Designing (restorative) remedies for abuses of dominance by online platforms

Abstract

The debate surrounding the application of European Union competition law to online platforms so far has predominantly focused on identifying anti-competitive practices. The design of corresponding remedies has been far less explored, although it is of equal importance to practice. In the absence of effective remedies, the mere identification of anti-competitive practices can do little to protect competition. Given the difficulties associated with identifying competition law infringements by online platforms, there is no reason to assume that the design of remedies intended to tackle the effects of such infringements will be straightforward. Therefore, this article aims to show the intricacies of designing effective remedies for Article 102 TFEU infringements by online platforms based on the case study of tying and bundling practices. In this process, the article clarifies the importance of incorporating restorative measures in competition law remedies that may require opting for structural interventions, which are difficult to impose under the scope of Regulation 1/2003. The article offers three potential options to overcome such difficulties within the ambit of this regulation, namely through the strategic use of interim measures, the introduction of flexible remedies, and the incorporation of the regulatory restraints of the Digital Markets Act.

1. INTRODUCTION

Applying the current European Union competition law framework to online platforms poses multiple challenges for competition authorities and courts at both the national and EU levels. Many of these challenges stem from the distinguishing characteristics of online platforms, such as network effects, economies of scope and scale, and the growing importance of data for competition in such markets, enabling platforms to quickly acquire significant degrees of market power.¹ This combination of attributes enables shifts from competition in the market to competition for the market with a winner-takes-all as a prospective outcome.² These insights have channelled most of the debate and research on platforms and competition law toward focusing on the competitive concerns associated with the accumulation of market power and its use by platforms in practice.³ Although clarifying such matters is instrumental for the adequate application of the (EU) competition law framework to online platforms, the application process is not limited to the ability to correctly identify and qualify harmful practices within this framework. An equally important yet often neglected aspect of the application process is the design of remedies apt for tackling undesired and harmful practices. In the absence of adequate remedies, the goal of effective enforcement can hardly, if at all, be reached.

In the context of online platforms, the matter of remedies rarely gets the attention it deserves. It commonly constitutes the smallest part of the broader discussions about the competitive concerns associated with them.⁴ When it comes to abuse of dominance cases, the topic of remedies is explored mainly when it concerns an actual case where an infringement has been identified and condemned. In such circumstances, the discussion is not concerned with prospective remedy design possibilities, but instead, it serves as an after-fact critique of the competition authority or court that has already implemented a remedy for such practices. This is best demonstrated by the case of Google Shopping and previously by the Microsoft cases that have been subject to extensive ex-post critique.⁵ Such an approach is unfortunate as much can be gained from examining potential theories of harm in tandem with the corresponding remedies that would be suitable to offset the potential or actual competitive harm they cover.⁶ The parallel exploration of remedies and theories of harm allows the construction of more robust cases against the concerned undertaking(s) that are more likely to survive the stage of judicial review and have a meaningful impact in practice. It can even be argued that the outcome of the parallel study of theories of harm should be considered when it comes to prioritization, especially when enforcement resources are limited. Pursuing cases that cannot be effectively remedied may reduce the capacity of competition authorities and courts to deal with cases that can be effectively resolved, which may lead to even more significant competitive harm. In this regard, bringing remedy design consideration to the forefront of the analysis alongside theories of harm may help identify situations where legal intervention is needed in the form of regulation rather than ex-post enforcement.

In the case of online platforms, the importance of constructing remedies in the process of developing theories of harm cannot be overstated. This is because the distinguishing characteristics inherent to these actors and the markets in which they operate may continue to sustain and even amplify the anti-competitive harm caused by abusive conduct even when such conduct is abandoned. Such a risk will occur, as will be discussed, where the abusive conduct of a platform fuels the positive feedback loop that drives its growth while also shifting the homing patterns of its customers in its favour from multi- to single-homing. Consequently, abuses that rely on theories of harm covering such risks must be followed by remedies that can counter such actions to be effectively redressed. Doing so in practice will mean that remedies must not only require the concerned platform to abandon its abusive practices and prohibit their repetition, as commonly done, but also introduce a restorative component designed to restore or re-establish the state of competition undermined by the abuse. Failure to do so may mean that the concerned platform will continue to benefit from its (past) abusive strategies even after abandoning them, with no ability from competition authorities or courts to intervene (once more). This undesired outcome can be seen in the past cases of Microsoft WMP and Google Android, where the imposed remedies did not effectively address the anti-competitive harm identified in each case.

Despite the importance of restorative measures for the effectiveness of remedies, the ability and likelihood of implementing them in EU competition law depends on the legal framework that regulates this option, which is currently Regulation 1/2003. The Regulation sets the legal boundaries for remedies accepted or imposed by the European Commission that determine which legal forms remedies can take, when these can be used, and which components they can include. Consequently, the ability to design effective remedies for abuses of dominance by online platforms also requires mapping out the legal space for remedy design choices in this Regulation. Unsurprisingly, given the novel nature of these actors, the current framework of Regulation 1/2003 is hardly a perfect fit, thereby making the design of effective remedies capable of surviving judicial review legally challenging.

Against this backdrop, this article seeks to address the matter of designing effective remedies for abuse of dominance cases by online platforms. Therefore, this article intends to answer the question of how effective remedies for abuse of dominance by online platforms should be designed in light of their multisided nature. By focusing on remedy design choices while building on the insights of past experiences, this article helps prepare the ground for cases where the legal boundaries and objectives of competition law remedies will have to incorporate considerations concerning the multisided nature of platforms. By covering the topic of remedies, this article complements the existing literature on the finding of dominance and the abuse thereof by online platforms, which have been at the forefront of the debate in practice and academia for almost a decade, thereby adding the last piece of the puzzle of the application process of Article 102 TFEU to these actors.

The article will use tying and bundling abuses of dominance by online platforms as an example to show how the various remedy design considerations play out in practice with respect to these actors. The choice for this type of abuse was made based on the following reasons: (i) tying and bundling is perhaps the most likely type of a (well-established) abuse to be pursued by online platforms in their natural path towards expansion or envelopment ‘attack’; (ii) tying and bundling abuses are likely to have a significant impact on the positive feedback loop on the concerned platforms, as well as on the homing patterns of platform customers; (iii) there are past tying and bundling cases concerning platforms that can offer insights for future remedy design choices, in light of their shortcomings; (iv) there are currently multiple on-going investigations concerning potential abusive tying and bundling practices by online platforms at the EU level.⁷

Despite the focus on tying and bundling abuses, the insights provided by this article with regard to remedy design are relevant to all price and non-price-related abuses by online platforms where the positive feedback loop and homing patterns of platform customer were affected by their prohibited practices. The term remedies in the context of this article covers only public enforcement remedies that directly intervene in the commercial conduct of the concerned dominant undertaking. Accordingly, financial penalties such as fines or periodic payments, as well as damages imposed in the context of civil claims, are not addressed as these are not suitable for directly mending the competitive harm caused by abuses of dominance.

In order to answer the question posed by this article and provide coherent guidance for current and future practice, this article is structured as follows. Section 2 presents the current EU legal framework for remedies in abuse of dominance cases. Section 3 discusses the specific attributes of platforms that require consideration in the process of remedy design. Section 4 addresses the design of remedies in the specific case of tying and bundling abuses by online platforms in light of these attributes and the legal boundaries of the current legal framework under Regulation 1/2003. This section focuses on the dire need to tap into restorative measures, consisting at times of structural intervention in the case of online platforms. Section 5 examines the possibilities of overcoming the legal challenges posed by Regulation 1/2003 for remedy design. This section discusses three possibilities: the strategic use of interim measures, the introduction of flexible remedies, and the combination of competition law remedies with the framework of the Digital Markets Act. This section provides alternative approaches to remedy design, which would allow remedies to effectively terminate a competition law infringement and counter its corresponding anticompetitive effects within the boundaries of Regulation 1/2003, followed by some final remarks and conclusions.

2. REMEDY DESIGN: AIMS, MEANS, AND LIMITATIONS

Aims

Designing appropriate remedies for abuses of dominance is challenging as it requires a clear vision of the functions and objectives pursued by their implementation. In the context of competition law, the main objectives of remedies are terminating the infringement, preventing re-occurrence, and restoring or re-establishing the state of competition.⁸

Bringing an unlawful practice to an end is the objective that is commonly considered the first step that needs to be taken in each case. Doing so in practice typically entails requiring the concerned undertaking to conduct itself in a manner that is opposite to its prohibited practice.⁹ Putting an end to the undesired practice may, however, often not be sufficient if the concerned undertaking can resume such practices or similar ones at a later stage. Therefore, terminating the abusive behaviour is commonly paired with a mechanism intended to prevent the concerned undertaking from doing precisely that. For example, in Tetra Pak II, the Commission indicated that Tetra Pak must end all the behaviours the Commission found to be abusive and avoid repeating such actions or actions having a similar effect. To achieve this, the Commission provided a specific list of practices Tetra Pak must implement and avoid, together with a reporting obligation for five years after the decision.¹⁰

The third objective, namely restoring or re-establishing competition, is concerned with eliminating or reducing the harm the concerned undertaking creates, which would otherwise continue to benefit from its abusive actions. If the dominant undertaking, through its abusive behaviour, has accumulated significant market power, ordering it to cease its abusive behaviour may not be sufficient to restore the state of competition in the short and long term. For example, in the case of predatory pricing or loyalty rebates, if the dominant undertaking managed to put its competitors out of business, adjusting the pricing strategy of the dominant undertaking to a non-predatory level as well as converting the rebate scheme into a quantitative one would still allow the dominant undertaking to reap the benefits of its abuse. This, in turn, may even motivate the dominant undertaking to pursue comparable strategies in the future.¹¹ Therefore, together with preventive orders, additional indications may be provided to restore the state of competition to what it was prior to the abuse or at least re-establishing a state of competition previously impeded by the concerned undertaking. This can be seen, for example, in the case of Akzo, where the Commission indicated that Akzo was prohibited from offering its rivals’ clients lower prices than it offered its own customers. This prohibition applied to the customers that Akzo unlawfully took from its competitors as well as the existing customers of its rivals.¹² By imposing such measures, the Commission and EU courts allowed Akzo’s rivals to win back the clients they lost to Akzo due to predatory practices and restore the state of competition that was present before Akzo implemented its abusive practices.¹³

Effective remedies are therefore expected to address all these objectives while at the same time not being overly strict so as to disincentivize fierce competition.¹⁴ With no clear instructions, this challenging recipe has been used with varying success by the Commission and national competition authorities (NCAs) when dealing with abuse of dominance cases over the years. The multisided character of online platforms further complicates this process. The cumbersome character of this task has already been displayed by the Microsoft and Google Android cases, where the adopted remedies have had varying degrees of success despite the similarities between such cases.¹⁵

Nevertheless, when taken together, it is not hard to see why the tri-fold combination of objectives is the sensible approach for competition law infringements, particularly abuses of dominance, where market power gains are often part of the benefits gained by the dominant undertaking. Admittedly, however, the restorative aspect of remedies does not appear to be as high a priority as the other two objectives. In fact, the inclusion of restorative measures in the context of abuses of dominance is not often seen in practice despite their evident importance.

In the context of this article, the restorative objective of remedies will, however, be treated on equal footing with the other two objectives. The reason for this is that, as explained further, not addressing the restorative objective in the context of online platforms will run the risk of market tipping in favour of the dominant platform when no readjustment of illegally gained market power is addressed. This is because the particularities of platforms may continue to drive growth even after the prohibited practice is abandoned following the initial finding of abuse. At such point, competition may not only be harmed to a degree where restoration is (almost) unattainable but also where there may be no legal ground for (further) antitrust intervention.

The ability to combine and translate such objectives into concrete remedies in practice depends, however, on the legal framework that facilitates the possibility of imposing remedies following a finding of abuse. The legal framework determines the means through which such objectives can be pursued and the legal boundaries for using such means. Therefore, just as finding an abuse of dominance in the case of online platforms requires exploring the boundaries of Article 102 TFEU, so does the design of remedies for such abuses, which requires looking into the legal framework that shapes their corresponding remedies.

Means and limitations

The legal basis of remedies

The legal framework that regulates the imposition of public enforcement remedies for abuse of dominance infringements at the EU level is found in Regulation 1/2003.¹⁶ Article 7 of the Regulation provides the Commission’s power to impose remedies that ideally combine the abovementioned objectives. The formulation of Article 7 mentions only the ability of the Commission to require the concerned undertakings to bring their practices to an end in the event they are found to infringe Article 101 or 102 TFEU. However, both the decision-making practice of the Commission and the case law of the EU Courts have clarified that this provision also enables the Commission to include preventive and restorative aspects within the scope of its remedies, albeit subject to some limitations. Terminating the respective prohibited practices can entail, in practice, an order to cease a particular form of undesired behaviour and the imposition of positive duties on the dominant undertaking needed to bring the infringement to an end.¹⁷ The implementation of prevention mechanisms is possible; however, it is limited in terms of scope and concerns only the repetition of the practices that were found to be abusive.¹⁸ Prevention in this regard refers to a situation where future practices would entail de facto a continuation of the established infringement and not other practices that may raise (other) competitive concerns.¹⁹

Although many remedy provisions often include orders for the concerned undertaking not to repeat infringements that have similar effects, such orders are considered to have a merely declaratory character.²⁰ Accordingly, purely preventive orders that encompass more circumstances than the established infringement fall outside the scope of the Commission’s powers under Article 7.²¹ How the preventive mechanism or orders take form in each case will, of course, depend on the nature of the abusive behaviour. In the case of an exclusionary practice targeting a competitor in a vertically related market, such as a margin squeeze, preventive measures would concern the future commercial relationship between the dominant undertaking and the harmed competitor(s).²² In situations where the exclusionary behaviour of the dominant undertaking targeted the customers of its competitors, such as predatory pricing, the preventive measure will concern adjusting the commercial relationship between the dominant undertaking and the past, present, and sometimes future customers of its competitors.²³ In the case of exploitative practices, such as excessive pricing, the preventive aspect of the remedy would deal with future-proofing the commercial relationship between the dominant undertaking and its customers.²⁴

Similar considerations apply with regard to restoring or re-establishing the state of competition, where the nature of the abuse determines the competitive relation that needs restoring. In this regard, it is worth noting that while the objectives of preventing repetition and re-establishing competition may be considered separate, this distinction is often not visible in practice as measures that are required to prevent repetition are also required to bring the situation to its status ex-ante.²⁵ Similar to preventing repetition, restoring or re-establishing the state of competition is possible under Article 7. However, this possibility is limited in scope. Accordingly, the restorative aspect of the remedy should be restricted to curing the distortion caused by the infringement and its persisting effects. It cannot extend to measures that create a different market dynamic than before the abuse. This could have occurred in Akzo, for example, if the prohibition to align its prices to those of its competitors was extended to the customers it had before implementing its abusive practices.²⁶ Such an extension would have put Akzo’s competitors, who were harmed by its predatory practices, in a better position than they had prior to the abuse since it would have prevented Akzo from competing with them even on non-predatory price settings.

In addition to the power conferred to the Commission in Article 7 to impose compulsory remedies upon the dominant undertaking following the establishment of abuse, Regulation 1/2003 includes the possibility of implementing interim measures before a final finding of abuse has been made. Accordingly, based on Article 8, the Commission can adopt interim measures in the event of a prima facie finding of abuse where non-intervention risks irreparable damage to competition. In this context, a prima facie infringement can be said to exist if there is clear evidence indicating that a particular practice is quite likely or probably prohibited under EU competition law rules.²⁷ This requirement does not go so far as to require the same likelihood as in the case of a final decision establishing an infringement.²⁸ Nevertheless, it has been argued that such findings would be more suitable in clear-cut cases concerning established theories of harm compared to cases concerning novel situations,²⁹ which is an interpretation that EU courts also seem to support.³⁰ The pre-emptive character of interim measures makes them advantageous from the perspective of restoring or re-establishing competition. In this regard, interim measures enable the Commission to prevent situations that cannot realistically be restored if intervention is permitted only once the infringement procedure is completed.³¹

The implementation of interim measures in the case of (potential) abuses of dominance has been done in various situations, including predatory pricing, discriminatory trading conditions, and tying. Given the preliminary nature of such measures, the remedies adopted in the context of those procedures were concerned only with stopping the potentially prohibited behaviour.³² However, this use of interim measures predates the implementation of Regulation 1/2003, which now explicitly regulates this option.³³ The use of interim measures by the Commission following the implementation of Regulation 1/2003 only occurred once, in the recent case of Broadcom.³⁴ The halt in using this option can, to some extent, be explained by the legal changes that occurred with the implementation of Regulation 1/2003.³⁵ First, the implementation of interim measures was restricted to an ex officio initiation by the Commission in contrast to previous practice where potentially harmed parties (competitors or customers) could request the Commission to adopt such measures.³⁶ Second, the benchmark for urgency in Article 8 concerns situations where irreparable harm to competition would be caused in the absence of interim interventions. This is a higher threshold than its predecessor that could have been met if the risk of irreparable harm to a specific party affected by the potential infringement could be proven.³⁷ These changes to the legal framework of interim measures have also been said to make it more challenging to implement them in the case of the digital economy,³⁸ even though the Commission has explicitly noted it is interested in using interim measures in this specific context. Accordingly, such measures may become a valuable option for restoring or re-establishing the state of competition with remedies under Article 7, provided that the legal hurdles associated with them can be overcome, as will be further discussed.

Finally, Article 9 also provides that the Commission, in situations where it intends to implement an infringement decision, can accept commitments offered by the concerned undertaking to remove the competitive concerns identified by the Commission. Although all commitments inevitably entail making design decisions, such decisions are predominately made by the concerned undertaking(s) as their cooperation in such procedures is a (legal) pre-requisite.³⁹ The Commission’s primary role in such decisions is to adequately describe the competitive concerns it identified and indicate whether the commitments offered can alleviate them so that further pursuing the potential infringement is no longer needed.⁴⁰ Due to the voluntary character of commitment-based remedies, this procedure will not be further addressed within the scope of this contribution. Nevertheless, the insights from the following section on remedy design remain relevant for such procedures in the context of the review and approval of commitments by the Commission and EU Courts.⁴¹

Remedy types

Generally speaking, implementing remedies in abuse of dominance cases can include two possible types of remedies: behavioural and structural.⁴² As their name implies, behavioural remedies concern measures that seek to regulate the commercial behaviour of the dominant undertaking. Such measures can, for example, address the pricing strategies and contractual terms of the dominant undertaking with respect to its customers and competitors.⁴³ When implemented, behavioural remedies can concern both the abusive behaviour that needs to be brought to an end and further steps that need to be taken to prevent the repetition of the respective infringement(s) and restore or re-establish competition to the situation ex-ante.⁴⁴ Accordingly, in practice, behavioural remedies can impose both negative and positive obligations upon the dominant undertaking.⁴⁵ Structural remedies concern interventions in the corporate structure of the concerned undertaking by requiring it to sever the links from certain assets it holds. In practice, this may translate into the divestiture of one or more of the divisions operated by the dominant undertaking. Structural remedies, therefore, attempt to change the competitive structure of the market(s) affected by the abusive behaviour. Such actions can then achieve all three previously mentioned objectives of remedies with one manoeuvre. In the case of margin squeeze, for example, if the dominant undertaking is required to sell its downstream entity, it will no longer have a clear interest in distorting competition in this market, which may allow the re-establishment of competition with minimal risks of interference or minimization.

When it comes to implementing remedies in practice, both behavioural and structural remedies have their respective advantages and disadvantages. Behavioural remedies that target the conduct of the concerned undertaking entail a very flexible and accurate tool for curbing the undesired practices to fit within the lines of competition policy. By addressing solely the harmful behaviour of the concerned undertaking, behavioural remedies can re-establish compliance with competition policy with minimal collateral damage for third parties.⁴⁶ At the same time, since such measures concern the restriction of conduct, they also require continuous monitoring by the respective competition authority or courts. This is because constraining an undertaking’s behaviour does not necessarily eliminate its incentives to continue with its prohibited practices. Therefore, designing behavioural remedies requires considering the various manners in which the concerned undertaking may attempt to minimize the impact of such remedies on its commercial practices.⁴⁷ This can occur, for example, in cases concerning discrimination, input foreclosure, and refusal to grant access to a facility, which require stringent and accurate measures to prevent the concerned undertaking from undermining the pursued effects of the remedies.⁴⁸ The more complex the abuse and the greater the scope of harm created by it, the more complex such remedies and their monitoring mechanisms would likely have to be.⁴⁹

When it comes to monitoring, structural remedies, on the other hand, can be said to offer a one-off solution. Once the divesture of certain assets of the concerned undertaking has been required and the restructuring is completed, there is little need for further compliance monitoring. This is because such remedies are capable of removing the incentives as well as the ability of the concerned undertaking to continue pursuing its anti-competitive practices. This, in turn, also reduces the possibility of minimizing the effect of the remedy by the concerned undertaking. At the same time, the possibility of imposing a divestiture depends significantly on the corporate structure of the concerned undertaking. If such a structure is relatively modular, meaning that the various subsidiaries and divisions of the concerned undertaking are spread across separate commercial entities, divesture would be more realistic than in a case where the concerned undertaking is vertically integrated throughout the entire supply chain. In the latter case, divesture is more challenging to administer and has a higher risk that the efficiencies gained through integration may be lost, which in turn may diminish incentives to innovate.⁵⁰ Furthermore, a successful divesture also requires market conditions to support such changes. If the market will not support creating an additional player that is supposed to arise following the divesture, then there is little sense in ordering it. Divestitures that lead to the creation of unviable businesses can do little to improve the state of competition in the markets affected by the prohibited practices of the dominant undertaking and are also more likely to produce undesired effects for third parties.⁵¹ This logic can also be seen in the context of EU merger control, where the approval of concentrations subject to divesture commitments requires that the severed entity is a viable business that should be acquired by a buyer capable of turning it into a genuine player on the market where it is active.⁵² Finally, since structural remedies primarily entail severing (some) ties between the dominant undertaking and its assets active across different markets, such measures would commonly make more sense in situations with market power leveraging abuses than single-market abuses. In the case of single market abuses, divesture remedies would consist of horizontal break-up measures, such as the famous Bell Company break-up in the USA, which are more challenging to administer and entail de facto a more far-reaching act of intervention and market regulation. Therefore, while structural remedies may tackle infringements in a more sweeping manner than behavioural remedies, they are not the ultimate solution for all abuses or market conditions. Consequently, in practice, the suitability of each option will depend greatly on the specific circumstance of the case.

In the context of EU competition policy, the division between behavioural and structural remedies is explicitly provided in Article 7 of Regulation 1/2003; however, these two options are also available (to some extent) in the context of Articles 8 and 9 procedures. The temporary nature of interim measures, as well as their objective to prevent irreparable harm from occurring, makes them less suitable for structural remedies. However, from a purely theoretical legal perspective, there is no reason why this would never be possible.⁵³ In the case of Article 9 commitment procedures, both options have been implemented in practice. It can even be said that structural remedies are more common in such settings than in the case of Article. 7.⁵⁴ The choice between behavioural and structural remedies, as well as between the possible variations of measures within each type of remedy or combination thereof is, however, not entirely subject to the Commission’s discretion in practice. The selected remedy must be both effective and proportionate, or it risks being overturned at the stage of judicial review.

Limitations—effectiveness and proportionality

Article 7 of Regulation 1/2003 indicates that in case of an infringement (of Article 102 TFEU), the Commission has the power to impose on undertakings any behavioural or structural remedy that is proportionate and necessary to bring the infringement effectively to an end. Structural remedies should, however, be implemented only in cases where behavioural remedies are not equally effective or are more burdensome for the concerned undertaking. A similar formulation is also included in Recital 12 of the Regulation, which adds that changes to the structure of the undertaking as it was before the infringement are only proportionate in situations where there is a risk of lasting or repeated infringement stemming from the structure of the undertaking. Therefore, it can be said that there is an initial preference for behavioural remedies in practice; however, it is limited to situations where structural and behavioural remedies are equally effective.⁵⁵ This is something that the Commission will have to show in its current investigation against Google in relation to ad tech, where it stated it would be interested in imposing a structural remedy.⁵⁶

When designing a remedy, the Commission must make sure that it is proportionate to the infringement and is necessary to bring it effectively to an end. Effectiveness in this context should be read as the ability to stop the abusive behaviour, prevent its repetition, and re-establish competition.⁵⁷ In EU law, proportionality (of measures) entails that EU measures must not exceed what is appropriate and necessary to achieve the objective pursued.⁵⁸ When there is a choice between several possible measures, the least burdensome should be selected, and the disadvantages must not be disproportionate to the aims pursued. In cases concerning compulsory remedies under Article 7 of Regulation 1/2003 or under the older Article 3 of Regulation 17/63, similar definitions for this principle were provided.⁵⁹ Although not explicitly addressed in the context of interim measures under Article 8, these remain, of course, EU measures and are thus inherently subject to this threshold. Nevertheless, since Articles 7 and 8 pursue different objectives, applying the proportionality principle in the context of both procedures will look different.⁶⁰

For remedies to be proportionate, these must correspond with the nature and scope of harm caused by the abuse of dominance. The remedy selected must be linked to the theory of harm identified in each case to be considered appropriate and necessary.⁶¹ The importance of this link has been observed in practice in cases where the remedy imposed by the Commission was annulled for going beyond the scope of the infringement and the theory of harm behind it.⁶² At the same time, this correlation should not mean that the Commission must impose identical remedies in all cases dealing with the same type of abuse. Such a requirement would negate the circumstantial differences across various cases. Nor should the link between the remedy and theory of harm be considered to determine the legal test of abuse that needs to be applied in the first place. According to the GC in Google Shopping, the remedy imposed by the Commission does not dictate which legal test should be applied to the practice at hand to reach a finding of abuse.⁶³ This conclusion addressed one of the main critiques expressed in the context of the Google Shopping case involving the link between the remedy and the discussed theory of harm, as the imposed remedy was claimed to entail a market access remedy, thereby requiring a corresponding test of abuse.⁶⁴ Therefore, it is only in cases where a significant discrepancy between the theory of harm and the imposed remedy can be identified that the suitability of such a remedy can be questioned. Such a discrepancy would occur in situations where the remedy would attempt to address a different, perhaps larger, competitive problem than the one used as the ground for finding an abuse. This, for example, could have happened if a remedy requiring the untying of the App Store from iOS had been imposed in the case of the Apple App Store when the established abuse concerns the anti-steering obligation imposed on App developers. The two issues are certainly connected but pertain to different competitive concerns both in nature and in magnitude. Deviations from previous practice, as such, should, however, not be sufficient to consider a remedy inappropriate and thus disproportionate as long as the case circumstances are also different. This would follow the Commission’s practice for the past few decades, where the same kinds of abuses were subject to different remedies that attempted to address the specificities of each case.⁶⁵

In the case of online platforms, allowing for deviations from previous practice while insisting on a close fit between the theory of harm and the designed remedy is imperative for the correct application of EU competition law to these actors. Online platforms and their respective commercial practices entail significantly different circumstances from the conventional market setting, where most of the Commission’s practice originates. Therefore, addressing the competitive concerns and harm that may arise in such circumstances will almost inevitably require deviating from previous practice to avoid economically unsound outcomes. Such deviations should, nevertheless, always maintain the consistency between the remedy and the identified theory of harm. To do so, the multi-sided nature of online platforms and the particular dynamics of competition in the markets in which they operate must be accounted for. Although this task may sound straightforward, the Commission’s experience in the Microsoft and the Google Android cases shows that performing it in practice is difficult.⁶⁶ The following section will, therefore, cover some of the primary considerations that must be addressed when designing remedies for abuses of dominance by multisided online platforms to increase their likelihood of success.

3. REMEDY DESIGN IN CONTEXT: PLATFORM CONSIDERATIONS

In the context of competition law policy, online platforms are commonly discussed in relation to their multisided market character, which places them in a different commercial reality compared to non-platform undertakings.⁶⁷ Such reality requires that the methods of assessing competitive harm and the means to mend it be adjusted to account for the differences between platform and non-platform market settings.⁶⁸ Broadly speaking, the different commercial reality stems from the fact that platforms create value by successfully facilitating (matchmaking) interactions between two or more separate customer groups.⁶⁹ Consequently, the success of platforms depends on their ability to bring the right customer groups ‘on board’ and coordinate the interaction between such groups in a profitable manner.⁷⁰ Therefore, when competing with other (platform) undertakings, platforms compete to attract the customer types they wish to get on board in optimal proportions, depending on their business model.⁷¹ An online marketplace, for example, competes with other comparable marketplaces with respect to both consumers as well as sellers since not having one of the two customer groups ‘on board’ means it cannot deliver the value it seeks to create, namely facilitate transactions between these two groups.

Igniting and diffusing network effects

In practice, successfully getting members from different customer groups on board can and has been achieved through various strategies.⁷² Generally speaking, because different customer groups may often have different degrees of demand for the (matchmaking) service offered by platforms, platforms often have to make different value propositions for such customer groups to get them on board.⁷³ This commercial reality manifests in the adoption of skewed pricing structures where (at least) one platform customer group (the subsidizing group) pays more for using the platform than another group (the subsidized group).⁷⁴ Such skewed pricing structures can be implemented through all the monetization possibilities that platforms have to make a profit on the (matchmaking) interaction they facilitate, such as pay-per-click, membership, and transaction fees, as well as a combination thereof. Like the pricing structure, the monetization mechanism is also essential for enabling different customer groups to get on board.⁷⁵ Once members of two or more separate customer groups join the platform, the (positive) indirect network effects at play between such groups are said to enable the growth of the platform to viability (also referred to as critical mass) and, in the long run, even help secure a monopoly position.⁷⁶

Going back to the example of the online marketplace, once some consumers and sellers are attracted to the platform, growth will proceed due to a positive feedback loop; the more consumers it attracts, the more attractive it becomes for sellers and vice versa. To manage such growth patterns, platforms will implement governance rules to reinforce such indirect network effects and prevent undesired practices on the platform that may diminish or reverse such effects. In the case of the marketplace, the platform owner can, for example, choose to accept more payment methods, which may attract more consumers and merchants, as well as remove merchants that sell counterfeit goods since they may cause consumers to stop using the marketplace, which in turn will also cause ‘good’ sellers to leave thus triggering a downward spiral to failure.⁷⁷ Accordingly, when it comes to nurturing the growth of platforms and thus their (relative) market power, platform undertakings will focus their efforts on price and non-price commercial practices that amplify the positive (direct and) indirect network effects at play on their platform as well as their ability to internalize and capitalize such effects. This, in turn, would mean that the anti-competitive behaviour pursued by platform undertakings will seek to achieve this same goal while choosing means that do not entail competition on the merits.⁷⁸

Therefore, when addressing the harm created by such practices, the remedy should seek to halt the self-reinforcing positive feedback loop generated by such (positive) network effects. Failing to do so would entail, in practice, that the concerned platform may continue to grow and gain market power at the cost of its competitors (and, at times, of its customers), even if the anti-competitive practice is abandoned. In this regard, it is worth noting that the remedy should not aim to set the network effects in reverse. This may put the concerned undertaking in a worse state than before it engaged in the abuse and even to market exit, which would be considered disproportionate. This is very challenging as it essentially means reversing the snowball effect caused by the abuse back to the top of the mountain while ensuring it does not go in the opposite direction.

Shifting and adjusting homing patterns

In addition to amplifying the network effects at play, platform entities can seek to obtain more market power by curbing their customer groups’' participation into a pattern that provides them with the most significant degree of (relative) market power. In the context of platforms, such participation patterns are referred to as single or multi-homing patterns⁷⁹ and are considered to be one of the main criteria for establishing the existence of market power.⁸⁰ Single-homing occurs when a platform customer group uses a single platform to fulfil one of its specific demands for a product or service. For example, this would be the case if consumers would use only Booking.com to make their online hotel reservations. Multi-homing occurs in the opposite scenario, where a platform customer group uses multiple platforms to fulfil the demand for the same service or product the platform offers. This occurs, for example, when consumers use multiple online marketplaces to purchase retail goods. As platforms inherently cater to two or more customer groups, such participation patterns can be observed with respect to each. Generally speaking, the settings that can then be observed are (i) multisided single-homing, (ii) multisided multi-homing,⁸¹ and (iii) single-homing on one side and multi-homing on another side of the platform,⁸² also referred to as a (competitive) bottleneck scenario. From a market power perspective, the platform owner would commonly have the incentive to ensure that at least one of its customer groups is single-homing, as this provides the platform with monopoly power with respect to the other platform customer groups that are interested in interacting with such single-homing groups. If consumers, for example, would only (or primarily) turn to Booking.com to reserve their hotel room, then Booking.com would have significant power over hotel owners wanting to reach consumers. This, in turn, would allow Booking.com to charge higher fees from such hotels in return for getting access to the platform. Where single-homing is achieved on all sides of the platform, such market power would be even more significant as it would apply, in principle, to all groups simultaneously.⁸³

By contrast, in a multisided multi-homing scenario, the market power of the platform with respect to its customer groups and competitors is constrained by the ability and willingness of such customer groups to use alternative solutions. Getting one or more customer groups to single-home through practices that entail competition on the merits requires, however, prevailing under the intense competitive conditions associated with single-homing tendencies.⁸⁴ Achieving this would require consistently providing customers with the most innovative service with the best quality at the (relative) lowest price. In the absence of high switching or multi-homing costs, there is, in principle, nothing preventing customers from using competing services simultaneously.⁸⁵ Given the decisive role played by the single-or multi-homing patterns of platform customers for the accumulation and preservation of market power by platforms, it is sensible to assume that the competitive and anti-competitive behaviour of such dominant platforms will aim at curbing such patterns in their favour through price and non-price practices. Accordingly, anti-competitive practices by online platforms will likely aim to get at least one of their customer groups to single-home and/or reduce the potential impact of multi-homing on their respective market power.⁸⁶ Therefore, when seeking to address harm (potentially) created by abusive practices of dominant platforms targeted at influencing the single-and multi-homing patterns of platform customers, the envisaged remedy should aim at preventing such patterns from steadily shifting in favour of the concerned undertaking and restoring them to their previous setting before the abuse.

Exploitative vs. exclusionary concerns

The insights mentioned above on network effects and platform customer homing patterns concern settings in which the competitive harm is of an exclusionary nature. In situations where the competitive harm has an exploitative character, the network effects at play on the platform and the platform customer participation patterns would play a different role and thus may require a different approach when considering remedies. In such a context, the manner in which network effects manifest allows the platform owner to know which customer groups are more likely to generate higher rents, as network effects have been identified as one of the main determinants of platform pricing.⁸⁷ In situations where the dominant platform undertaking seeks to exploit the customer group it considers most susceptible to such practices, the remedy may not have to aim at addressing the network effects. In such cases, the remedy could aim to prevent the concerned platform from making use of an opportunity that it would not be able to obtain in a competitive market vis-à-vis the respective customer group.⁸⁸ By doing so, even cases where the exploitative practice assists the platform in amplifying the (indirect) network effects on the platform may be resolved as removing the exploitative element from such practices essentially takes away (some of) the fuel needed to sustain such undeserved advantages.⁸⁹

Similarly, the participation patterns of platform customers will determine to a great extent which customer group is likely to generate the highest participation fees. Generally speaking, single-homing customers are considered more valuable than multi-homing customer groups, and thus, the latter are more likely to be subject to higher participation fees.⁹⁰ In the case of abusive practices with an exploitative character, the remedy would, in principle, not have to recalibrate the participation patterns of platform customers but instead target the exploitative practice. This is, however, conditional on the fact that the exploitative practice did not also assist the platform in shifting the participation patterns of its customer groups in its favour. This could occur, for example, when unfair pricing or trading conditions also increase the switching and/or multi-homing costs of the harmed customer group or help the concerned platform to secure single-homing by other customer groups. In such a case, the remedy may also need to aim to enable such participation patterns to return to their pre-abuse state. Doing so would, in essence, require intervening with the practices that accompanied the implementation of excessive prices and caused such patterns to shift. For example, suppose the additional rents extracted from one customer group were used to finance a lock-in mechanism intended to secure the single-homing of a different customer group. In that case, the envisaged remedy should tackle both aspects. This would entail intervening in the price setting of the concerned dominant platform and dismantling the lock-in mechanisms financed by such abuse that would allow for multi-homing to resume. Not addressing such participation patterns may otherwise lead to a situation where the harmed customer group will continue to be subject to relatively higher participation costs even once the exploitative practice is halted.

Identifying and preventing tipping

In light of the above, it is essential that at the phase of remedies, not only the immediate harm to competition is observed and addressed but also the future implications of the anti-competitive behaviour for the competitive process. In this regard, addressing the impact of the anti-competitive practices of the concerned dominant platform on the network effects at play in each case and on the participation patterns of platform customers will also prevent market conditions from tipping in its favour. Market tipping refers to the situation where a platform succeeds in securing an (impassable) advantage over its competitors that persists until such platform becomes a (quasi) monopoly.⁹¹ The tendency of market tipping in the context of platforms has been identified as one of the main challenges of current competition policy, which will require adjustments to possess the tools to prevent its occurrence.⁹² This is particularly so in situations where tipping results from legitimate competition.⁹³ In such cases, the impact on the network effects on the respective platform and the homing patterns of its customer groups may occur following well-planned legitimate business practices. Consequently, despite the undesirable outcome of such practice, the result of tipping, as such, will not justify legal intervention under Article 102 TFEU, even when it involves a dominant platform. Intervening in this undesired outcome at such a stage can only be done through other regulatory tools, which target the (market) outcome rather than the permissive or prohibited commercial practices that led to it, such as the DMA.⁹⁴

When, however, the tipping effect or tendency is amplified through anti-competitive practices adopted by a dominant platform, the grounds for intervening are present, and tackling such an undesired outcome can be done at the remedy phase. The manner in which this will occur in practice will differ from case to case based on its specific circumstances. Nevertheless, this objective is unlikely to be attained without addressing the effects of the anti-competitive behaviour on the network effects at play in each case and the homing patterns of platform customers. This is because these two aspects constitute perhaps the two most important factors for assessing the facilitating and mitigating circumstances for tipping, which apply to all platforms regardless of the sector in which they operate.⁹⁵ As previously mentioned, the network effects at play on a platform can ignite a positive feedback loop provided that the correct customers are brought ‘on board’ and managed adequately. This positive feedback loop allows the platforms to grow quickly. Once an initial market share advantage is secured, if managed properly, this feedback loop will sustain and gradually enlarge such advantage to the point where the initial frontrunner prevails as the undisputed winner.⁹⁶

This outlook is particularly relevant in market settings where (at least) one of the platform customer groups is prone to single-homing. In such a scenario, if the market tips on the single-homing side of the platform, it essentially means such a player becomes the gatekeeper for the entire single-homing market. For example, if consumers would generally use only one platform to make their hotel room reservations, then hotel owners would have to list their properties on that platform if they want to reach consumers. By doing so, the hotel owners would reinforce the tendency of consumers to stick to the respective platform as it will have all or the most offers compared to other platforms. In the long run, this can also reduce the incentive for hotel owners to list properties on other platforms, leading to single-homing on both sides of the platform, which provides the platform with a near monopoly position with regard to both customer groups.⁹⁷ Therefore, by addressing these factors in each platform-related case, the envisaged remedies will be better able to obtain their common tri-fold objective and address the relatively new challenge of market tipping, which has been identified as a major concern in the greater context of digital markets.

The following section will use the example of tying and bundling cases to showcase how the platform considerations mentioned above could manifest in practice and how these could be addressed within the existing framework of Regulation 1/2003.

4. REMEDIES IN TYING AND BUNDLING CASES

Tying and bundling practices in the case of multi-sided platforms entail several variables that need to be considered for remedy design. First, tying practices in the case of multisided platforms can take two forms, namely on-platform tying or cross-platform tying.⁹⁸ On-platform tying refers to situations where the matchmaking functionalities offered on a platform are tied to one another.⁹⁹ Cross-platform tying, by contrast, refers to situations where (at least) two separate platforms are tied to each other.¹⁰⁰ Secondly, the multisided nature of platforms means that the tying practices can be implemented with respect to more than one customer group. For example, in theory, Amazon could tie the Amazon Pay platform with consumers and merchants using the Amazon Marketplace. Thirdly, the intermediary role played by platforms can cause the tying with respect to one customer group to extend across the platform to other customer groups. This occurred, for example, in Microsoft and Google Android, where the tying practices applied to OEMs also resulted in tying in the case of consumers.¹⁰¹ Accordingly, when dealing with abusive tying practices of a platform, the remedies need to account for these variables as they might impact the scope of competitive harm caused and thus also the kind or remedies required for tackling it. To clarify the competitive concerns posed by such practices and the challenges these entail for remedy design, comprehensively, on-platform and cross-platform tying will be discussed separately.

On-platform tying—competitive concerns and remedy design challenges

On-platform tying, involving two or more matchmaking functionalities of a platform, will most likely concern only the consumer side of the respective platform. This is because the various matchmaking functionalities on a platform involve bringing together consumers and one or more commercial customer groups. So, while consumers could perhaps be required to utilize more than one match-making functionality, the same is often not possible with regard to the commercial customers of the platform since it would essentially mean that these companies would have to offer more than one service to consumers. For example, Booking.com could tie its matchmaking functionalities with respect to consumers and thus require them to use the hotel reservation service when reserving an airline ticket on the platform. Since consumers commonly need both when going on vacation or a business trip, such a requirement would not be unrealistic from a commercial perspective. Applying the same strategy to the commercial customers of the platform would, however, be unrealistic since it would require airlines also to offer hotel rooms on Booking.com. Consequently, this type of tying is unlikely to extend from consumers to other customer groups.

When dealing with abusive on-platform tying, the competitive harm that must be prevented will commonly concern the markets of the tying and tied matchmaking interactions and potentially the emersion of a new market.¹⁰² To prevent harm from occurring in such markets, the tying obligation of the platform must be removed. Given the wording of Regulation 1/2003 and in the absence of any evident exceptional market conditions, it can be expected that such steps will be attempted predominantly based on behavioural remedies, which, as will be seen, have their limitations.

When the tying element is contractual and thus forms part of the terms and conditions of the concerned platform, the remedy should first require its removal, as was done in the case of Tetra Pak and, more recently, in the case of Google Android.¹⁰³ When the tying of matchmaking interactions is facilitated through technical means,¹⁰⁴ such technical elements should be removed so that one matchmaking functionality does not automatically trigger or depend on the usage of a second functionality. Such requirements would, in principle, end the practice that directly harms competition in markets of the tying and tied matchmaking functionalities. In the absence of the tie, competitors of the concerned platform in either market could resume competing for consumers on a stand-alone basis. However, if such practice helped the platform fuel the positive feedback loop on the platform, meaning it has attracted a more significant consumer and commercial customer base, removing the tying vehicle may not suffice to remove the harm created by it. This is because the (indirect) network effects at play may still make it difficult for competitors to compete for consumers. If Booking.com manages to get more consumers, airlines, and hotel owners on board by tying their hotel room reservation and airline ticket search functionalities with respect to consumers, removing this tie does not necessarily mean consumers will quickly consider switching to competitors in the markets of either functionality. Therefore, the positive feedback loop accelerated by the tying practice may proceed even after the tie is removed.¹⁰⁵

The perseverance of the positive feedback loop would be most evident in cases where the tying practices lead to the formation of a consumer bias in favour of the concerned platform for the (previously) tying and tied functionalities.¹⁰⁶ Consumer bias in this regard refers to a situation where the consumers internalize the choices imposed by the tying practices as their default choice. Once such a bias is created, the positive feedback loop will not only continue to fuel the platform’s growth but may also gradually drive the participation patterns of the customer groups previously involved in the tie toward single-homing and overall market tipping. Breaking through such a self-reinforcing consumer bias can be very challenging, requiring consumers to reconsider their default choices.¹⁰⁷ To do so, a remedy would have, in addition to removing the tie, also prevent the respective platform from engaging in practices that reinforce such bias. This could occur, for example, if the tying of functionalities is replaced by constant nudging strategies that aim to preserve the same consumer behaviour previously imposed by the tie. Going back to the example of Booking.com, this would occur if Booking.com would stop tying its hotel room reservation and airline ticket search functionalities and display instead constant reminders to consumers to use such functionalities in tandem, and perhaps even link such use to some promotional fee.

Furthermore, actions that are targeted at making multi-homing by consumers as well as by the commercial customers of the platform more difficult should also be prohibited as these may have a similar effect of preserving the bias and driving platform participation towards single-homing. This would occur, for example, if hotel owners that add their property listing to other platforms would receive a less favourable ranking on Booking.com. While such actions would not directly form part of the tying practice directed at consumers, they would nevertheless undermine the effectiveness of the untying remedy by preventing hotel owners from switching to other platforms following the imposition of such a remedy. In other, words such actions would prevent the situation from going back to the status ex-ante by locking-in the growth achieved (ie, increase in hotels) trough anti-competitive tying. Suppose such actions are, however, not sufficient to halt or at least slow down the positive feedback loop ignited by the tying practices. In that case, there is little room for resolving such a situation with a behavioural remedy. The last resort in such a scenario would be to reduce the visibility of the tied functionality on the platform when consumers are using the tying functionality and vice versa to nudge consumers into actively reconsidering their default choices. Although such a measure may raise some of the search and transaction costs experienced by consumers temporarily,¹⁰⁸ the benefit of healthy competition may outweigh such considerations in the long run. Of course, as with any behavioural measures, constant monitoring of compliance with such requirements would be needed and, given their subtle and complex nature, would likely be costly.

Minimization by the concerned platform can thus be expected, as it would have an incentive to look for other means to keep reinforcing the consumer bias it managed to create. If, despite these measures, the consumer bias remains persistent and the positive feedback loop is not halted or slowed down, there is little that can be achieved through other behavioural measures. Consequently, the last resort option would be to require the two functionalities to be split across two separate platforms. Such a separation would entail a quasi-structural remedy as it does not need to go as far as to request the divestiture of the commercial division behind a specific functionality. A genuine divesture in such a case would likely be disproportionate as a change of ownership would not be necessary to influence consumer behaviour. The structural separation would only concern how the functionalities are offered to consumers, namely on two stand-alone platforms instead of one covering both functionalities. Such an approach would undoubtedly have the potential to impact consumer behaviour and be suitable for dealing with consumer bias scenarios. Nevertheless, it could only be considered proportionate in the most severe cases involving on-platform tying, where the prospect of market tipping is almost imminent. This is because the previously mentioned behavioural measures may sometimes suffice while being less cumbersome for the concerned undertaking and imposing fewer additional costs on consumers. This could occur, for example, in cases where the network effects on the platform are moderate and multi-homing by the platform customer groups persists despite the tying practices. Such circumstances mitigate the possibility of the concerned platform aggregating market power and its ability to make the affected markets tip in its favour.¹⁰⁹

The ability to assess ex-ante whether the suggested behavioural remedies will be effective is, however, very limited in practice. Although platform participation patterns can be studied with relative ease, correctly assessing the nature of the network effects at play on each platform and measuring their intensity is far more complex. Despite the extensive literature on network effects, there is currently no consensus on how such effects should be measured and evaluated in the context of competition policy. Similarly, there are very limited tools for predicting and assessing the tipping tendency of platform markets.¹¹⁰ Without legal tools that convincingly show that a respective case displays the most alarming anti-competitive character, it is hard to see how a structural remedy such as the one previously mentioned could be considered a proportionate first-pick solution. Recital 12 of Regulation 1/2003 indicates that changes to the structure of the undertaking would be proportionate when the anti-competitive risk arises from its structure as such. However, such a conclusion would be hard to reach in the absence of legal tools capable of assessing the nature and intensity of the network effects at play and the market-tipping tendency in a given case. Consequently, although structural remedies may be needed to prevent the anti-competitive harm caused by on-platform tying to progress after such behaviour is brought to an end, the current legal framework makes it very cumbersome to adopt such measures under Article 7 of Regulation 1/2003. Developing new legal tools for assessing network effects and market tipping tendencies is therefore certainly desired to increase the likelihood that more stringent (structural) remedies can be imposed at a (relatively) early stage, that is before tipping takes place or becomes inevitable.

In the worst-case scenario, in which tipping has occurred, there would be a strong argument towards a full-on structural remedy entailing a divesture that would be proportionate. In the event of tipping, the concerned platform would have essentially managed to monopolize at least one of the two markets for the tied functionalities. Accordingly, allowing both functionalities to be offered and thus monetized by the same corporate entity would, in essence, entail forfeiting such markets to it for the long run with the prospect of future expansions and leveraging to other related markets. The seriousness of this outcome seems to be acknowledged by the GC in Microsoft, where the Court indicated that reversing the anticompetitive effects of competition law infringements in markets characterized by network effects is extremely difficult.¹¹¹ This can be argued to be the case with the more recent investigation of Microsoft concerning its tying of Teams to Microsoft 365 and Office 365 packages. The tying practices, amplified during the COVID-19 pandemic by a surge in demand for video conferencing and cooperation software,¹¹² have potentially tipped the market in Microsoft's favour beyond the point of return. In such circumstances, anything short of divestiture will allow Microsoft to benefit from the market power advantage it acquired through its tying actions even after ending these practices.¹¹³ The possibility to do so under Article 7 of Regulation 1/2003 remains, however, subject to the ability to argue not only that structural remedies would be effective and proportionate but also that behavioural remedies are not sufficient.

Cross-platform tying—competitive concerns and remedy design challenges

Cross-platform tying, where two or more stand-alone platforms are tied with respect to one platform customer group using both platforms, can, in principle, occur with respect to both customers and the commercial customer group(s) of platforms. The possibility to impose the tying practices with respect to either side of the platform will occur mostly in situations where the matchmaking functionalities offered by the tied platforms concern complementary services from the perspective of both consumers and the commercial customers of the platform. For example, Amazon may tie its digital payment platform, Amazon Pay, to its marketplace platform for merchants and consumers, either separately or simultaneously. This is because such a payment platform is a suitable complement for enhancing the services provided on a marketplace platform. However, not every strategy in this respect would be equally viable as trying to tie the marketplace services to Amazon Pay is highly unlikely to succeed, as a marketplace cannot be considered an essential complement to a digital payment platform. Accordingly, cross-platform tying can sometimes occur with regard to more than one platform customer group. In such situations, tying two or more platforms for one customer group may extend to another. As previously mentioned, if Amazon Pay were tied to the Amazon marketplace for consumers, merchants would also likely accept such a payment method. Such an outcome can be expected even when the tie does not require consumers to use Amazon Pay as their exclusive payment method. Therefore, when considering remedies, the variations in how cross-platform tying can manifest should be considered, as they may require slightly different types of interventions.

The competitive harm that can be expected in the case of cross-platform tying would commonly concern the exclusion and deterrence of competition in the markets of the stand-alone tying and tied platforms. Where cross-platform tying is applied to consumers and does not extend to other platform customer groups, a behavioural remedy that removes the tie would go a long way toward bringing such an infringement to an end, preventing its repetition and allowing for the state of competition to return to its pre-abuse status in principle. The removal of the tie should, of course, be complemented by the prohibition of adopting other means aimed at preserving and reinforcing consumers' joint use of the previously tied platforms. Such actions could include some form of preferential treatment between the previously tied platforms, such as providing monetary incentives, undermining interoperability with third parties, and nudging notices constantly pulling consumers back to their previous behaviour. Although tying two platforms can certainly contribute to reinforcing the positive feedback loops of each platform, removing the tie in such a situation would remove a great deal of such effects across platforms. Accordingly, once the tie is removed, each platform’s network effects may continue to fuel growth on each platform separately. However, the two platforms would not continue to feed off each other’s success as in the case of on-platform tying. Going back to the example of Amazon, if Amazon Pay were no longer tied to Amazon Marketplace, an increase in the use of Amazon Pay by consumers and merchants would not necessarily translate to a similar effect on Amazon Marketplace and vice versa.

Nevertheless, the persistence of the positive feedback loop on each separate platform can still lead to changes in homing patterns and, eventually, even in the tipping of markets in favour of each separate platform. This is particularly so if, after the removal of the tie, the customer group that was subject to the tie continues to use the previously tied platforms in parallel. Where the tie and its effects can extend across more than one customer group due to the complementary relationship between the previously tied platforms, tackling such an outcome is even more challenging since it requires changing the behaviour of multiple customer groups simultaneously. For example, in the early days of eBay, following its acquisition of PayPal, the use of PayPal was tied to the use of eBay for both consumers and merchants. Recently, this obligation has been modified into an optional choice for both, meaning the tie has been removed. However, getting consumers and merchants on eBay to switch from PayPal requires overcoming the indirect network effects present on PayPal and igniting such effects on a separate payment platform. This can prove to be very difficult in practice. Suppose consumers choose to continue using PayPal on eBay. In that case, there is no reason for merchants to stop accepting it, further reinforcing consumers' preference to stick to PayPal and vice versa. Such a lingering effect would be powerful where the tying practices created a consumer bias concerning the joint use of the previously tied platforms. This reality has been clearly understood by eBay, which voluntarily started actively phasing out PayPal from eBay a few years after the platforms split into two separate public companies.¹¹⁴ Accordingly, instead of letting consumers and merchants switch to other payment platforms, eBay actively excluded the possibility that PayPal would remain the default choice for such parties as eBay’s tying practices had previously intended.

The previous practice of the Commission in the Microsoft (WMP) and Google Android cases can be said to have dealt with similar circumstances. However, the remedies implemented in these cases were far less intrusive than the steps taken by eBay, which is perhaps also why they were not very successful. In Microsoft, the remedy allowed OEMs to choose between a version of Windows OS with Windows Media Player (WMP) and a version without it that was priced identically. Unsurprisingly, OEMs had little incentive to choose the version without WMP, which meant that consumers also continued to receive (and expect) their PCs to come with WMP. Consequently, the remedy failed to reset the status quo created by the tying practices, thereby allowing the harm to persist. Ironically, the (rejected) remedy offered by Microsoft, where WMP would be separated from Windows OS and provided on a separate CD,¹¹⁵ would have likely been more effective. Such a remedy would have removed the tie for OEMs and, at the same time, allowed consumers to reconsider their previously imposed default choice of having WMP, thereby potentially breaking free from their bias.¹¹⁶ Later in Microsoft (tying), concluded via a commitment procedure, the remedy overcame this shortcoming by requiring Microsoft to present consumers with a choice screen for their default internet browser once they turned on their PC for the first time.¹¹⁷ Nevertheless, implementing this remedy did not go smoothly, as Microsoft was later fined for undermining and minimizing its effects.¹¹⁸ The same approach can also be seen in Google Android, where consumers are presented with a comparable choice screen for their default internet browser despite the tying abuse initially directed at OEMs.¹¹⁹ Pursuing remedies that follow this logic in the case of online platforms will require continuously adjusting the design of remedies to their different technical and commercial characteristics that constantly evolve.

Generally speaking, cross-platform tying cases can take several forms. The tie can be technical in the sense that the use of one platform requires signing in with another platform. This would occur, for example, when using Instagram would require signing in with a Facebook account. In such a case, removing the tie would require creating other sign-in possibilities for Instagram for users to choose from, including through platforms competing with Facebook. Another form of tying would entail the data generated on the tying and tied platforms. This would occur if using Instagram would require agreeing with the sharing of data generated on Instagram to be shared with Facebook.¹²⁰ In such a case, removing the tie would require allowing users to choose whether they wish to do so actively, thus without a pre-programmed opt-in or out of such an option. A third way of tying would combine the previous two by requiring users to create a profile account that automatically creates additional profiles on a separate platform and enables sharing data across such platforms to feed the profiles made on both platforms. In such a case, removing the tie would allow users to refuse this cross-platform creation of profiles and data sharing across platforms when joining one of the platforms and allow such users to change their choice at any time.

Finally, the fourth option for implementing a tie between two separate platforms requires using two (complementary) platforms in tandem, as was the case with eBay and PayPal. In such a case, removing the tie would require other platforms that provide the complementary service as the tied platform to be accepted for integration and use with the tying platform. In the case of eBay and PayPal, this would mean requiring eBay to accept other payment solutions and that this choice can be made during every separate transaction. In the case of this last type of tie, where the tie also extends to other customer groups, the remedy must do so as well. Accordingly, in the example of tying a payment platform to an online marketplace, allowing consumers or sellers to use other payment platforms must occur simultaneously. Implementing such remedies in the case of cross-platform tying would entail translating the remedies accepted by the Commission in Microsoft (tying) and Google Android to the context of online platforms. Although this increases the chances that such remedies will be considered proportionate in future cases,¹²¹ it does not mean these will always be equally effective. This is particularly so in situations where the markets of the tying and tied platforms have tipped in favour of the dominant undertaking due to its tying practices.

In cases where tipping has occurred or is on the verge of occurring, solely removing the tie and presenting the platform customer groups with the possibility to make different choices actively may not bring much change in practice. If, by virtue of the tie, the concerned undertaking has become the undisputed leading platform in the product market of either the tying or tied platforms, presenting its customers with alternatives may not have much value. In such a scenario, more far-reaching structural or quasi-structural solutions may be required. These solutions can be either a full divesture of the tied platform to a third party or temporarily ceasing interoperability between the previously tied platforms. A divesture solution would prevent the concerned undertaking from continuing to profit from the positive feedback loops it managed to fuel through its tying practices for profit or strategic purposes.¹²² In the example of Amazon Marketplace, this would mean requiring Amazon to divest Amazon Pay to a third party, thereby preventing it from leveraging its market power from the retail market to the digital payment solutions market. Imposing a comparable measure would require the Commission to take a step it has rarely taken until today,¹²³ which may cast doubt on the possibility that such a measure would be considered a proportionate first pick under Article 7 of Regulation 1/2003. Nevertheless, in cases where tipping has occurred, and the effects of the abuse persist even after the tie between the platforms has been removed, one may argue that the structure of the undertaking prevents effective enforcement, thus justifying a structural intervention.

In cases where such structural separations are considered too burdensome or too risky for the Commission to attempt, a technical separation would entail a less permanent form of intervention that may potentially achieve similar results. Such a quasi-structural measure would require the dominant platform to switch off the interoperability between the previously tied platforms for a specific, reviewable period. Accordingly, in the case of a mutual sign-in system, this would require not only allowing the usage of other sign-in options but also removing the possibility of signing in on the tying platform with an account linked to the tied platform. In the previous example of Instagram, this would mean enabling users to sign in with their email account but not their Facebook account. Similarly, where platforms provide complementary services, the tying platform should be open to be used with other complementary platforms but not with the tied platform. In the example of Amazon Marketplace, this would mean requiring Amazon Marketplace to accept various payment systems but not Amazon Pay.

Although quite radical, such a remedy would effectively prevent the concerned undertaking from continuing to profit from the positive feedback loops it managed to fuel through its tying practices without intervening in its corporate structure. By imposing a technical separation of the two platforms, the positive feedback loops on such platforms can no longer feed off each other. Furthermore, by obligating the concerned undertaking to link the tying platform to competitors of the tied platform, the market power of the concerned undertaking in the tying market can be used to leverage the position of existing small players or new entrants in the product market of the tied platform. Similar to a divesture, this option also allows existing and potential competitors to compete with the tying platform without necessarily having to enter the tied platform’s market.¹²⁴ By doing so, a comparable remedy would possibly bring changes to the homing patterns of the platform customers, which in turn may make the tipped markets contestable once more. This can occur particularly when consumers opt for the tied platform solely because they want to use the tying one. Without the link between the previously tied platforms, customers can be expected to abandon the tied platform. This, in turn, will also likely reduce the adoption of the tied platform by other customer groups that opted for the tied platform only to reach such consumers, thereby also slowing down the positive feedback loop on either platform. Consequently, such a separation may significantly reduce (at least some of) the undeserved growth achieved by the dominant undertaking on both platforms.

It is worth noting, however, that despite the intrusiveness of this remedy, a technical separation does not mean it would put the tied platform out of business. eBay’s permanent separation from PayPal does not appear to stand in its way of further growth following the announcement that eBay will no longer support it.¹²⁵ Therefore, a similar measure, which is only temporary, is less likely to be disproportionately detrimental to the commercial state of the dominant platform. The period for which such a remedy should apply would differ from case to case depending on the nature of competition in the affected markets and should preferably include review possibilities where the Commission can withdraw the measure. In cases where the tying practices led to market tipping, such technical separations may nevertheless not suffice, and a divestiture would be required. In such cases, the network effects at play on each separate platform may continue to fuel their respective growth, thereby allowing the dominant undertaking to profit from its previous anti-competitive practices after termination. This can be seen to some extent in the case of Google, which continues to profit from the additional market power gained through its tying practices concerning the Android ecosystem. The remedy imposed by the Commission did little to curb any of the advantages it obtained, and Google continues to enjoy in the markets of its previously tied products, namely Google Chrome and the Google Search App.¹²⁶

Nevertheless, the mere fact that the above-mentioned suggested measures and the like could (in theory) all be proportionate under Article 7 of Regulation 1/2003 does not mean they can easily be applied. Opting for the most intrusive measures, even if desired, requires overcoming the limited ability to assess the effectiveness of milder remedies ex-ante, which, as will be seen, requires some legal creativity within the existing framework of Regulation 1/2003 and outside of it.

5. OVERCOMING REMEDY DESIGN LIMITATIONS OF REGULATION 1/2003

The discussion in the previous sections exemplified the multiple challenges involved in designing effective remedies for abuses of dominance by online platforms. Identifying these challenges is, of course, only half of the solution. Designing effective remedies requires overcoming those challenges by finding ways to tackle the risk of persisting (and evolving) competitive harm posed by the multisided character of online platforms in a manner that fits within the boundaries of Regulation 1/2003. While more radical (theoretical) solutions can always be found, the article focuses on those feasible within the current ambits of the law in light of ongoing cases and future ones that will face these challenges soon before Regulation 1/2003 is updated. Ideally, the future iteration of Regulation 1/2003 will allow more legal space for remedy design choices involving novel circumstances, as the evolution of technology and the digital economy will inevitably produce these in the future. Until that moment comes, it is worth considering three approaches to remedy design that could contribute to the effectiveness of remedies in the case of online platforms. These concern the strategic use of interim measures, the introduction of flexible remedies, and the reliance on other regulatory frameworks outside of (EU) competition law. The discussion of these three approaches will refer to the tying abuses examples addressed previously to consistently showcase their potential added value.

Strategic use of interim measures

The strategic use of interim measures combined with final ones can greatly contribute to implementing suitable behavioural and structural remedies in the case of platforms. Such a combination could allow for an ex-ante evaluation of the effectiveness of remedies that would otherwise be imposed as final ones. This, in turn, could allow the Commission to consider and impose structural remedies more easily when needed. Returning to the examples of on- and cross-platform tying, this approach would look as follows.

When practices that may qualify as abusive (on-platform or cross-platform) tying are identified, an interim measure consisting of the behavioural remedies mentioned in Section 3 can be imposed at a (relatively) early stage of the case. The period from the implementation of the interim measure to the point where a final decision on the existence of abuse can be made would then serve as a quasi-trial period for the remedy. If, during such time, the remedy fails to achieve its desired effect, namely slowing down the positive feedback loop on the respective platform(s) and concerning shift in homing patterns, far-reaching structural remedies could be considered more seriously. In the case of on-platform tying, such a remedy could require the previously tied functionalities on one platform to be separated across two stand-alone platforms by the same corporate entity or divested to a third party. In the case of cross-platform tying, a comparable remedy would entail that the two previously tied platforms no longer inter-operate or that one of them is divested to a third party.

Following such an approach would make the eventual implementation of structural remedies more likely to be proportionate as the trial period following the interim measure would prove that behavioural remedies are ineffective for tackling the competitive harm caused by the tying practices as required by Article 7 of Regulation 1/2003. If, however, the behavioural remedy imposed in the context of the interim measure appears to deliver the desired effect, it can be kept or slightly tweaked when the final decision on abuse is delivered under Article 7. Such an approach would align with previous Commission practices where interim measures preceded final decisions in abuse of dominance cases.¹²⁷

However, using interim measures under Article 8 of Regulation 1/2003 requires showing the risk of irreparable harm to competition and a prima facie finding of abuse. Although proving the existence of irreparable harm to competition would typically be quite cumbersome, one could argue that the tendency of platform markets to result in winner-takes-all outcomes where the market tips in favour of one platform entails such a risk.¹²⁸ If a prohibited practice is indeed capable of causing the market to tip in favour of the concerned undertaking, not intervening at an early phase in such a process may result in the elimination of competition in at least one market.¹²⁹ If that occurs, even a structural separation of the concerned platform at the final decision stage under Article 7 may not suffice to re-establish a state of competition.¹³⁰ This is because new entrants to markets that have tipped would face significant barriers to entry due to the network effects that remain in place even after the structural separation.¹³¹ Reversing the anticompetitive effects by re-establishing the state of competition in markets characterized by network effects has been acknowledged in case law to be extremely difficult, and a tipping situation can, therefore, be expected only to worsen it. The likelihood and urgency concerning the manifestation of such harm, which would justify intervention prior to a final decision, would depend on the effects of the investigated practices on the growth patterns of the respective platform(s) following their implementation.

Going back to the example of tying, if the introduction of (on-platform or cross-platform) tying practices results in an evident surge in the volume of both consumers and commercial customers on the respective platform(s) compared to the situation before the tying, it could be argued that the manifestation of (irreparable) anti-competitive harm is sufficiently foreseeable.¹³² This is particularly so if such changes are observed in combination with a shift in the homing patterns of these actors.¹³³ The finding of a prima facie infringement in such a context would require that the investigated practice at least appears to fulfil the criteria for which the Commission carries the burden of proof in tying cases.¹³⁴ The theories of harm in the context of abusive tying practices have been extensively studied in academic literature and covered several times in practice by both the Commission and EU Courts. Thus, such theories can hardly be considered novel, even when concerning platforms. Accordingly, finding a prima facie infringement in such a context should be less likely to be struck down at the stage of judicial review in a similar fashion to what happened in IMS Health.¹³⁵ The availability of such evidence will, however, depend on the specific circumstances of each case.

Utilizing interim measures in tandem with infringement decisions when dealing with platforms will help transform structural remedies into a feasible and effective solution in the most extreme abuse of dominance cases. This will be valuable not only to platform cases but also to the enforcement of competition policy as such, as it would help revive the interim measure procedure and the use of structural remedies currently considered theoretical rather than practical options. Nevertheless, although the theoretical reasoning behind this approach is relatively straightforward, putting it into practice effectively will not be easy due to its practical limitations and challenges. Firstly, the behavioural remedy imposed as an interim measure must be the most far-reaching for this approach to enable more use of structural remedies. In the tying examples given above, this would entail the quasi-structural remedies discussed. Otherwise, the following best remedy in line for consideration at the final decision phase will still be behavioural. This leads to a dilemma. On the one hand, structural remedies will evidently not be needed (or possible) in every case, so leading with the most far-reaching remedy as an interim measure may go too far and, therefore, risk annulment on the other hand not going forward with the most far-reaching behavioural remedy as an interim measure risks not being able to reach the stage of structural remedies despite these being needed to achieve the tri-fold objective of remedies. It is unclear which of the two options would be preferred. Secondly, if such an approach is implemented, the duration between the interim measure and the final decision on abuse must be managed with diligence. If the duration is too short, there may be a risk of remedies becoming (overly) strict as it is (relatively) less likely for the interim measure to show a representative effect suitable for indicating the way forward. Conversely, if the duration between the interim measure and the final decision becomes too lengthy, concerns about proportionality and the length of the procedure may arise. Thirdly, implementing such an approach in practice requires the Commission to use interim measures for an additional, related, yet different purpose than these were intended to serve. Against this background, it can be said that even though the strategic use of interim measures may be sound and justifiable from a legal theoretical perspective, the combination of the abovementioned concerns and legal challenges posed by the wording of Article 8 of Regulation 1/2003 limit their practical implementation.

Flexible remedies

An alternative solution to the strategic use of interim measures that could achieve a similar result has been discussed and, to some extent, applied in the context of EU merger control in the form of so-called ‘flexible remedies’. Flexible remedies entail a more extensive menu of measures consisting of multiple intervention layers that are triggered depending on how market conditions evolve after the remedy is imposed.¹³⁶ By introducing remedies, which consist of measures that apply only if and when specific competitive concerns materialize or at least become more apparent, the Commission (or NCAs) would be better able to deal with the dynamic nature of competition in the context of online platforms. Accordingly, following an abuse of dominance, the Commission could impose a remedy with a behavioural measure as a starting point. If the concerned abuse amplifies the network effects at play and causes changes in the homing patterns of the platforms’ customer groups, which do not appear to be affected by the respective measure, a more intrusive measure may be triggered later. Such a mechanism can proceed to introduce more intrusive measures in correspondence with the growing tendency of the affected market to tip in favour of the concerned dominant platform due to its abusive behaviour. Such an intensifying scale going from the least intrusive measure to the most intrusive one could include both behavioural and structural measures, which are triggered based on the developments in the affected markets over time.

Returning to the example of on-and-cross platform tying, the initial remedy layer in such a construct would entail simply removing the contractual or technical tie while including a restorative element, as was done in the Microsoft (tying) and Google Android cases. If such measures do not manage to slow down the positive feedback loop on the respective platform(s), readjust the homing patterns, and challenge an undesired bias, the following layers can go further. These layers would predominantly focus on challenging the (consumer) bias, for example, by further intervening in the display and interoperability setting of the interactions provided by the respective platform(s). At the end of the intensifying scale, the ultimate option of divesture measures would be included as a last resort measure and as a deterrent element against the minimization of the previous layers by the concerned platform(s).

Consequently, in the context of flexible remedies, the structural and most radical forms of intervention would only be imposed when market conditions justify their use, namely where market tipping appears imminent or, in the worst case, has already occurred. Going back once more to the case of tying, such an approach would allow the Commission to go beyond what was done in Microsoft (tying) and Google Android if needed while remaining proportionate at the same time.

Implementing flexible measures that combine behavioural and structural elements needed to achieve the tri-fold objective of remedies, although certainly unconventional, could be accommodated within the existing framework of Article 7 of Regulation 1/2003. First, the provision itself is not formulated in a way that would prevent the implementation of such a mixed remedy, nor would such remedies be entirely exceptional. In the context of EU merger control, commitments made to enable a transaction to go through commonly include both behavioural and structural measures. Secondly, such a mix would also be more compatible with the principles of effectiveness and proportionality than a rigid measure, as it would allow for a more customized approach to the harm created by the abuse until it is removed. Accordingly, it would also increase the chances of such remedies to achieve their tri-folded objective. Furthermore, the innovation-related concerns often voiced with regard to competition law interventions in the case of platforms and digital markets, in general, would be mitigated as such intervention would only be reserved to the cases where it is necessary and its scope limited correspondingly.

The theoretical potential of such measures does not guarantee, however, implementation in practice. Designing remedies with multiple layers that apply depending on different market conditions or developments entails inevitably predefining an exhaustive set of circumstances attached to each of the respective layers. Returning to the example of on-and cross-platform tying, this would entail having a benchmark for undesirable homing patterns, which would require further intervention and a corresponding strategy. Doing this in practice will inevitably require balancing legal certainty and effectiveness. Suppose each enforcement layer’s ‘trigger’ scenarios are vaguely described and defined. In that case, legal certainty concerns may arise for the concerned dominant platform, as it would not be able to assess when additional, more intrusive measures may be triggered. At the same time, the more accurate such scenarios become, the less flexible the remedy becomes with respect to its ability to deal with unforeseen developments, which may make the entire remedy inapt. Consequently, it is imperative that the Commission has a clear view of the potential future concerns it wishes to tackle but is unsure whether these will materialize.

A workaround for this problem would be to introduce a review clause that would allow for the impact of the imposed remedy layer(s) to be reviewed in light of the market conditions at the time of review and for adjustments. This would avoid situations where the predefined measures are undermined by market changes not foreseen when the remedy was defined, as occurred in the case of EDF/Segebel,¹³⁷ where unforeseen market developments de facto diminished the alternative options included in the remedy.¹³⁸ Despite such difficulties and the inevitable challenge involved in ongoing monitoring comparable measures, the added value of utilizing flexible remedies in the context of online platforms (and even beyond) would be significant. Of course, if such measures were to be utilized, the legal boundaries of effectiveness and proportionality would apply to the flexible remedy in its entirety and to each layer (with its corresponding scenario) separately since each of these could become the final one. Accordingly, it requires that the depicted scenarios not only describe well-defined, realistic market developments but also that the corresponding remedies are very well matched with the theories of harm in such scenarios.

When comparing this alternative solution to the strategic use of interim measures, as previously mentioned, it can be argued that flexible remedies offer a more suitable enforcement tool that is more likely to achieve the tri-fold objective of competition law remedies. Contrary to interim measures, the hurdles that need to be overcome in the application of flexible remedies are practical rather than legal-procedural ones. Nevertheless, the lack of preventive intervention as done with interim measures may mean that the harm that must be addressed in such a context may be relatively more significant. In this regard, perhaps the best approach would be to utilize both means in tandem. Interim measures would then be used, as intended, to prevent market conditions from significantly deteriorating, and the flexible remedies introduced at the stage of the final decision on abuse would strive to bring the infringement to an end, prevent its repetition, and restore or re-establish the state of competition. Should both approaches mentioned so far fail the test of reality, the third approach to enhancing the effectiveness of competition law remedies for online platforms is coordinating such remedies with other regulatory frameworks applying to these actors.

Competition law remedies and regulatory frameworks: the role of the DMA

The interplay between competition law and regulation is a challenging and multifaceted one.¹³⁹ The obligation to account for regulatory constraints in the process of applying EU competition law continues to intensify due to the growing importance of the more economic, effects-based approach in EU competition law, which demands that the entire legal and economic context of each case is accounted for.¹⁴⁰ Only when such context is complete can the legality of the investigated practice be assessed in light of its likely impact on the objectives pursued by EU competition law, including but not limited to consumer welfare, total welfare, market access, and the integrity of the EU internal market. This approach can be seen through the entire EU competition law spectrum covering EU antitrust rules and EU merger control. Commonly, the reference to the legal and economic context is directed to the integral role of regulatory frameworks in assessing the permissibility of the pursued commercial behaviour or envisaged concentration.¹⁴¹ A much less explored but equally important role played by regulatory frameworks as part of each case’s legal and economic context concerns designing the corresponding remedies to tackle the identified anti-competitive effects. In this context, the respective regulatory framework that applies to the facts of the case in tandem with competition law shapes the competition law remedy. This interplay between competition law and regulation can be best seen in the context of EU merger control. In the acquisition of Sanofi by Google, for example, data protection legislation was used as an alleviating argument for the competitive concerns identified by the European Commission.¹⁴² By contrast, in the acquisition of Fitbit by Google, data protection legislation was considered insufficient to remove the identified (data-related) concerns of the proposed transaction, thereby requiring data-specific remedies.¹⁴³

Returning to the specific context of online platforms, the most directly relevant regulatory frameworks that can be used in such ways are the Digital Services Act,¹⁴⁴ the Digital Markets Act (DMA),¹⁴⁵ and the Platform to Business Regulation (P2B).¹⁴⁶ As this article focuses on abuses of dominance by online platforms through tying and bundling practices, the discussion in this section will focus only on the DMA that covers explicitly various business practices of online platforms with significant market power that could be identified as such under Article 102 TFEU. Nevertheless, in practice, when the potentially prohibited practices of a dominant online platform may be covered by the DSA, P2B, or any other regulatory framework, the insights in the section can, to some extent, be applied by analogy.

According to the text, the DMA aims to complement competition policy and help create fair and contestable markets in the digital sector, which may sometimes be undermined by the business practices of providers of platform services that possess a high degree of market power.¹⁴⁷ For this reason, the DMA is also designed to apply solely to actors with control over the digital sector's gateways. Accordingly, the application scope of the DMA is limited to providers of a core platform service that have reached the gatekeeper status. The term core platform service does not cover all platforms, nor is it limited solely to services that meet the criteria of multisided markets as defined by economic literature.¹⁴⁸ Instead, the term core platform service, defined in Article 2 of the DMA, refers to a non-exhaustive list of commercial services commonly provided by multisided platforms. A provider of a core platform service may qualify as a gatekeeper when it (i) has a significant impact on the internal market, (ii) serves as an important gateway for both consumers and business users to reach each other, and (iii) enjoys an entrenched and durable position in its operations or it is foreseeable that it may reach such a position in the near future.¹⁴⁹

The criteria of Article 3 of the DMA are inspired by the notion of dominance under Article 102 TFEU. However, unlike a finding of dominance, the threshold of gatekeeper does not require defining the relevant market, which is problematic in the case of platforms because of their multisided nature.¹⁵⁰ Instead, the DMA relies firstly on the predefined quantitative criteria in Article 3 that are assumed to capture significant degrees of market power, with the possibility to rely on qualitative criteria instead where the respective platform does not meet these benchmarks.¹⁵¹ Once the designation is made the undertaking concerned is bound by the obligations set out in Articles 5–7 of the DMA, which can play an important role in the design of effective restorative remedies for tying practices.

The DMA’s added value for designing effective remedies for abuses of dominance under Article 102 TFEU by online platforms depends on whether these two legal frameworks intersect with respect to the same platform undertakings, which is case-dependent. Such intersection will first depend on whether the concerned platform offers a core platform service as defined by the DMA. Secondly, the added value of such intersection will depend on whether the respective platform has been designated as a gatekeeper under the DMA and qualified as dominant under Article 102 TFEU. Although the two qualifications relate to undertakings with significant market power they do not seamlessly overlap. Dominance is established based on the market power of the undertaking concerned in relation to that market in which it is active and with respect to its direct and potential competitors.¹⁵² This is also arguably the case with the qualitative benchmark of the DMA but not so much with respect to the quantitative one, which is based initially on evidence of financial power and market reach that does not always portray an accurate image of market power within the ambit of competition policy. With these considerations in mind, the relationship between the DMA and Article 102 TFEU can be divided into three possible scenarios that can offer different benefits for remedy design: (i) the dominant platform is also a gatekeeper, (ii) the dominant platform is not a gatekeeper (yet), and (iii) the gatekeeper platform is not dominant. The designation decisions by the Commission and its practice so far can be said to cover all scenarios.¹⁵³

In the first scenario, the DMA will be the most relevant for remedy design. In such situations, the DMA will introduce specific obligations that may complement the obligations imposed following an abuse of dominance finding or otherwise contribute to the design of remedies that need to be imposed once an investigation of abuse reaches the final stage. Accordingly, if an abuse of dominance is caused by a platform that is also a gatekeeper, the remedy imposed for such abuse will have to take into account the obligations that already apply to it under the DMA. The competition law remedy could then complement the DMA’s (general) obligations by more tailor-made measures directed, for example, at the consumer bias that the abusive practice may have created. This may allow a more comprehensive package of measures, which would otherwise be impossible as a competition law remedy under Article 7 of Regulation 1/2003. Where the abuse of dominance concerns a practice that was meant to be covered by the DMA, such as a specific form of cross-platform tying, the competition remedy for such abuse may take the eventual shortcomings of the DMA into account. This, in turn, may increase the likelihood that more far-reaching remedies will be considered proportionate in the context of Article 7 of Regulation 1/2003. If a dominant platform abused its position despite the behavioural obligations it faced under that DMA, it would be harder to argue that similar remedies in the context of competition law proceedings would be effective, consequently making a stronger case for the implementation of stricter behavioural remedies and even (quasi-) structural remedies.

In the second scenario, where the dominant undertaking is not (yet) a gatekeeper, the DMA may serve as a safety net or reinforcement for the competition law remedies in the long run. Accordingly, the behavioural or structural remedies implemented following a final decision of abuse may be reinforced or reintroduced later under the DMA.¹⁵⁴ Of course, such a development may not always be a positive outcome since it implies that in some situations, the already dominant platform has continued to grow, following its abusive practices and corresponding remedies, to the point it has also become a gatekeeper under the DMA.¹⁵⁵ Nevertheless, if the dominant platform qualifies as a gatekeeper because the service matchmaking functionality it offers counts as a (new) core platform service under Article 2 of the DMA, this transition may not necessarily imply a worsening of competition in the market(s) where the concerned platform is active. Accordingly, when such a specific evolvement occurs, the obligations in the DMA may complement the effectiveness of the competition law remedies to the extent these are not the same. In such a scenario, the competition remedy should be designed so that, should it fail, the corresponding obligation from the DMA would serve as a safety net. Accordingly, in such cases, there would be less urgency to turn to (quasi-) structural measures following a finding of abuse. In essence, the DMA obligation in such a context would serve the same purpose as a second layer of a flexible remedy.

In the third scenario, where the gatekeeper platform is not dominant for the purposes of an Article 102 TFEU investigation, the DMA will not be directly relevant for remedy design, as in the absence of dominance, no legal intervention is possible under this provision.¹⁵⁶ Nevertheless, the effects of the DMA’s obligations can be used as material for imposing remedies on the gatekeeper platform if it also becomes dominant and its business practices breach Article 102 TFEU. In this regard, if the DMA’s obligations for the dominant platform, which applied due to its earlier gatekeeper status, did not prevent it from abusing its market dominance later, it could be argued that a more intrusive and perhaps structural remedy under Article 7 of Regulation 1/2003 may be easier to justify. In this regard, the period of application of the DMA’s obligation on the gatekeeper platform can be seen as a test period for the effectiveness of similar remedies that could be imposed in the event of abuse once the gatekeeper platform becomes dominant. Consequently, such cases will require less experimentation with interim measures to find the most effective remedy at the final decision stage. At the same time, this test period may allow for using interim measures more often since the transition from a gatekeeper platform to a dominant one would indicate a market tipping tendency that requires immediate intervention to prevent irreparable harm.

Finally, it should be noted that the implications, as mentioned earlier, of the DMA for remedy design concerning abuse of dominance cases are currently relevant to the limited scope of core platform services and the respective theories of harm that were the basis for the obligations included in Articles 5 and 6. Many such obligations concern the prohibition of various forms of on- and cross-platform tying similar to those discussed in this article, drawing on past experiences with tying and refusal to supply cases.¹⁵⁷ When it comes to pricing practices, however, the relevance of the DMA is currently more limited as it appears to tackle very few potential pricing problems that could arise in the context of platforms. The most prominent pricing-related practice concerns prohibiting wide price parity clauses,¹⁵⁸ which has been the subject of procedures in the context of hotel room booking platforms and Amazon’s eBooks.¹⁵⁹ Beyond this, the matter of pricing primarily concerns access to the data generated by and on the platform providing the core platform service.¹⁶⁰ Such matters can relate to the Commission’s experience in previous merger cases where the respective concentrations raised potential concerns about the data possessed by the concerned undertakings.¹⁶¹ Obligations that would simulate more ‘traditional’ price-related abuses under Article 102 TFEU are, however, not included. This situation is arguably justified by the fact that the Commission’s experience with online platforms is relatively limited for now. Accordingly, including ex-ante obligations that regulate such matters would have faced significant resistance, and for good reason. Nevertheless, such implications may gradually diminish in the course of future updates of the DMA.¹⁶² However, this would first require the Commission to have experience with additional theories of harm concerning online platforms, which would then be translated into additional ex-ante obligations that could apply to a more extensive selection of core platform services. In practice, this may take a significant amount of time as such experience requires not only reaching the stage of a final decision on abuse by the Commission but also ensuring that such a decision survives the stage of judicial review by EU courts. In the case of Google Shopping and the idea of self-preferencing as a stand-alone abuse, such a clarification process has been ongoing since the initiation of the Google Shopping case, which started off with self-preferencing being an example of anti-competitive leveraging. It is only recently that this possibility seems to have become a realistic prospect with the opinion of AG Kokott.¹⁶³ The final stage of this process thus remains to be reached, approximately seven years after the Commission delivered its decision on the matter.

Admittedly, there is no legal obligation to follow this process. New obligations can be inserted into the DMA even without prior practice under Articles 101 or 102 TFEU. In fact, in certain situations, the difficulty in bringing certain practices under the ambit of these provisions could be the reason for the inclusion of such (novel) practices in the DMA (or any other similar regulatory framework). Nevertheless, doing so may create tension with the motivation behind the DMA, which is also included in its recitals, namely, to tackle practices that raise competitive concerns based on (prior) experience. In the absence of actual case practice experience, the motivation behind such new additions should, at the very least, concern practices that clearly touch upon contestability and fairness in the realm of digital platforms. Typically, such practices should also be considered contrary to the rationale or objectives of Article 101 or 102 TFEU so as to stay true to the initial stance of the DMA. Alternative approaches, while legally possible, can be expected to be (rightfully) met with significant objection.

6. CONCLUSIONS AND FINAL REMARKS

The design of effective and proportionate remedies in the case of online platforms is a complex matter. For remedies to be effective, they must terminate the abusive practice, prohibit their repetition, and introduce a restorative component designed to restore or re-establish the state of competition undermined by the abuse. The last component is essential in the case of abuses of dominance by multisided online platforms. Although the nature of the competitive concerns that can arise from the abusive behaviour of such players is similar to traditional market settings, the manner in which such concerns manifest in practice is quite different as it can extend to more than one market due to their multisided nature. This, in turn, also requires remedies to extend across markets correspondingly. When doing so, remedies should account for the impact of the abusive practices on the positive feedback loop(s) on the concerned platform(s) and the user homing patterns to prevent competitive harm from persisting and even amplifying after the respective abusive behaviour is abandoned. Achieving this in practice necessitates introducing restorative components at the remedy phase. In the absence of such a possibility, there will be a luring risk that market tipping will take place in favour of the dominant platform. Once that occurs, competing platform (and non-platform) undertakings will no longer be able to successfully compete with the dominant platform even when the latter does not resort to other strategies than competition on the merits.

The current legal framework of EU competition policy does, in theory, provide the Commission with the tools it requires to design effective remedies in the case of platforms, despite the challenges posed by such actors. However, the manner in which this framework has evolved and has been used in practice can make the implementation of effective remedies in the case of platforms very challenging, particularly when structural remedies would be required. The adoption of such measures is limited to situations where behavioural remedies cannot attain similar results and is subject to the principle of proportionality. This combination of requirements makes the implementation of structural remedies as a first-choice solution highly unlikely, even in the case of platforms. This is confirmed by the Commission’s past practice, which shows little to no traces of structural interventions in the case of abuses of dominance, in stark contradiction to its practices in the case of EU merger control.

Moving away from this impasse, heading towards stricter measures, and even tapping into the possibility of imposing structural measures within the existing framework of Regulation 1/2003 requires more experimentation by the Commission with interim measures and/or flexible remedies. Such experimentation would allow for overcoming the difficulty of assessing ex-ante the effectiveness of remedies before their imposition and thereby create more legal space for the implementation of more far-reaching measures when needed. Although both approaches to remedies are theoretically possible within the ambit of regulation 1/2003, the practical hurdles involved may prevent them from materializing in the near future. To some extent, this may be understandable when trying to make strategic use of interim measures, which also raises the question of whether such use aligns with the purpose and objectives of such remedies. However, in the case of flexible remedies, this would be a missed opportunity. Passing on both approaches would be a rather unfortunate outcome in the case of online platforms where the Commission’s current practice when it comes to remedies will not consistently deliver the desired result.

Given the challenging dynamics of the current legal framework, it can be said that the arrival of the DMA entails a welcome addition. The DMA could complement the remedies that the Commission would impose following a finding of abuse by a dominant platform and, at times, even prevent such platforms from implementing abusive behaviour. As it is an entirely new tool, its application in practice would be less restrained than the current competition law framework, thereby allowing for more strategic use of the two frameworks to merge to deal with challenging cases involving platforms. However, the DMA’s complementary nature will often be limited in scope due to its jurisdictional thresholds and set scope of substantive obligations, which cover only a limited number of competitive concerns. Nevertheless, such complementary relations between the DMA and competition practice may evolve as the use of qualitative criteria for gatekeeper designation are used and the scope of obligations, as well as core platforms services covered, are periodically updated. Accordingly, the adoption of the DMA may help increase the effectiveness of competition law remedies by paving the way to the use of more far-reaching remedies, which would be required for achieving their tri-fold objective that is of great importance when dealing with abuses of dominance by multisided online platforms.

Footnotes