaurant reservations) by both end-users and business users.⁴² Moreover, the Commission explained that Booking.com was a single CPS even if it provides multiple services on the same platform (eg hotels and flights). The reason given for this is that businesses and end-users rely on Booking.com’s various services for the same purpose: ‘to intermediate the distribution of travel services available on it’.⁴³ Moreover, the Commission observed that a service like Rentalcars.com, which is available via Booking.com, was also part of the CPS because it is offered to users ‘as a conduit to the same travel services offering provided through the Booking.com online interface’.⁴⁴ This approach is in line with the Commission’s findings in Booking/eTraveli, where it considered that because of (among others) consumer inertia, a consumer booking a flight on a platform offering accommodation services would then likely use that same platform to find a hotel. In other words, the various services offered by Booking.com are part of a multi-product ecosystem: a ‘wide range offer of services that cover multiple facets of the travel experience’.⁴⁵

The same conclusion, that a gatekeeper may be designated even if it only has that status in a single CPS, is also found in the Amazon decision.⁴⁶ The firm provides a platform that allows business users to offer goods and services (Amazon Marketplace), its own retail services (Amazon Retail), a delivery system that is available to business users selling via Amazon (Fulfilment by Amazon), and a special subscription service (Amazon Prime) that gives users access to Prime Video, Amazon Music as well as free and fast delivery. According to the Commission, this ecosystem contains one CPS: Amazon Marketplace. The other services were not considered part of the CPS. Insofar as Amazon Retail and Fulfilment by Amazon are concerned, these are distinct services that should not be considered together with Amazon Marketplace. Nor are these two services considered online intermediation services because they do not allow business users to offer goods or services to consumers. The Marketplace is thus a unique, self-standing gatekeeper that provides a single service.

Rather than thinking of the task of CPS delineation as being about finding an ecosystem, the key point for assessing the question of what services constitute a CPS requires the identification of criteria to ensure the exercise is predictable and suitable to the task at hand. The starting point in the designation decisions is the precise definition of each CPS, found either in the DMA or other secondary legislation referred to in the DMA. The second step is the application of the principle of device neutrality. The only time this has not been applied is with respect to OSs, where the Commission also considered the purpose of use of the CPS by the end-user, business user, or both, and the marketing efforts of the firm with respect to the relevant service(s). The third step is an identification of which service(s) form part of the CPS by reference to how business users and end users interact with these services.

Core platform service classification

Delineation is about the boundaries of services within a CPS category. Classification disputes turn on whether a service is a CPS at all, or if so, in which CPS category it fits best.

First, a potential gatekeeper may argue that its service does not constitute a CPS.⁴⁷ According to Microsoft, for example, its email service Outlook.com did not constitute a number-independent interpersonal communications service (NIICS) under the DMA because it was intended only for personal use and had no business users.⁴⁸ Emails travel over the open Internet, so it is not Microsoft intermediating between the two user groups. The Commission disagreed because, notwithstanding the service’s terms and conditions, users do use Outlook.com to send commercial and professional emails.⁴⁹

Secondly, the potential gatekeeper and the Commission may disagree about the classification of a service within this or that CPS category. ByteDance notified TikTok as a video-sharing service, that is a service-providing programmes or user-generated videos to inform, entertain, or educate.⁵⁰ The Commission did not deny that TikTok met that definition but argued it should be classified as a social network, that is ‘a platform that enables end users to connect and communicate with each other, share content and discover other users and content …, in particular, via chats, posts, videos and recommendations’.⁵¹ As TikTok met both definitions, the question was how to choose between them. This makes a difference as social networks have heavier obligations.⁵²

ByteDance argued that what matters is the service’s primary purpose and core function. It recognized that TikTok offered some social networking features (likes, comments, shares), but its service was not built around a social graph or the need to identify friends; rather, its primary purpose was to serve end-users short-form video content based on their interests.⁵³

The Commission instead decided that the classification as a social network ‘best reflects the characteristics and breadth of TikTok’s features and functionalities’.⁵⁴ It considered TikTok a fully-fledged social network, of which video-sharing was just one feature.⁵⁵ The totality of its features, including the video-related ones, enables users to engage in ‘shared activities and experiences’ and ‘foster a sense of community and social interactions’.⁵⁶ It was not important that TikTok users do not necessarily know the users they interact with.⁵⁷

ByteDance referred to market definition precedents from competition cases to support its classification as a video-sharing service. These precedents were not mentioned in the Commission decision, but one can think of the 2022 Bundeskartellamt decision, where the authority held that content services such as TikTok are ‘not part of the market for social networks’.⁵⁸ The Commission was, however, unwilling to consider competition law precedent in the CPS classification exercise.⁵⁹

A related issue of classification is whether a gatekeeper could provide two CPSs in the same CPS category and what criteria apply to distinguish them. As can be seen from Figure 1, Meta has two social network CPSs and two NIICSs. The criteria for delineation rest on the purpose for which the service is used by end users, business users, or both.⁶⁰ Applied to social networks, two sets of indicators were used. The first indicator is the manner in which these services are offered by Meta: they are branded differently, marketed as distinct services, accessible via different software applications, and have distinct user interfaces. Users may opt into a system that creates a single account and links the two services, but they are offered as distinct services.⁶¹ The second indicator is the purpose for which business users and end-users take advantage of these two services. Instagram was seen as primarily a site to share images, while the social interactions facilitated by Facebook are wider and more features are available to end-users.⁶² Facebook end-users tend to be older than Instagram end-users, too.⁶³ Likewise, business users tend to post images on Instagram but provide a wider range of information on Facebook.⁶⁴ This was confirmed by advice Meta offers to business users, directing them to select which of the two apps is best for their business.⁶⁵

When it came to distinguishing between WhatsApp and Facebook Messenger, the decision focused more on explaining why Messenger was not an ancillary service to Facebook but a stand-alone service which ‘can be used, and is effectively used, irrespective of whether the end users … also use the Facebook online social networking service’.⁶⁶ No explanation was then provided as to why it is a different service to WhatsApp. The Commission notes that it is the second most widely used messaging service after WhatsApp and then uses this to suggest that this is evidence reinforcing contestability and fairness in the NIICS market because gatekeepers provide these services ‘as part of their platform ecosystem’, which raises entry barriers for rival NIICSs.⁶⁷ This may be true, but it would have been helpful to explain the ways in which the two services are distinct; for example, they are at present not interoperable.

Multiple gatekeepers per core platform service

As may be seen in Figure 1, more than one gatekeeper was designated for some CPSs. One might consider that the concept of a gatekeeper implies that there is only one firm across whose gates, to continue the metaphor, a business user must pass to reach an end-user. As soon as there is an alternative gate, then there is competition. The presence of multiple gatekeepers can be explained in the following way.⁶⁸

When it comes to OSs, app stores, and browsers, end-users are largely locked into one of the two major providers (Google or Apple). These two firms furnish an ecosystem comprising a variety of services, and business users have to make their services available via both ecosystems. For example, in Apple iPadOS, the Commission explained that developers find it costly to multi-home and tend to make their apps available to a maximum of two OSs, and ‘developers of the most popular applications are usually present on both Android and iPadOS’.⁶⁹ This means that the two CPSs are not substitutes. Competition authorities have similarly distinguished the ecosystems of Google and Apple, putting them in different markets.⁷⁰

When it comes to intermediation services,⁷¹  Figure 1 further classifies these to capture their differentiation. Online travel agencies and online maps are intermediation services that attract different business users, serve different end-user requests, and can easily be distinguished from each other and the rest of the services. App stores are another type of intermediation service,⁷² and here, as discussed above, end-users are locked into the hardware they bought, and developers normally have to be present in both systems.

However, in the social network CPS category, there are four players. Using the principles above, LinkedIn may be distinguished because it is oriented to professional use.⁷³ But the other three could be grouped in a sub-sector of services where there is competition. ByteDance contested a finding that it is a gatekeeper, given its rivalry with two other well-established rivals (Instagram and Facebook). However, as discussed above,⁷⁴ the Commission considered that Facebook and Instagram were distinct services because of the manner in which they were branded and marketed, and the manner in which end-users and business users engaged with the two platforms differed.⁷⁵ In turn, the Commission found that TikTok could also be distinguished and that content creators who multi-home may do so out of necessity: to reach different kinds of end-users or to increase their audience.⁷⁶ The Court confirmed this. The following is worth noting: ByteDance observed that business users and end-users tend to multi-home. The Court agreed this was a relevant factor, but that it is not of itself dispositive.⁷⁷ It explained that content creators may multi-home ‘to increase their exposure to as wide an audience as possible’.⁷⁸ Others might single-home and then ‘face switching costs’ to migrate to another platform.⁷⁹ In both scenarios, it is plausible to conclude that the content creator depends on the gatekeeper even if there is some competition between gatekeepers on the other side of the market. Likewise, with end-users, the Court held that multi-homing did not per se establish that they are not locked into the TikTok app; for example, a category of users might exhibit stronger engagement with this platform than others.⁸⁰

Roughly the same reasoning applies to marketplaces. Amazon, Facebook, and Google (Shopping) serve slightly different audiences. Amazon provides a typical marketplace for professional sellers, while Facebook Marketplace is reserved for (second-hand) goods sold by non-professional sellers.⁸¹ Google Shopping is an advertising service that links to professional sellers and other comparison shopping engines.⁸² Nevertheless, there is a degree of competition between these marketplaces. Meta claimed, for example, that it faced competition from other e-commerce platforms. Again, the Commission stated that the mere presence of competition in a CPS category does not preclude a finding of gatekeeper status. On the one hand, Meta is free to provide evidence that it has no power. On the other hand, the presence of some gatekeepers does not call into question the presence of entry barriers for new entrants.⁸³

Finally, when it comes to advertising, it is worth recalling that advertising services are only a CPS when the firm offering these services is also a gatekeeper in one of the other CPSs. This is because advertising happens through that CPS and holding gatekeeper power in (at least) two services creates the incentive to impair contestability and fairness. From this, it follows that the ad tech stacks of Alphabet, Amazon, and Meta are all gatekeepers in their respective closed ecosystem. This also explains why Microsoft’s ad services were not designated: very limited scale of use, revenues significantly smaller than those of the other three gatekeepers, and absence of scale effects within the Microsoft ecosystem (given Bing’s modest share of the market where most of Microsoft’s advertising business is concentrated), modest interest by business users, and a trend towards mobile advertising combined with Microsoft’s limited presence in that segment.⁸⁴

3. THE RHYME OF DESIGNATION

Once CPSs are delineated and classified, there is little controversy in assessing whether the presumption has been met. At times, there is disagreement about how to compute certain numbers (eg identifying business users on Facebook, Instagram, Microsoft’s PC OS and LinkedIn, and end-users for Microsoft’s browser), but the Commission does not pursue this once the numbers are high enough to cross the threshold.⁸⁵ Only if parties contest the presumptive value of their numbers does the Commission carry out a more detailed analysis.

In Section 4, we discuss how the Commission has responded to requests not to designate. Here we discuss the only designation decision that has resulted from a market investigation: Apple’s iPadOS.⁸⁶ Apple did not meet the thresholds in Article 3(2)(b) and (c) in relation to this CPS. The Commission considered a number of indicators listed in Article 3(8) to rule that the CPS should be designated. It is hard to generalize on the basis of one decision, but two features are worth noting.

The first is the comparisons drawn with rival tablets. The Commission compared the number of applications on the iPadOS with those on Android devices (roughly comparable at 1–1.5 million) and with those on the nearest rival, Amazon’s Fire (less than 1 million).⁸⁷ When it came to end-users, it seems the numbers were too low to trigger the presumption but these were growing between 2020 and 2022. This growth ‘illustrates the position of iPadOS as a major platform intermediating between an increasing number of business and end-users’.⁸⁸ This conclusion was bolstered by explaining that these numbers were similar to those of Android (which had already been designated) and much higher than those of Amazon’s Fire OS (which had not been designated).⁸⁹

The second is the analysis of the iPad’s importance for developers: one out of three developers makes 80 per cent of their tablet applications revenue via the iPad,⁹⁰ and generally, iPad end-users spend 2.5 times more on in-app purchases than Android users.⁹¹ More specifically, developers of certain apps (eg games, note-taking, and word processing) consider the iPad an important platform.⁹² In addition, multi-homing by developers is costly because different software is required for each tablet, confirmed the majority responding to the Commission’s market investigation by stating that they make their apps available to a maximum of two OSs.⁹³ Apple’s conduct adds to the lock-in effect by making it easy to adapt an iOS application for the iPadOS.⁹⁴ Moreover, developers cannot afford to code for all tablets and pick the most lucrative ones, which leads to increased end-user demand for these tablets that offer a higher number of apps. These network effects have made the market tip to Apple and Android; other entrants (eg Microsoft and Amazon) have not been able to enter at scale because of a lack of apps on their tablets.⁹⁵ This approach suggests that gatekeeper status is in part assessed by the market position of the firm relative to other competitors. It also shows the importance of analysing the P2B relationship and the extent to which business users are dependent on a platform. We return to this observation in Section 5 as it offers clues about the nature of gatekeeper status.

4. THE RHYME OF NON-DESIGNATION

CPS-providing firms can rebut the gatekeeper presumption. The Commission can accept a rebuttal immediately after the firm reports meeting the quantitative thresholds, as was the case for Samsung’s browser and the email services of Microsoft and Google. This happens when the firm not only manifestly calls into question the presumption but also ‘clearly and comprehensively demonstrate[s] that the requirements of that CPS constituting an important gateway … is not fulfilled’.⁹⁶ In cases where the firm ‘only’ manifestly calls into question the presumption, the Commission opens a market investigation. After investigation, the Commission accepted rebuttals for various Microsoft services, Apple’s iMessage, TikTok Ads, and X’s social network and advertising service. In contrast, the Commission outright rejected ByteDance’s arguments rebutting the gatekeeper presumption for TikTok’s social network, though not for lack of trying on the part of ByteDance.⁹⁷

In this section, we examine the non-designation procedure. We first tackle the evidentiary question: which rebuttal arguments are admissible, who has the burden of adducing them, and what is the standard to accept them (‘Rebuttal burden and standard of proof’ section)? Next, we examine the Commission’s practice of non-designation (‘The practice of non-designation’ section).

Rebuttal burden and standard of proof

The DMA strictly circumscribes the possibility of rebutting the quantitative gatekeeper presumption. The burden of proof is on the CPS-providing firm.⁹⁸ The standard of proof required to get the Commission to open a market investigation is high: the firm must advance sufficiently substantiated arguments that ‘manifestly call into question’ the fulfilment of the qualitative criteria.⁹⁹ The Court has specified that this requires ‘a high degree of plausibility’; mere ‘doubts’ or ‘prima facie’ evidence does not suffice.¹⁰⁰

In a further restriction, a recital limits the rebuttal evidence to elements that ‘directly relate to the quantitative criteria’,

namely the impact of the undertaking providing core platform services on the internal market beyond revenue or market cap, such as its size in absolute terms, and the number of Member States in which it is present; by how much the actual business user and end user numbers exceed the thresholds and the importance of the undertaking’s core platform service considering the overall scale of activities of the respective core platform service; and the number of years for which the thresholds have been met.¹⁰¹

The idea is that rebuttal, like designation, must stick closely to the quantitative criteria. But how closely? The recital contains a confusing mix of ‘namely’, which suggests a closed list, and ‘such as’, which suggests an open-ended character.¹⁰² In ByteDance, the Court sensibly held that the list of elements is non-exhaustive.¹⁰³ Even if arguments must relate directly to the quantitative criteria, the firm may provide a mix of quantitative and qualitative evidence to rebut the presumption.¹⁰⁴ The line is drawn at evidence that is explicitly excluded by the DMA, that is arguments ‘seeking to enter into market definition or to demonstrate efficiencies’.¹⁰⁵

The ByteDance judgment does not appear fully aligned with legislative intent.¹⁰⁶ Originally, the Commission proposed that the qualitative designation procedure and the rebuttal of the quantitative presumption be symmetrical: in both cases, non-quantitative evidence, for example on barriers to entry, was admissible.¹⁰⁷ The legislature did away with this symmetry, adapting the text so that only the qualitative designation procedure referenced digital entry barriers.¹⁰⁸ As we discuss below (‘The practice of non-designation’ section), the Commission did not completely disregard arguments related to entry barriers in rebuttal procedures. However, it is still helpful that the Court held that the evidentiary allowances of the different designation procedures are more symmetrical than what the DMA’s text suggests.

The practice of non-designation

To rebut the quantitative presumption, it suffices to show that one of the Article 3(1) criteria is not fulfilled, so that the CPS-providing firm either (i) does not have a significant impact on the internal market; (ii) does not serve as an important gateway for business users to reach end-users; or (iii) does not enjoy an entrenched and durable position.¹⁰⁹ We discuss each gatekeeper criterion in turn.

Significant impact on the internal market

Fulfilment of the ‘significant impact on the internal market’ criterion was not controversial in most non-designation discussions, as most CPS-providing firms significantly surpassed the EU revenue threshold (€7.5 billion), the market capitalization threshold (€75 billion), or both. It was, however, contentious in ByteDance and X.

ByteDance did not meet the revenue threshold and could not meet the market cap threshold as it is not publicly listed. But its fair market value, the alternative to market cap for non-listed firms, was estimated to significantly exceed €75 billion. ByteDance countered that this is driven by its China-focused business, with its EU revenue significantly below the threshold.

The Commission found it irrelevant that ByteDance’s EU revenue was relatively low,¹¹⁰ which the Court considered an error in law.¹¹¹ The Court held that it is legally possible that meeting only the fair market value threshold, not the revenue threshold, could be an indicator that the firm does not have a significant impact on the internal market.¹¹² This might sound counterintuitive because the two thresholds are alternatives, so they should indicate the same fact. But the Court takes the view that the two thresholds ‘reflect similar, but distinct, situations’.¹¹³ As intimated in a DMA recital, EU turnover shows present ability to monetize users, while high market capitalization reflects future potential.¹¹⁴ However, on the facts, ByteDance did not show how missing out on this element affected its gatekeeper status. As the Commission had shown satisfactorily that other elements showed a significant impact on the internal market, namely ByteDance’s rapid growth in the EU and a fair market value well above the thresholds, the Commission’s error had no impact on the lawfulness of the decision.¹¹⁵

Supporting arguments by ByteDance were equally unsuccessful in disproving its significant impact on the internal market. It pointed to the losses it had been suffering to build out its customer base.¹¹⁶ The Commission found that argument insufficient as, in line with the above recital, its significant fair market value, in itself, ‘reflects a relatively significant potential to monetise … users in the near future’.¹¹⁷ It also shows ByteDance’s ability to access financial markets to further strengthen its position (eg through acquisition).¹¹⁸ Another intriguing argument was that as a firm largely based in China, it faced cultural and regulatory barriers to expansion abroad.¹¹⁹ The Court accepted the argument as a matter of law, but found that ByteDance did not substantiate as a matter of fact how such barriers prevented it from benefitting from its high fair market value.¹²⁰ The existence of such barriers was hard to reconcile with TikTok’s meteoric rise in the EU.¹²¹

The situation of X was different. X’s social network and advertising services are part of X Holdings, which is—together with SpaceX, X.AI, Neuralink, and the Boring Company—operated by Elon Musk.¹²² The Commission found that Musk has ultimate control—meaning, as in the EU Merger Regulation, decisive influence—over these different firms.¹²³ Hence, the entire Musk Group constitutes a single undertaking, which means the fair market value threshold must be assessed at the level of the group.¹²⁴ X argued that the DMA should only apply to undertakings active in the EU digital economy, or those with ‘platform power’, which would exclude some of Musk’s other firms, but to no avail.¹²⁵ As a result, the DMA can apply to diversified groups of which only a part, even a minor part, operates a CPS. As such, the DMA’s scope appears overly broad, even if correction (based on other criteria) is still possible. In the case of X, the Commission did correct course by accepting rebuttals for X’s social network and advertising service (in the latter case, without market investigation).¹²⁶

The fringe cases of ByteDance and X show that the market capitalization/fair market value threshold can be a poor proxy for a CPS’s significant impact on the internal market. First, valuations concern the potential gatekeeper firm as a whole, not any specific CPS, as seen in X. Secondly, valuations are not necessarily based on the potential gatekeeper’s EU business; in other words, they do not necessarily concern the internal market, as shown in ByteDance. The Commission’s answer is that a high valuation reflects a firm’s current financial capacity and shows strong potential to monetize users in the future,¹²⁷ though that will not always be the case. At the end of the day, mistakes are avoided primarily due to the cumulative nature of the gatekeeper criteria, given that the user number thresholds do relate to a specific CPS and its popularity in the EU.

Important gateway for business users to reach end-users

The main rebuttal battleground has been the second qualitative criterion of serving as an important gateway for business users to reach end-users, which is presumed to be fulfilled when the CPS has 45 million end-users and 10,000 business users. Successful rebuttals usually rely on a combination of factors that put those user numbers in perspective.

The primary evidence to rebut that quantitative presumption relates to the relative unimportance of the CPS in question.¹²⁸ The leading argument usually goes that, while the CPS may have over 45 million end-users, other CPSs in the same category boast many more still. Sometimes, for the relative importance of CPSs to become clear, it is useful not to look at monthly active end-users (MAUs), as prescribed by the DMA, but at more granular and representative measures of usage (webpage views, queries, and clicks).

The web browsers of Samsung and Microsoft (Edge) were both saved from designation in large part because their usage pales in comparison to those of others in the same CPS category. Google Chrome and Apple’s Safari receive 60 and 20 per cent of webpage views, while Samsung’s browser and Edge have to make do with less than 5 per cent each.¹²⁹ The trend is also important. Samsung’s browser usage is declining, while Microsoft was unable to significantly increase usage of Edge by integrating its AI assistant, Bing Chat.¹³⁰ As for Microsoft’s Bing search engine, looking at the number of queries and clicks—rather than MAUs—showed the search engine’s ‘very limited scale of activities’.¹³¹ Its share of those queries and clicks is less than 5 per cent, compared to Google’s 90+ per cent.¹³² Reliance on user numbers was greatest in Apple iMessage. The Commission stressed iMessage’s low number of end-users compared to the other designated NIICSs, Meta’s WhatsApp, and Messenger.¹³³ iMessage’s relative unimportance became even clearer when observing its low intensity of usage compared to the other NIICSs.¹³⁴

Having a significantly lower number of end-users than designated gatekeepers in the same CPS category makes it difficult, if not impossible, for another CPS to be an important gateway for business users to reach end-users. As the Commission held with regard to Bing: ‘Limited usage of an online search engine by end users to find businesses … necessarily translates into limited reliance by those business users of online search engines to reach end users.’¹³⁵ Nevertheless, the Commission tends to summarily examine the business user side as well, in a similarly comparative fashion. In Apple iMessage, for example, the Commission focused on the ‘extremely limited user base’ of Messages for Business, Apple’s B2C version of iMessage, especially compared to Meta’s B2C messaging tools.¹³⁶

In comparing user numbers between services within the same CPS category, the Commission comes perilously close to examining market shares, which the DMA explicitly precludes.¹³⁷ The Commission is cautious to refer not to market shares,¹³⁸ but to usage shares (often awkwardly called ‘scale of use’).¹³⁹ However, as some CPS categories can overlap with antitrust markets,¹⁴⁰ one can tolerate the use of the same indicators.

In addition to usage shares, the Commission examined arguments relating to barriers to entry, though sparingly. Apple argued the lack of network effects of iMessage, for example, but the Commission did not pick up on that argument.¹⁴¹ This may be for reasons of procedural efficiency: if an examination of user numbers suffices to accept the rebuttal, there is no need for the market investigation to turn into a more resource-intensive, antitrust-like inquiry.

When the Commission does take a closer look at barriers to entry, it is usually within the context of ecosystems. Just about every rebuttal assessment features a discussion of the CPS-providing firm’s ecosystem.¹⁴² The Commission in this context conceives ecosystems as multi-product groups offered by a single firm, rather than as multi-actor groups.¹⁴³ The focus is on how the (potential) gatekeeper provides a set of complementary products, not so much on how independent complementors contribute to the ecosystem. Based on this conception, CPS-providing firms have been unsuccessful in convincing the Commission that they do not operate an ecosystem. But the Commission has accepted that specific CPSs do not significantly benefit from the ecosystem and the barriers to entry (network effects, data-driven advantages, lock-in) that come with it.

In Samsung, the Commission described how none of Samsung’s other CPSs currently have gatekeeper status. Therefore, those adjacent services ‘do not in any way reinforce the importance of [Samsung’s browser] as a gateway for business users to reach end users’.¹⁴⁴ Microsoft did have an ecosystem that included designated CPSs, but the Commission noted how ‘Bing does not benefit from any significant advantage due to its vertical integration in Microsoft’s platform ecosystem’.¹⁴⁵ Indeed, even within the Windows ecosystem, which affords Bing advantages like default placement, Google is by far the most-used search engine.¹⁴⁶ Dependence on other ecosystems also helps rebut the gatekeeper presumption. The browsers of Samsung and Microsoft run on Google’s Blink browser engine, which shows that the CPS-providing firm is not fully in control of its gateway.¹⁴⁷ The dependence is even stronger for Samsung, as all its devices run on Google Android.¹⁴⁸ Having your key technological architecture controlled by another ecosystem operator is a particularly strong reason to rebut the important gateway criterion.¹⁴⁹

The approach in ByteDance was slightly different. ByteDance argued that, in contrast to Meta, it lacks a wider ecosystem that it could leverage, in particular, to sell advertising. The Commission was largely unwilling to consider the lack of a TikTok ecosystem, holding that the pertinent comparison was between social networking services, not the related advertising services.¹⁵⁰ The Commission’s rejection of the Meta comparison mostly ended ByteDance’s ecosystem argument, even if the Commission added that, in any case, having a multi-product ecosystem is not a strict requirement to be an important gateway and, in fact, ByteDance does operate such an ecosystem (that it has not been very successful in exporting outside of China).¹⁵¹ In rejecting ByteDance’s comparative approach to ecosystems, the Commission avoided arguments that it had accepted in other decisions. Like Samsung, ByteDance had just one potential gatekeeper service. Like Microsoft’s Bing, ByteDance’s social network did not appear to significantly benefit from its wider ecosystem. But while those arguments played a role in Samsung and Microsoft’s non-designation, they did not for ByteDance.

The Court primarily faults ByteDance for the perhaps partial treatment of the ecosystem argument in the designation decision: it was ByteDance that built its argumentation around the premise that it did not have an ecosystem, which it then failed to substantiate.¹⁵² However, the Court also noted that the Commission, by focusing on online advertising, did not fully engage with ByteDance’s arguments.¹⁵³ This did not invalidate the assessment, as the Commission did pick up on the fact that, even without an ecosystem (comparable to Meta), TikTok still grew to 125 million users in a couple of years, reaching half of Meta’s size.¹⁵⁴ Clearly, the lack of an ecosystem does not mean TikTok does not benefit from network effects, which limit contestability.¹⁵⁵ In short, the lack of an ecosystem can help disprove gatekeeper status; it just did not in this case.¹⁵⁶

A final argument to rebut the fulfilment of the ‘important gateway’ criterion relates to interoperability. The email services of Google (Gmail) and Microsoft (Outlook) boast enough users to be designated. In their ‘current configuration’, however, these services enable business users and end-users to effectively exchange messages with users of any other email service.¹⁵⁷ This interoperability goes back to the open standards that email is based on (in particular, the Simple Mail Transfer Protocol or SMTP).¹⁵⁸ As the existence of such open standards, outside of the control of gatekeepers, is unique, this rebuttal ground is unlikely to find wider application.

In sum, successful important gateway rebuttals tend to rest on the cumulative force of a number of indicators. The most important one is the scale of a service compared to the designated ones in its CPS category. For Apple’s iMessage, its small usage share was essentially enough not to be designated. In other decisions, the Commission put weight on the firm’s multi-product ecosystem. It seems impossible to convince the Commission that a potential gatekeeper does not have such an ecosystem, though that would be relevant, but it is possible to show that the service in question does not benefit significantly from that ecosystem, or—even more convincingly—that the service in question depends on the ecosystem of another gatekeeper. Other evidence is available but given less attention.¹⁵⁹

Entrenched and durable position

Most gatekeeper services have been around for some time. Therefore, the criterion of having an entrenched and durable position, presumed when meeting the user number thresholds during the last 3 financial years, was not often in dispute. A recital relates the entrenched and durable position to the DMA’s contestability goal, holding that such a position exists ‘where the contestability of the position of the undertaking providing the core platform service is limited’.¹⁶⁰ ByteDance seized on this recital to argue that, on the contrary, its position is not ‘unassailable’.¹⁶¹ The Commission found this to be a subversion of the recital and instead focused on user numbers: TikTok boasted more than 125 million end-users in 2022 and was still growing rapidly.¹⁶²

In an argument that cuts to the core of the gatekeeper concept, ByteDance suggested that, as a ‘challenger’ to the incumbent ecosystems of Meta and Alphabet, it could not also be a gatekeeper.¹⁶³ That challenger position was not only reflected in its smaller size but also in its conduct, as it facilitates interoperability and multi-homing (eg by allowing cross-posting)—typical challenger behaviour, according to ByteDance. The Commission did not believe the positions of challenger and gatekeeper were mutually exclusive.¹⁶⁴ TikTok had already grown to half the size of Facebook and Instagram over the last 3 years and had a particularly high engagement rate with young users.

5. THE REASON OF GATEKEEPER DESIGNATION

Having surveyed all the Commission’s designation decisions to elicit the how of designation, we now turn to the reason of designation. We start by explaining the approach to the production and assessment of evidence, which places a heavy burden on firms, both at the notification and rebuttal stage, and how this has been checked by the Court in its first DMA judgment (‘Evidentiary approach’ section). We then turn to discuss the delineation and classification approach that has been followed, discussing its rationale and framework (‘Anti-circumvention in delineation and classification’ section). Now that the Commission has designated 24 CPSs with gatekeeper status and has not designated 10 other CPSs, we discuss if we are any closer to understanding what a gatekeeper is (‘What is a gatekeeper?’ section).

Evidentiary approach

The DMA relies on firms making a notification to trigger the proceedings. The evidentiary burden on the Commission is lightened as a result of this. Similarly, it is for the gatekeeper to contest the designation by bringing arguments that are sufficiently substantiated. Reading through all the Commission decisions, the pattern that emerges is that the Commission sets out and responds to each and every counterargument levelled, and in its market investigations gathers data from a number of stakeholders to secure as much information as possible before reaching its decisions. The Court explains that while the burden of proof is often on the firm, this does not require it to take excessive or unrealistic steps; frequently, the gatekeeper will be in the best position to provide evidence and explain it.¹⁶⁵

However, there may be certain circumstances where the Court is overly optimistic. For example, the firm may benchmark itself with other designated gatekeepers to show that it is less significant than others. In doing this, it will have data about how and how much its users engage with it, but it is less likely to have the information to be able to explain why it is less economically significant relative to others. At least at the stage of rebuttal, there is a risk of the DMA setting out a probatio diabolica. (This risk is reduced at the market investigation stage by the Commission’s richer engagement with stakeholders, but that does not matter unless the firm can rebut the presumption in the first place.) One might counter this criticism by indicating that a presumed gatekeeper could use its network of business users to gauge their use of other platforms or even further that this knowledge is so essential for the evolution of the platform’s own business model that the gatekeeper has sufficient incentives to discover data about its performance relative to rivals.

Importantly, neither the Commission nor the Court outright rejects a lot of the arguments made by gatekeepers: it is perfectly plausible to rebut a presumption that one is a gatekeeper by showing that there is no ecosystem, that users multi-home, that there are cultural differences that create entry barriers, or that the gatekeeper is really a challenger. However, the burden of proof is on the firm to substantiate these claims.¹⁶⁶

In this light, one should not read the ByteDance judgment as displaying a deferential approach by the Court. The Court accepts the legislative intention expressed in the DMA (which was not challenged by the applicant) and uses it as a basis for evaluating the arguments. It frequently faults the applicant for not bringing enough evidence to the table to support its claims.¹⁶⁷ Moreover, the Court does not shy away from criticizing the Commission. It does so on three occasions. Most significantly, it finds an error in law when the Commission rejects as irrelevant an argument about the significance of ByteDance not having a sufficiently high EU turnover.¹⁶⁸ In addition, the Court faults the Commission for not interpreting the applicant’s arguments completely when it came to its claims that it did not have an ecosystem,¹⁶⁹ and in a related matter, the Court finds that the Commission did not respect the applicant’s right to be heard because it concluded that ByteDance did have an ecosystem but did not inform the applicant of this.¹⁷⁰  Ex abunanti cautela, while the Court reminds the applicant that the DMA procedures mean that they cannot submit, for the first time before the Court, arguments or evidence that were not submitted during the administrative procedure, it considers the merits of these anyway.¹⁷¹

The takeaway for CPS providers is to prepare possible counterarguments with reliable data about its products and users. This is not to say that the lines drawn by the decisions and the judgment are always clear, as we discuss below.

Anti-circumvention in delineation and classification

The drafters were aware of the risk that gatekeepers might delineate their CPSs creatively to escape the DMA. In the Annex to the DMA, it explains that when counting users of a CPS, the firm cannot draw arbitrary lines saying that users of CPS.it are different from users of CPS.nl, for example. Conversely, it indicates that two CPSs may be different even if they belong to the same category and even if the users are overlapping, provided that they are used for different purposes.¹⁷² Firms may be strategic in delineation in two ways: first, they might try and notify services in narrow segments so that each falls below the threshold; secondly, they might notify services in a wide way so as to avoid certain access or interoperability obligations. In the decisions we have surveyed, we see manifestations of this and of how the Commission has addressed it via its delineation approach.

Perhaps the clearest example of strategic delineation is Meta’s attempt to bundle several services into one.¹⁷³ By claiming that Facebook Messenger is just an ancillary product to its social network and not a stand-alone NIICS, it would avoid the interoperability obligations in Article 7 DMA.¹⁷⁴ The Commission rejected this claim.¹⁷⁵ Likewise, it rejected claims that Facebook and Instagram are the same CPS, that Facebook Dating and Gaming Play are ‘an integral part of its online social networking experience’, and that Meta’s marketplace is not a distinct CPS.¹⁷⁶ Designating these services as a single CPS would have made compliance with Article 5(2)(b) DMA unnecessary, as it would not be combining personal data from two CPSs. Similarly, unsuccessful claims by Meta that online ad services are ‘part of a single ad-supported online social networking service’, or claims by Alphabet that certain specific ad services (eg Google Analytics) are not part of an online advertising service would serve to either eliminate or limit the scope of the obligations relating to ad transparency (Article 5(9) and (10) DMA).¹⁷⁷ Finally, Apple’s unsuccessful claims that it has five app stores (one per type of device) and three distinct web browsers would have meant that it only met the thresholds for the iOS App Store and Safari on iOS, so that the DMA would not have applied to the other devices.¹⁷⁸

Classification can be an equally strategic exercise. Many of the DMA’s obligations are tied to a specific type of CPS. While none of the obligations relate specifically to video-sharing services, for example, some apply to social networks.¹⁷⁹ When anticipating compliance, firms may therefore try to have their CPS classified in an ‘obligation-light’ category. This explains the dispute regarding TikTok’s classification.¹⁸⁰

As we saw above, in engaging with these arguments, the Commission considers product characteristics and the uses made by businesses and consumers. There is perhaps little to distinguish this kind of exercise from the inquiry into the physical characteristics criteria that have been used as the starting points of market definition in competition law cases.¹⁸¹ Moreover, we see similar strategies: just like in antitrust where market definition can be determinative to establish the application of competition law so here we find endeavours to define services widely or narrowly to escape the DMA. However, the Commission remains adamant that the delineation of a CPS ‘has no bearing on the definition of the relevant market for the purpose of applying EU competition rules (and vice versa).’¹⁸² Perhaps the way to explain this is that a delineation for the purposes of applying the DMA should not be taken as a legal indicator of what the relevant product market might be for the purposes of competition law. Methodologically, the two approaches resemble each other, but the legal consequences differ.

What is a gatekeeper?

The DMA’s gatekeeper definition has a formal part (the list of CPSs) and a functional part (the criteria of Article 3).¹⁸³ The qualitative criteria give an idea of a gatekeeper’s features—a significant impact on the internal market, serving as an important gateway, and holding an entrenched and durable position—but those features are not defined; the DMA only provides quantitative thresholds to presume them. The DMA’s recitals provide further clues about a gatekeeper’s key characteristics, but in the end also refer to a ‘combination of features’.¹⁸⁴ It is difficult to ascertain, first, what each gatekeeper characteristic entails, and secondly, which characteristic—if any—defines a gatekeeper. Below, we go over the characteristics that seem to be inherent in the gatekeeper concept and play a significant role both in the text of the DMA and in the designation decisions. The first characteristic, dominance, is taken from competition law. Two other characteristics, single-homing users and ecosystems, have been developed by economists to better understand market power in platform markets. The final characteristic, being a challenger, is the opposite of a gatekeeper.

Dominance

The DMA’s objectives of fairness and contestability are ‘complementary to, but different from’ competition law’s goal of undistorted competition.¹⁸⁵ The DMA was adopted to make up for competition law’s limitations, including that it is ‘limited to certain instances of market power, for example dominance on specific markets’.¹⁸⁶ And yet, it is tempting to see some similarities in analytical framing: the process of delineation and classification is but a basic version of market definition,¹⁸⁷ raising questions about how far the Commission might in future feel compelled to engage in a more economics-based analysis to determine the scope of a CPS with more precision.

However, when it comes to designation, it would seem that the scope of the DMA cannot be determined by competition law’s conception of dominance. On the one hand, the DMA qualifies gatekeepers as having ‘considerable economic power’.¹⁸⁸ Most designations also seem to track the degree of market power, at least in the sense that the absence of it is a strong argument to rebut quantitative designation. The Commission examines the ‘scale of use’ of CPSs and tends not to designate those that show minor usage compared to others in the same category.¹⁸⁹ Having less than 10 per cent usage share, such as Bing, Edge, or iMessage, seems to immunize a firm from designation. Similarly, firms with 10 per cent market share can be certain of their lack of dominance.

On the other hand, the dominance analogy breaks down at slightly higher usage shares. Safari, for example, was designated with a 20 per cent share in the browser market. Moreover, many CPS categories have multiple gatekeepers.¹⁹⁰ ByteDance is designated in the social network CPS category together with the (by user numbers) larger services of Facebook and Instagram. And the Commission has found Facebook dominant in the market for personal social networks.¹⁹¹ In competition law terms, there could not be multiple firms with significant power in a relevant market unless there is a finding of collective dominance.¹⁹²

The absence of dominance does little to insulate a CPS-providing firm from designation. The Court confirmed that the DMA covers gatekeepers ‘who do not necessarily hold a dominant position within the meaning of competition law’.¹⁹³ Two alternatives present themselves. Perhaps, the DMA operates based on a lower threshold of market power, not the substantial market power from Article 102 of the Treaty on the Functioning of the European Union (TFEU). A more convincing explanation is that the DMA has a different conception of economic power. Its conception could be more granular, focusing not on the overall market but on specific user groups (See ‘Single-homing’ section). Or the DMA could go beyond a narrow conception of markets to a broader understanding of power in ecosystems (See ‘Ecosystem’ section).

Single-homing

The gatekeeper concept is built around the idea that there is a single gate through which businesses and end-users must pass to reach each other.¹⁹⁴ The DMA stresses that the gatekeeper concept implies dependence by both businesses and end-users.¹⁹⁵ If there are multiple gates, businesses and end-users are not dependent on any one platform.¹⁹⁶ Therefore, it appears crucial that a gatekeeper is ‘a monopolist when it comes to its user base’.¹⁹⁷ This is in line with economic literature, which has long recognized that single-homing by one user group creates a competitive bottleneck.¹⁹⁸ When users multi-home, in contrast, there is no platform monopoly or dependence.¹⁹⁹ For this reason, scholars have proposed a safe harbour from designation for platforms that face multi-homing on both sides.²⁰⁰

In designation rebuttals, CPS-providing firms often point to their multi-homing users. The Commission’s answer generally goes that ‘the existence of multi-homing is not in itself an indication that a CPS does not constitute an important gateway for business users to reach end users’.²⁰¹ The Court confirmed that single-homing is not a conditio sine qua non for a CPS to be regarded as an important gateway.²⁰² It more generally observed that ‘in order to consider that business users of a CPS “depend” on it in order to reach their end users, it is not necessary for that CPS to be the only channel through which those undertakings can reach those users. It is sufficient for it to be an important channel for that purpose’.²⁰³

The Court added an interesting policy argument to reject single-homing as a key gatekeeper characteristic: if single-homing was key, ‘the consequence would be that no, or virtually no, online social networking service would be regarded as being such a gateway.’²⁰⁴ As the DMA includes social networks as a CPS category, this result would go against its ratio legis.²⁰⁵ This reasoning is quite outcome-oriented. An alternative reading, which aligns with the DMA, is that social networking currently constitutes a contestable CPS category, as evidenced by the recent entry of TikTok, which makes designation unnecessary. After all, there need not be designation in every CPS category.²⁰⁶

In addition to refining the importance of single-homing for gatekeeper designation, the Commission took a closer look at user homing patterns. CPS users are not homogeneous, so it may be that at least a sub-group of them single-homes.²⁰⁷ In some cases, it is easy to find single-homing users on one side of the platform. End-users of OSs and related services (such as app stores) tend to single-home as they own only one mobile device.²⁰⁸ In other cases, there is multi-homing on both sides. ByteDance, for example, faces significant multi-homing by both content creators and end-users. But the Commission pointed out how such multi-homing may be, in part, asymmetrical: some content creators are mainly active on TikTok, using other platforms as ‘advertisement’ for their main platform; some end-users, especially younger ones, spend significantly more time on TikTok than on other social networks.²⁰⁹ ByteDance’s multi-homing argument failed because the multi-homing numbers it brought to the table did not account for this varying intensity of use.

It is difficult to see how a CPS-providing firm with significant multi-homing on both the business and end-user side can be a gatekeeper. Nevertheless, the Commission and Court have minimized the importance of multi-homing for designation. The evidentiary burden to rebut a presumption, in any case high, seems particularly difficult to meet here: the firm would have to show how different sub-groups multi-home and do so to a sufficiently significant extent, accounting for intensity of use. The principle that is used to animate this evidentiary burden is to test if the business user is economically dependent on a platform. Dependence does not prevent multi-homing: a developer may depend on being present across multiple channels, for example. If this is the key to understanding gatekeeper power, then the basic approach to market delineation may well suffice to ensure optimal designation. In this regard, guidance on the required size of sub-groups and their required intensity of multi-homing would be welcome.

Ecosystem

The Commission treats the ecosystem concept in almost every rebuttal decision.²¹⁰ This is remarkable, as the text of the DMA contains scant references to ecosystems.²¹¹ Similar to single-homing, however, the Commission is careful not to make ecosystems a decisive factor. In ByteDance, for example, it held that ‘nothing in [the DMA] suggests that having an ecosystem is an absolute pre-requisite to be an important gateway’.²¹² It finds support for that statement in a DMA recital that describes how ‘some’—not all—gatekeepers ‘exercise control over whole platform ecosystems’.²¹³

Even if this was not absolutely required, the Commission found that every designated gatekeeper controlled an ecosystem. The Commission did not define ecosystems, but decisions show how it sees them mainly in terms of interconnected products owned by a single gatekeeper, not in terms of the different actors complementing those products.²¹⁴ Apple’s ecosystem, according to the Commission, consists of hardware such as the iPhone and software such as iMessage.²¹⁵ In contrast, Apple—as much of the management literature²¹⁶—focuses on how different actors co-create value within the ecosystem.²¹⁷

The Court did infer an ecosystem definition from the DMA.²¹⁸ First, like the Commission, the Court stressed ecosystems’ multi-product nature, describing how ‘a digital platform ecosystem may consist of one or more CPSs and other services connected to them, for example by means of technological links or interoperability’.²¹⁹ But the Court added a multi-actor perspective by stating that an ecosystem exists where ‘several categories of suppliers, customers and consumers are brought together and interact within a platform’.²²⁰ Confirming the Commission, the Court held that having an ecosystem is a typical characteristic of some gatekeepers.²²¹

Presence in a single CPS does not prevent a firm from being an important gateway.²²² Inversely, the presence of an ecosystem does not necessarily mean it is such a gateway. What matters is whether the CPS benefits significantly from its links with other products in the ecosystem. In the words of the Court, ‘each ecosystem must be examined on a case-by-case basis, taking into account the benefits or lack of benefits deriving from such a business model’.²²³ In rebutting a designation, it is therefore a matter of showing how ecosystem advantages like default placement, favourable ranking, and other ecosystem privileges do not limit the contestability of the CPS in question.

In sum, the importance of a gatekeeper’s ecosystem in designation depends on its strength and the linkages with the CPS under consideration. The strongest type of ecosystem consists of multiple CPSs with gatekeeper status. As these CPSs constitute important gateways, they are likely to confer a significant advantage on connected services. It is then up to the gatekeeper to show how a specific CPS does not benefit from the wider ecosystem. While this may be difficult, it is not impossible: Microsoft managed to do so for Bing and Edge, and Apple for iMessage. In both cases, this was mainly due to the exceptional strength of another gatekeeper in the same CPS category (Google in search and browsers; Meta in messaging). Weaker ecosystems are built around a single CPS that meets the quantitative thresholds. Booking, ByteDance, and Amazon are all one-product gatekeepers. The other services making up the ecosystems can be CPSs without gatekeeper status, that is gateways but not important ones in the sense of the DMA. A combination of such gateways could still significantly benefit the ‘main’ CPS, but this is not a given. Samsung operated a number of CPSs, but they did not reinforce each other. Alternatively, the other ecosystem services do not qualify as CPSs, which makes it hard for them to significantly benefit the gatekeeper CPS. In describing TikTok’s ecosystem, for example, the Commission referred to other ByteDance services (eg in healthcare), which are not CPSs and did not appear to provide TikTok with any advantage.²²⁴ An exception is Amazon: its Prime service is not a CPS, but nevertheless an important part of its ecosystem, which includes Amazon Retail and Fulfilment by Amazon.²²⁵

Challenger

A gatekeeper is characterized by an entrenched and durable position. A position as a challenger to established ecosystems or gatekeepers would seem to be the opposite, and thus a good reason for non-designation.²²⁶ When ByteDance argued so, the Commission disagreed, holding that nothing in the DMA ‘suggests that an undertaking cannot be simultaneously a “challenger” to certain gatekeepers and a gatekeeper in its own right’.²²⁷

The Court confirmed the Commission but put more stress on the temporal dimension. A CPS-providing firm is presumed to be entrenched when it has met the user thresholds for three years. TikTok noted how it only had been operating in the EU for a total of five years.²²⁸ The Court found that TikTok was a challenger upon its entry in 2018 but had consolidated rapidly afterwards and far exceeded the user thresholds in 2023.²²⁹ The difference is that, while the Commission was willing to designate TikTok even as a challenger, the Court found the designation appropriate because TikTok was no longer a challenger. Nevertheless, the Court appears to echo an ambivalence found in some earlier antitrust precedent: dicta that the concept of gatekeeper ‘does not preclude the existence of a certain degree of contestability’²³⁰ resemble the statement that dominance ‘does not preclude some competition’.²³¹

The policy rationale not to designate challengers, even if they meet the quantitative thresholds, is that such designation may impede their challenge. It can thus go against the DMA’s goal of contestability.²³² The mechanism is that designation subjects the challenger to the DMA’s obligations, which—depending on the CPS category and enforcement intensity—can come with a significant compliance cost. Moreover, the obligations competitively constrain the challenger. A designated challenger is, for example, limited in the ways it can rely on user data to design or market new services.²³³ The incumbent gatekeepers that grew large before designation were not similarly constrained.

The Commission and Court were not willing to consider this link between designation and obligation, as it does not directly relate to the gatekeeper criteria.²³⁴ Microsoft, for example, claimed that if Bing were designated as a gatekeeper service, it would have to comply with Article 6(3) DMA, paragraph 2.²³⁵ This would require Microsoft to show Edge users a search engine choice screen, which would further entrench Google Search. The Commission rejected this argument as inadmissible under the DMA, where the emphasis is on the characteristics of the service.²³⁶ However, when discussing Edge, the Commission examined whether Microsoft could use the Windows ecosystem to drive Edge usage. It found that this was not possible, inter alia, because Article 6(3), paragraph 1 obliged Microsoft to facilitate the entry of rival browsers by making the uninstallation of apps easy.²³⁷ These two positions may be reconciled: when a gatekeeper has obligations that lead it to weaken its market position in another CPS, this can count as a factor that reduces its gatekeeper power in that CPS. Conversely, a gatekeeper cannot claim that the obligations imposed by the DMA would weaken it vis-à-vis other gatekeepers, because they are bound by the same obligations, and the idea of the DMA is to facilitate the entrance of firms that are not existing gatekeepers.

Normatively, the Commission seems to make a trade-off in favour of wholly new entrants and against challengers that already hold gatekeeper status, even if they are not as powerful as the incumbent platforms. The preference for wholly new entry may align with the DMA’s setup but can also make it more difficult to attain its contestability goal. Firms that are large but not quite as large (such as TikTok) or firms that are very large but not in the CPS category in question (such as Microsoft in search) are likely to prove most capable of contesting the incumbent’s position. If their ability to do so is constrained, this may harm rather than promote contestability.

6. Conclusion

Our comprehensive review of all gatekeeper designation decisions reveals that the Commission has applied the standards one expects from the DMA, but it has also enriched these, for example by considering how to best classify and delineate CPSs, by examining when technology neutrality is not a helpful concept, and by explaining how to disaggregate complex services. This is necessary not only for the purposes of designation but also for compliance. Its rebuttal decisions have utilized a narrow conception of ecosystems (focusing on the services of the firm under consideration) and have also provided guidance on the kind of evidence that a firm may produce to challenge gatekeeper designation such as the role of business user dependence and market power relative to other gatekeepers. Our method was to stress test the legal framework: we do not wish to second-guess any decision. Overall, we find that the Commission’s decisions are well-reasoned and its exercise of discretion explained fully.

However, once we step away from an explanation of the ‘law of designation’, the ontology of gatekeeper status remains elusive. Ecosystems with multiple, dominant gatekeeper CPSs are clearly within the scope of the DMA. A dominant CPS provider is a gatekeeper even if it does not have a multi-product ecosystem and is present in one CPS only. But, if there is contestability among gatekeepers, then the precise contours of gatekeeper status remain undefined. Perhaps business user dependence on one side of the market, with competition on the other side may be the lower bound of gatekeeper status. Refining the line between gatekeepers and emerging challengers will prove essential to guarantee that the DMA facilitates the entry of new CPS providers.

Footnotes

Acknowledgements

The article is current as of 10 December 2024, which is also when all links have been last accessed. The authors are grateful to the referees, as well as to Giacomo Tagiuri and other participants in the DMA Enforcement: Trends and Gaps in the First Year of Application conference (21–22 November 2024), for their questions and comments. According to the ASCOLA Declaration of Ethics, Giorgio Monti discloses that he wrote an expert opinion in relation to one designation process and is obliged to keep this confidential. Friso Bostoen has no declarations to make.