Regulating digital platforms as the new network industries

Direct, indirect and algorithmic network effects

It is today widely accepted that network effects play a fundamental role in digital markets. This fact has been identified early on by academics like Shapiro and Varian (1999), who in 1999 published their influential book entitled Information rules. A strategic guide to the network economy. Another significant intellectual milestone was Rochet and Tirole’s paper on two-sided markets already back in 2003 (Rochet & Tirole, 2003).

We propose to distinguish between three types of network effects in digital markets: direct, indirect and what we can call algorithmic network effects. Direct network effects exist when the value of a product or a service is affected by the number of users. Positive direct network effects are well known to academics working on network industries, as they were identified more than a century ago by the early telephone managers: “A telephone without a connection at the other end of the line is not even a toy or a scientific instrument. It is one of the most useless things in the world. Its value depends on the connection with other telephones and increases with the number of connections” (AT&T Annual Report for 1908, p. 21). The same network effect plays in digital networks such as messaging services (WhatsApp), social networks (Facebook), and videoconferencing services (Zoom): “I think that network effects shouldn’t be underestimated with what we do” (Zuckerberg in Ammirati, 2016). Digitalization further enhances direct network effects as the new networks connect not only millions but billions of individuals. As described by Zuckerberg when Facebook acquired WhatsApp: “if you look at the number of things that have reached a billion people, they all end up being incredibly valuable and important things” (Zuckerberg in Levy, 2020, p. 215).

Indirect network effects play an even more important role in digital markets, which are typically multi-sided. As digital technologies reduce transaction costs, new and more ambitious complementarities can be pursued. Technology empowers the interaction not only among homogenous groups (such as telephone users), but also among heterogeneous groups, as Rochet and Tirole in their seminal paper have so eloquently conceptualized. Interestingly, Tirole had originally worked on, among other things, the regulation of network industries, particularly telecommunications. In fact, Rochet and Tirole explicitly state that their theory builds on network economics. Their premise was that “many (probably most) markets with network externalities are two- (or multiple-) sided markets. A market with network externalities is a two-sided market if platforms can effectively cross-subsidize between different categories of end users that are parties to a transaction. That is, the volume of transactions on and the profit of a platform depend not only on the total price charged to the parties to the transaction, but also on its decomposition” (Rochet & Tirole, 2003, p. 1018).

Over the years, a consensus has developed that multi-sided markets typically involve two or more distinct types of users who interact via a third party, namely the “platform” (OECD, 2009). Key elements are the presence of positive indirect network effects, which have to be strong enough to affect business conduct (Evans & Schmalensee, 2016), as well as the leading role that the platform plays in distributing the benefits among the cross the interacting parties thanks to pricing and other design decisions. Digital champions such as Uber describe themselves as a heterogeneous network: “Our massive, efficient, and intelligent network consists of tens of millions of drivers, consumers, restaurants, shippers, carriers, and dockless e-bikes and e-scooters, as well as underlying data, technology, and shared infrastructure” (Uber Form S-1 for the IPO, April 11, 2019). Actually, creating network effects on top of previously fragmented assets is the business model in the most successful Silicon Valley ventures: “We had an internal thesis that other industries might benefit from a network layer on top of them” (Bill Gurley, Partner in Benchmark Capital, in Lashinsky, 2017, p. 102).

Finally, algorithmic network effects result from the analysis of ever larger amounts of data by ever more sophisticated and ever more powerful algorithms. And these algorithms are ever more powerful precisely because they are trained on ever bigger amounts of data, produced by ever growing numbers of users. The European Commission described such algorithmic network effects for the example of search engines: “because a general search service uses search data to refine the relevance of its general search results pages, it needs to receive a certain volume of queries in order to compete viably. The greater the number of queries a general search service receives, the quicker it is able to detect a change in user behaviour patterns and update and improve its relevance” (Commission Decision of 27.6.2017, Google Search (Shopping), para. 287). In short, such algorithmic network effects are central in digital markets: with more users, more data is harvested, the more efficient the algorithm becomes, and the better the service gets for the users, which in turn will attract more users Interestingly, Uber presented itself in its IPO not simply as a network, but as “a network that becomes smarter with every trip” (Uber Form S-1 for the IPO, April 11, 2019).

Platforms, concentration and market power

Platforms as creators and curators of new and massive network effects have grown increasingly relevant over the past two decades. For some, this was an unexpected evolution of the internet, which was created and designed to be a fully distributed network with no hierarchies. However, for others, the open internet has precisely fostered a unique opportunity to capture the value created by the new network effects thanks to the internet. And network effects, as shown above, are leading to market concentration so that a very small number of digital platforms are monopolizing the value created by network effects. In this way, the most successful platforms have risen to unprecedented economic power.

To recall, the internet was the product of a unique collaboration between governmental authorities (mostly the US Department of Defense) and academia, led by a small group of computer scientists. From 1969, when the first connection between UCLA and the Stanford Research Institute was established, to 1995, when the number of internet users was around 45 million, the internet was basically a non-commercial venture. It was purposely built as a distributed network with no central control, no hierarchies, and open to everyone, with no specific purpose and no aim to monetize the immense network effects it made possible. These origins misled the first theorists studying the societal impact of the internet to believe that the internet would lead to a society, including an economy, with much more distributed power, characterized by a non-hierarchical network of decentralized, yet interconnected social actors, as so optimistically described by Castells in The Rise of the Network Society (Castells, 1996).

Nonetheless, starting around 1995, a group of individuals clustered in Silicon Valley understood that the interactions in the internet could be structured in specific mechanics and for a profit. “[T]here are network effects around social products and a finite number of different social mechanics to invent. Once someone wins at a specific mechanic, it’s difficult for others to supplant them without doing something different.” These words from Zuckerberg are repeatedly quoted in the antitrust suit filed by the US Federal Trade Commission against Facebook in December 2020.

Twenty years later, some academics, especially the ones from business schools, focused their attention on how platforms could be specifically designed to create and capture network effects as a business model (Cusumano et al., 2019; Parker et al., 2016). Platforms could actively build multisided markets around them by leveraging potential complementarities, which would then become as many new sources of new direct, indirect network and algorithmic effects. Targeted investments could be made so as to build the scale necessary to trigger and enhance such network effects, for example by attracting new groups of users, often through discounts, subsidies, and other incentives. In addition, platforms could actively manage the multisided market, by curating the multi-sided market, thus dynamically deciding on the distribution of the value created by the network effects to the different parties in the broader ecosystem. Sociologists had already identified decades earlier that “Inventors, organizers, and managers of technological systems mostly prefer hierarchy, so the systems over time tend towards a hierarchical structure” (Hughes, 1987, p. 49). They had come to this conclusion by analyzing the evolution of one of the traditional network industries: electricity. But the very same dynamic is now taking place in the new network industries, spurned as it is by the very same underlying network effects.

Industries where network effects are at play tend toward concentration, as a critical mass of users or more generally partners in the network is necessary to make network effects effective, or in other words “to ignite the platform.” Indeed, scale is always necessary. Consequently, mature multi-sided markets will always have a limited number of competitors. Under some circumstances, the market might tip into domination by a single firm. Both the traditional and the new network industries present a similar development path. At an early stage of industry development, a larger number of companies might compete in the market. This was the case of the early telephony or electricity markets in the US, and it was the case of many digital markets, from social networks to delivery and ride-hailing apps. Such competition actually accelerates market growth, as competitors reduce prices and they even subsidize services in order to grow scale and market share.

But after such an initial phase, network markets tend toward concentration. As the market matures, competitors with less financial resources cannot continue subsidizing growth, they grow at a slower pace than their competitors and the value of their network effects is outpaced by the larger competitors. As a result, the market consolidates around a small number of firms. This trend is often accelerated by the fact that competitors with the deepest pockets (that is with potent investors behind them) often enter into an acquisition spree in what has been described as a Genghis Khan strategy “join the network and share the wealth, or face annihilation” (Wu, 2010). Monopolization if often the result of aggressive business practices so as to prevent competitors from growing, for example by predatory pricing in the more competitive market segments, exclusivity clauses prohibiting customers to contract with smaller competitors, artificial barriers to switch providers, strategies that could already be identified as having lead to the rise of the telephone monopoly in the early 20th century in the US. Concretely, the objective of such business practices is to limit the possibilities of unlocking network effects by the smaller competitors (Mueller, 1997).

Some digital markets have grown to be dominated by one single company, such as Google in searches, Facebook in social networks, and Amazon in marketplaces. Other digital global leaders face more competition, especially in geographically defined local markets, such as for example Airbnb in accommodation or Uber in ride hailing. Actually, the traditional network industries show similar market structures.

Key will be whether network effects get exhausted after reaching a certain threshold or whether they can grow almost indefinitely parallel to the increase in platform users. In the first case, the market be somewhat competitive in some form of tight oligopoly. Regulatory remedies, in this case, will pertain to fostering competition by reducing barriers to entry as well as by pushing for interoperability. In the second case, the market will tip into a winner-take-all dynamic. The key regulatory question here is whether market dominance by a platform is the result of a winner-takes-all evolution or whether it results from an anticompetitive strategy by the market leader. Regulatory remedies, in this case, will essentially be structural in nature.

Data portability

Data portability is an instrument to reduce these barriers to market entry which derive from lock-in effects. Reducing switching costs is indeed a successful strategy to further competition in deregulated network industries: regulators can identify lock-in effects and eliminate them. Number portability in the telecommunications industry can considered to be an early version of regulatory intervention to reduce lock-in effects. Regulators had identified that telephone users were reluctant to change carrier if they also would have to change their telephone number. Consequently, regulators obliged the carriers to work together so as to ensure that customers would have the right to take their telephone number with them when switching from one carrier to another.

Data is the raw material in the digital industries and thus data portability has been identified early on as one of the barriers for customers to switch platforms, thus leading to a barrier to entry for newcomers. Users might have spent years feeding their social network profiles with pictures, videos, comments, etc. They will hardly be in the position to switch to an alternative social network if they cannot take such content with them. The same problem is faced by all kinds of applications fed with user data, such as applications tracking exercise, sleeping patterns, and so on.

Data portability is clearly inspired by number portability regulation and has thus been proposed as a remedy for this obvious lock-in effect. Platforms would be obliged to create a feasible solution for customers to take their data with them when leaving to another platform. As a matter of fact, data portability is already included in the EU General Data Protection Regulation (GDPR) when stating: “The data subject shall have the right to receive the personal data concerning him or her, which he or she has provided to a controller, in a structured, commonly used and machine-readable format and have the right to transmit those data to another controller without hindrance from the controller to which the personal data have been provided” (Art. 20). But data portability under GDPR only applies to data uploaded by users and not to data that is inferred about them, which is most likely the majority of the data that platforms possess about their users anyway.

However, data portability requires a very sophisticated management of data, coordinated across platforms by setting common standards. Furthermore, it is necessary to structure and standardize the data in a way that can be processed by the new platform. Regulatory intervention is obviously indispensable to make data portability a reality.

Multi-homing

The promotion of multi-homing and the prohibition of exclusivity is another pro-competition regulatory intervention. By multi-homing one means the parallel and simultaneous usage of competing platforms by the same user. Be they services providers (property owners in the case of Airbnb), final users (the guests in the case of Airbnb), or both. Hosts often display their accommodation units in Airbnb and other competing platforms at the same time. Guests often look for accommodation on Airbnb but also in competing platforms. The term “multi-homing” has its origin in computer lingo, when a computer was connected to more than one network, usually to increase reliability.

Multi-homing reduces barriers to entry, as newcomers can build scale by selling services to customers already engaged with another platform. Multi-homing enables newcomers to replicate the scale of incumbents, as there is no obstacle to sign as many customers as the largest platform. As a matter of fact, all market players can build similar network effects by pooling together the same customers.

Multi-homing imposes a burden on users (both service providers and consumers) as they must make concurrent use of more than one system. They have to work with two or more applications, learn to use them, take on the potential costs, etc. Sometimes this effort can be perfectly assumed, as multi-homing is simple and does not trigger high transaction costs. This is also the case of app developers, who tend to work both with IOS and Android to make their apps available to all smartphone users. The burden on users is sometimes so onerous that multi-homing is not feasible. Social network users are also reluctant to multi-home, as uploading information and replicating the upload of the same pictures, stories, etc. is perceived as too time-consuming and of little interest.

But, sometimes multi-homing is not feasible because the dominant platform actively disincentives it. Dominant platforms can obstruct multi-homing by imposing exclusivity on customers, mostly to services providers, even though services providers usually prefer to work with more than one platform, so that they do not depend on an exclusive intermediary to commercialize their services. However, facing the choice to exclusively work with the market leader (with the largest pool of users) or be excluded from the largest pool of users and contracting only with a smaller platform, service providers will rationally choose the largest player. Such behavior accelerates market concentration, reaching a tipping point when network effects enjoyed by the largest platform cannot be replicated by competitors. But there are more subtle ways to impose exclusivity: platforms can grant rebates to customers, such as to guests in accommodation platforms, incentivizing to grow the volume of business made with one platform. Such rebates disincentivize working in parallel with another platform.

Competition authorities have already identified exclusivity clauses imposed by platforms as being the most obvious threat to competition in multi-sided markets. Consequently, the most common remedy in platform mergers has been to prohibit exclusivity clauses to facilitate multi-homing. This has been common in food delivery (Decision of the Spanish Comisión Nacional de los Mercados y la Competencia, Just Eat/La Nevera Roja, 3.31.2016) and ride-hailing mergers (Decision of the Competition and Consumer Commission of Singapore, Grab/Uber, of 4.13.2018.). Today, this obligation is expanding beyond mergers as a regulatory obligation imposed upon platforms beyond certain thresholds in terms of revenue, number of users or market shares. For example, the European Commission, in its Digital Markets Act proposal, includes an obligation on gatekeepers to “allow business users to offer the same products and services to end users through third party online intermediation services” (Art. 5(b)).

Data openness

Data openness is another regulatory instrument to reduce barriers to entry an as such proposed in the Furman Report (Furman et al., 2019). As described above, algorithmic network effects rely on the availability of massive datasets to be used for machine-learning purposes. Access to such datasets is a competitive advantage, such as, for example, in the search market. Sharing datasets would certainly reduce the barrier to entry for newcomers and allow them to challenge the position of incumbent platforms.

The European Commission is proposing this obligation for the search market, as the market leader, Google, would be obliged to: “provide to any third party providers of online search engines, upon their request, with access on fair, reasonable and non-discriminatory terms to ranking, query, click and view data in relation to free and paid search generated by end users on online search engines of the gatekeeper, subject to anonymisation for the query, click and view data that constitutes personal data” (Art. 6(1)(j) Digital Markets Act—DMA).

But being a highly intrusive obligation, as well as being risky in terms of personal data protection, the Furman Report advised for certain checks and balances: data exchanges could only be made in a controlled environment with pre-approved entities, as well as under the supervision of the data protection authorities. They would furthermore be limited to opening up raw underlying data that is an input to the service, rather than processed information, where companies have invested in order to derive insights and inferences from the original data.

Common standards

Common standards are necessary for different networks to interoperate, be they digital platforms or traditional network industries. In the traditional network industries, standardization directly affected infrastructure. An early example can be traced back in 1846, when the UK Parliament adopted the “Act for regulating the gauge of railways.” Private investors developed individual lines, each one with the gauge they preferred. When two lines with a different gauge reached a city, a train would not be able to circulate from one to the other. Regulatory intervention subsequently became necessary to determine a common gauge. As a result, such standardization enabled railways to interconnect and interoperate, which was a precondition for the creation of a national rail network, created on top of the fragmented infrastructure (Montero & Finger, 2020).

In the case of the digital platforms what is needed in order to interoperate is the development of application programming interfaces (APIs), a set of programming code that enables data transmission between different software applications according to technical specifications describing the data exchange. Digital platforms often develop APIs in order to interface with third parties, but such APIs are usually defined on the platform’s terms, so as to reinforce its central position in the ecosystem that it aims building around it. For instance, mobile operating systems allow third parties to develop apps for their app stores, Facebook allows third parties to develop games and other applications to be integrated into its platform and so on. These API are the technical tool for platforms to create network effects on top of third parties’ services and assets.

However, incumbent platforms are reluctant to develop APIs to interoperate with competitors in such a way that the different network effects can be shared with competitors. The most obvious examples of platform interoperability can probably be found in the communications platforms. Just as emails can be exchanged across platforms, instant messages could be exchanged between communications platforms, such as WhatsApp, Google Hangouts, Apple’s Messages, and WeChat. This is not the case now; Facebook even used Federal Criminal Law in the US against a small startup called Power Ventures, which was trying to make different platforms interoperable (Stigler Committee, 2019, p. 16). The leading communications platforms are closed systems, competing to grow the largest network effects for themselves.

Developing APIs becomes more challenging when they require the agreement of a large number of actors, than when they can be unilaterally defined and imposed by a platform. This is the case of the definition of the APIs for the exchange of data by transport providers in a city so as to develop a Mobility-as-a-Service solution which coordinates all transport modes. Therefore public authorities have to play a more active role, for example by promoting the definition of standards for platforms to interoperate, and perhaps one day to impose them. Furthermore, public authorities are becoming increasingly active when it comes to defining standards of interoperability for the different actors in multi-sided markets, including standards for data sharing. This is for example, the case of the “data spaces” that are being proposed by the EU in the finance, transport, energy, health and seven other markets in its Data Governance Act. But the definition of standards is only a first step. Forcing market actors to adopt these standards will be another step. Indeed, the availability of standards does not automatically mean that actors are forced into interoperability.

Interoperability obligations

Different market players have different incentives to make their networks interoperable. The largest players often consider their largest network effects as their main competitive advantage. Smaller players, on the contrary, want to interoperate and neutralize the network effects enjoyed by the dominant players as a competitive advantage. Yet, at the same time, imposing an interoperability obligation disincentives investments for the creation of network effects. For these reasons, the obligation to interoperate is used a regulatory tool when market power becomes entrenched and the market shows no contestability. Such conclusions were reached so as to advance the liberalization of when some traditional network industries We may have reached the same point, at least in some digital markets.

In the early days of the telephony industry, the largest provider in the US, the Bell System, refused to interconnect with smaller carriers, so that the different telephone networks were not interoperable. It was necessary to contract with all the local carriers (usually two, sometimes three, and even four) to reach all subscribers in a town. Overtime, subscribers tended to contract with the largest carrier, enabling them to connect to the largest number of subscribers. And this was one of the main reasons behind the monopolization of the telephone market in the US (Mueller, 1997). It was argued that such a monopoly ensured what was called “Universal Service,” that is, the possibility to connect to all subscribers in town and beyond. In Europe and in most of the world, state owned monopolies became the preferred option in order to ensure the universality of the service.

The reintroduction of competition in telecommunications during the last decades of the 20th century relied on the obligation imposed on all carriers to ensure universal interoperability. Any telephony subscriber would be able to reach any other subscriber, as carriers would be obliged to interconnect their networks to ensure interoperability. Carriers terminating calls on behalf of other carriers would be remunerated based on the cost of the service, not on the value derived from the different network effects.

But interoperability led to a dual effect in the liberalized telecommunications industry. On one hand, it eroded the competitive advantage of the incumbent with the largest customer base. All newcomers would ensure universal access just as the incumbent did. In this way, regulation tamed the winner-take-all tendency created by direct network effects in communications networks. On the other hand, interoperability ensured the full exploitation of network effects, and the societal benefits derived from them. Compulsory interoperability enabled the coexistence of competition with the full exploitation of network effects. Both in the US and in Europe, regulated competition in the traditional network industries is focused on the same objective. Market power has to be disciplined by fostering competition from other services providers, whereby network effects are recognized as a leading source of such market power. But the objective is not to destroy the network effects, but to impede them from becoming a decisive competitive advantage. As a matter of fact, the value of such effects is recognized, honored, and protected.

This approach can, in our view, also be extended to the regulation of digital platforms. As a matter of fact, a first illustration of a regulation imposing platform interoperability can again be identified in the European Union. The new European Electronic Communications Code foresees the possibility of imposing interoperability on “number-independent interpersonal communications services”; that is platforms such as WhatsApp and Skype “which reach a significant level of coverage and user uptake” (Art. 61(2) Directive 2018/1972). Interoperability can be expanded to more complex interactions intermediated by platforms. In social networks, content uploaded onto a platform would become accessible from another social network, as proposed by the Stigler Committee (Stigler Committee, 2019, p. 118). These are examples of interoperability obligations that would share direct network effects among competitors for the provision of the same services (communications and social networks).

Interoperability has yet another dimension when it comes to the digital economy, as platforms create not only direct but also indirect network effects. Thus interoperability could not only be imposed so as to enable access to same-side users in otherwise competing platforms. Furthermore, fostering interoperability could also be envisaged so as to facilitate the access to different groups across competing platforms. For instance, a potential guest using a small platform might interact with the host using a different accommodation platform, but of course only if both platforms were connected and interoperable. Such mandated interoperability could be explored in markets with a winner-take-all dynamic, where measures to increase contestability are not enough and where multi-homing is not feasible.

Finally, interoperability is important as platforms aim to create walled gardens encircling not just one but a series of connected multi-sided markets and ancillary services. For instance, the large Silicon Valley behemoths have their own email platforms, instant communication platforms, identity recognition systems, payment systems, and so on. The European Commission, for instance, has proposed legislation to prohibit gatekeepers from forcing the use of their own ancillary services (for instance identification services) and has imposed interoperability obligations to make competing ancillary services available for platform users (DMA Arts. 5 and 6). In this way, network effects are not diminished by mandating interoperability to platforms, but can actually be reinforced, as more users are included in the multi-sided market. The experience of traditional network industries suggests that a comprehensive regulatory framework, including institution building, is necessary in order to make such interoperability regulation effective.

Horizontal divestiture

Horizontal divestiture—that is, the break-up of a company to diminish its size—challenges the very source that makes platforms efficient. Divestiture is certainly an effective measure to reduce market power also in the network industries, traditional and new. Horizontal divestiture affects the very source of market power, namely the level at which the network effects that provide the ultimate competitive advantage ignite. However, by reducing the network effects, value is also being destroyed, not only for the divested company, but also for users and for society as a whole.

Also, horizontal divestiture might be ineffective if a market naturally concentrates when the market leader with the largest network effects has lower costs or another competitive advantage. The experience of the break-up of the Bell System shows that even though the monopoly was separated into to seven regional monopolies (the so-called Baby Bells), these monopolies ended up being concentrated again a couple of decades later into two carriers.

However, horizontal divestiture might also create more benefits than damage. This might be the case of markets monopolized by mergers or through anticompetitive practices that went unnoticed at the right moment. This may have been the case of Google and Facebook. There have indeed been growing criticisms of the weak antitrust enforcement in the digital world. For example, the antitrust suit filed by the Federal Trade Commission (FTC) against Facebook in December 2020 has formally requested the divestiture of Facebook’s Instagram and WhatsApp. The FTC identified that the scope of both acquisitions was an obstacle to the growth of potential competitors. Similar calls have been made for the divestiture of DoubleClick from Google (Warren, 2019).

Vertical unbundling

Vertical unbundling is a more nuanced approach to fighting some of the negative effects of market monopolization because of network effects. Digital platforms would be allowed to intermediate in the provision of good and services, but they would not be allowed to provide the underlying service in competition with third parties. Examples would be Amazon intermediating at times between providers, but sometimes also producing their own goods or a railway company operating a mobility-as-a-service app while at the same time operating trains. Separation can be total, when activities would be undertaken by independent companies, or only functional, in the sense that different corporations are required to provide the intermediation and the underlying service, even though they belong to the same business group.

Vertical separation is a regulatory measure that has been adopted in Europe in different network industries (OECD, 2016), particularly electricity (separation of electricity production from transmission and distribution) and railways (separation of infrastructure management and transport service provision). It is a structural obligation to be adopted so as to exclude self-preferencing. Discrimination in favor of the downstream arm is excluded (ownership separation) or at least made more transparent (functional separation).

Vertical unbundling does not have the same negative economic effects as does horizontal divestiture, as it does not reduce the value of the different network effects in a specific market. On the contrary, the scope of vertical unbundling is precisely to structurally impede the extension of market power generated by network effects downstream. In this case, a marketplace would not be allowed to sell its own goods, and so on. As a matter of fact, vertical unbundling has been called for in the case of the Amazon marketplace (Kahn, 2019). For example, authorities in India enacted a regulation, effective in February 2019, prohibiting e-commerce marketplaces such as Amazon from selling their own inventory. Amazon is only allowed to operate marketplace platforms where other parties sell goods to retail consumers. In this way, the authorities expect to put an end to Amazon’s supposedly predatory practices of duplicating successful products that were commercialized by third parties and then discriminating in favor of its own products.

But vertical separation is not limited to matchmakers. A company managing an app store would not be allowed to exploit its own apps. It has also been proposed for non-transaction platforms: Google, for example, would have to separate the ad exchange platform it manages from the underlying ad display ventures, such as the search engine and YouTube.

Data separation

As data is key for digital platforms, a new form of separation can be proposed as a remedy, namely in the form of data separation. Specific units in a platform would not be allowed to share data under certain circumstances. In this way, algorithmic network effects are somewhat reduced. For example, this was the remedy imposed by the German antitrust authority upon Facebook, forbidding it from pooling together data from Facebook and data from WhatsApp (Bundeskartellamt Decision February 6, 2019 in case B6-22/16).

The European Commission has proposed a data separation obligation to be imposed upon the largest digital platforms in its Digital Markets Act when imposing the ex ante obligation to: “refrain from combining personal data sourced from these core platform services with personal data from any other services offered by the gatekeeper or with personal data from third-party services, and from signing in end users to other services of the gatekeeper in order to combine personal data, unless the end user has been presented with the specific choice and provided consent in the sense of Regulation (EU) 2016/679” (Art. 5(a) DMA).