Interoperability Doesn't Set Users Free. It Redraws Who Holds The Network.

Based on the research ofGuofu Tan and Danhou Li, "A Network Approach to Interoperability," Platform Strategy Symposium, 2026

A platform's network effect is not a fixed asset it owns. It is a byproduct of who it is allowed to talk to, and changing the wiring changes who benefits.

A new theoretical paper presented at the 2026 Platform Strategy Research Symposium, "A Network Approach to Interoperability" by Guofu Tan and Danhou Li, builds a formal model of exactly this wiring problem. It treats interoperability among competing platforms as a weighted network rather than a yes-or-no switch, and it asks what happens to prices, competitive position, and user welfare as connections are added, mandated, or withheld. The headline result cuts against the reflexive policy instinct that openness is simply good: whether interoperability helps or hurts depends on where a platform sits in the network and on the specific shape of the interconnection, not on some general property of "more openness." For a dominant incumbent, mandated interoperability can drain the very scale advantage that made it dominant. For a smaller platform, voluntary interconnection can import a rival's network effects and lift its users' welfare. Both things are interoperability. They are not the same policy.

That distinction sounds abstract until you watch regulators and companies live through it in real time. The past three years have handed us a working laboratory: the European Union forcing messaging giants open, Apple choosing exactly how much of its own wall to remove, two rival protocols racing to build the next open social web, and a two-decade-old antitrust remedy that already tested what mandated interoperability does to a dominant incumbent's server business. Read together, they are the paper's mechanism playing out on live platforms.

The EU's Live Experiment on a Real Incumbent

Article 7 of the EU's Digital Markets Act does precisely what the Tan and Li model studies: it forces a designated gatekeeper's network to open to outside connections it did not choose. Meta was designated a gatekeeper for WhatsApp and Messenger on September 6, 2023, and under Article 7 it must let rival messaging providers request interoperability with its basic functionality, free of charge, through a published reference offer. The obligation is phased. One-to-one text and file sharing had to work by March 7, 2024 for WhatsApp and September 7, 2024 for Messenger. Group chats follow by September 7, 2025 for both services. Full voice and video calling is not required until September 7, 2027. According to the European Commission's own developer portal, the first two third-party messaging services actually interoperable with WhatsApp launched in November 2025, more than a year and a half after the first compliance deadline.

That gap between the legal deadline and a working outside connection is itself informative. A mandate does not instantly rewire a network; someone has to build the bridge, and an incumbent has limited incentive to make that bridge fast or elegant. This is the paper's core tension in miniature: WhatsApp's scale was built inside a wall, and Article 7 is a regulator deciding, functionality by functionality, how much of that wall to remove and on what timetable. The phasing itself is a network-structure choice: text now, calls not until 2027, because the Commission understands that different functionalities carry different amounts of the network effect across the wire.

Apple Chooses Its Own Topology

If Brussels illustrates a mandate imposed on a network, Apple's RCS rollout illustrates a dominant platform choosing its own interoperability topology under pressure, which is exactly the "coalitional" case the paper's model treats as distinct from full interoperability. After years of resisting Google's push to adopt Rich Communication Services, Apple announced in November 2023, and shipped with iOS 18 in 2024, support for the RCS Universal Profile for messages between iPhones and Android devices. Industry coverage at the time, including MacRumors's own accounting of the decision, pointed to regulatory pressure, including the EU's push to make iMessage more interoperable, as a factor Apple could not ignore.

Look at what Apple actually opened, though. RCS improves read receipts, typing indicators, and media quality for iPhone-to-Android texts, the green-bubble conversations that were never where Apple's core network effect lived. iMessage-to-iMessage conversations, the blue bubbles, kept their existing features, their encryption, and crucially their visual distinction from everyone outside the wall. Apple interconnected the commodity layer of messaging while leaving its highest-network-effect layer, the thing that makes an iPhone feel like it belongs to a closed club of other iPhones, untouched. That is not incoherent. It is a strong incumbent selecting exactly the slice of the network to open, minimizing how much of its own advantage crosses the new connection while still answering regulatory and competitive pressure. The model's finding that a dominant platform's exposure depends on which connections form, not whether any connection forms at all, describes this move almost exactly.

Two Protocols Betting on Opposite Topologies

Not every interoperability decision is regulator-driven. Two social networks built after Twitter's decline made opposite bets about how open to be from day one, and the difference is a live test of the paper's claim that the specific structure of interconnection, not just its presence, decides the outcome. Mastodon runs on ActivityPub, a W3C standard where each server, or instance, is independently operated and your identity is tied to that instance's address. Bluesky runs on its own AT Protocol and, according to Bluesky's own February 2024 announcement of its federation rollout, deliberately separated identity from hosting: a user's handle, followers, and posts can move to a different server without loss, the way a phone number survives a change of carrier. Bluesky's own comparison of the two systems is explicit that the two protocols are not interoperable with each other by default, despite both calling themselves federated and open.

That non-interoperability between two ostensibly open networks is the paper's point stated the other way around. Both Mastodon and Bluesky voluntarily interconnect independent servers within their own protocol, importing a kind of shared network effect for anyone who joins any node. Neither protocol has built full interoperability with the other, because doing so would mean redesigning identity portability, moderation, and discovery, exactly the network-topology variables the model shows drive the welfare and competitive result. Two protocols can each be internally open and structurally incompatible with each other, which is why bridging tools built by outside developers, not the platforms themselves, have had to stitch the two networks together piecemeal rather than the platforms simply flipping a switch.

Interoperability is not a light switch. It is a wiring diagram, and where you sit in it decides whether opening the wall lifts you or drains you.

An Older Mandate Already Tested the Incumbent Case

The paper's claim about mandated interoperability eroding a strong incumbent's advantage is not hypothetical. It already happened once. In its March 2004 decision, the European Commission found Microsoft had abused a dominant position by refusing to supply competitors with the interoperability information needed to build work group server products that could talk to Windows PCs and Windows servers on equal terms. The remedy, upheld by the EU's General Court in a 2012 judgment, forced Microsoft to disclose complete and accurate protocol specifications so that non-Microsoft servers could achieve full interoperability with the Windows environment. Microsoft's dominance in that market rested substantially on the fact that only Windows servers could fully and reliably talk to Windows desktops. Mandated interoperability did not add a nice-to-have feature; it removed the specific structural asymmetry that had been generating Microsoft's advantage in that adjacent market. This is the incumbent side of the Tan and Li result, decided by a court two decades before the paper formalized why it happens.

Telecom number portability offers the calmer, symmetric counterpart. When U.S. regulators required carriers to let customers keep their phone number after switching providers, no single carrier held a structural network advantage the way WhatsApp or Windows servers did; every carrier interconnects with every other network by design, and identity, the phone number, was the only thing still locking a customer in. The FCC's current consumer rules require that a simple port be completed within one business day and forbid a losing carrier from blocking a port over an unpaid balance. Because the underlying networks were already interoperable and only account switching was walled, removing that one wall mostly lowered switching costs symmetrically across roughly equal competitors, closer to the paper's full-interoperability, welfare-maximizing case than to the WhatsApp or Windows scenario.

The throughline across all four cases is the same one the model formalizes: interoperability is not a single lever with one setting. It is a set of choices, about which functionality connects, which participants are included, and how much of the incumbent's structural advantage crosses the new wire, and each of those choices redraws the map of who wins. A regulator or an operator who treats "more interoperability" as an unqualified good, or an unqualified threat, is answering a network-structure question with a slogan.