Stop Copying Apple's App Store. Your Ecosystem May Need the Opposite.

Based on the research ofDavid R. Clough, "Governance Structures and Coordination Trade-offs: A Discriminating Alignment Theory of Innovation Ecosystem Architectures," Strategy Science, 2026

David Clough's new theory of ecosystem governance, published in Strategy Science, extends Oliver Williamson's markets-and-hierarchies logic to innovation ecosystems and arrives at an unglamorous conclusion: there is no universally best governance architecture. Single-firm-controlled platforms, decentralized coalitions, and "platformless" ecosystems with no central owner at all each facilitate coordination differently, and each carries its own trade-off. The question that determines which one wins is not which owner holds the tightest grip. It is whether the architecture's coordination costs match the ecosystem's actual coordination needs and its uncertainty. Grab more control than the environment demands, and you pay for machinery nobody needed. Grab less than it demands, and incompatibility costs pile up until something forces convergence. Three real ecosystems, pulled from opposite ends of the governance spectrum, show both failure modes in progress.

Apple's App Store: the control that outran its own justification

Apple's App Store is the textbook case of tight single-firm control: one company owns the operating system, the hardware, the review process, and the payment rails, and charges a standard 30 percent commission on digital transactions (cut to 15 percent for developers earning under $1 million a year under the Small Business Program, per Apple's own published terms). That degree of control has a real coordination rationale, a single, consistent runtime and review bar genuinely reduces fragmentation risk for over a billion devices. But Clough's framework implies that the right amount of centralization tracks the environment's actual coordination needs, not the platform owner's appetite for control, and by that test Apple has been overshooting for years.

The clearest evidence is regulatory and legal rather than commercial. The European Commission fined Apple €500 million in April 2025, its first non-compliance decision under the Digital Markets Act, specifically because Apple's restrictions stopped developers from telling customers about cheaper offers outside the App Store, control exercised over a transaction (informing a customer of a price) where no plausible technical coordination need existed at all. Apple is appealing, but the fine is itself a data point: regulators are, in effect, pricing the gap between the coordination Apple actually needs to run a secure, consistent app ecosystem and the control it has exercised beyond that. When tight single-firm governance starts extracting rules that have no coordination rationale left to defend, the "control" stops being alignment and starts being rent, and rent draws antitrust scrutiny, developer flight to arbitrate elsewhere, and the EU's growing list of alternative-distribution carve-outs that Apple has had to concede, piecemeal, since 2024.

The tell that an ecosystem is over-centralized is not complementor complaints. It is a regulator pricing the gap between the coordination you need and the control you took.

Kubernetes: the coalition built for an environment nobody could predict

Cloud infrastructure in 2014 was exactly the kind of environment Clough's theory says a decentralized coalition should fit: high uncertainty (nobody knew which container orchestration approach would win, or how fast the category would move), many interdependent vendors who each needed the standard to succeed without any one of them trusting a rival to own it outright, and a coordination need too large for bilateral deals to solve but too contested for one firm's control to be accepted by competitors. Google's actual move matched the diagnosis: rather than keep Kubernetes as a proprietary advantage, Google donated it in 2015 to the newly formed Cloud Native Computing Foundation, run by the Linux Foundation, with founding members including AT&T, Cisco, IBM, Red Hat, Docker, and more than a dozen other companies who would otherwise have been racing Google rather than building on its project.

The governance structure that emerged is genuinely decentralized rather than decentralized in name only. The Kubernetes Steering Committee is a seven-member body elected by the contributor community, and its own members describe its job less as directing the project than as deliberately not directing it: nearly all technical authority is delegated to Special Interest Groups and Working Groups, and the Steering Committee's actual day-to-day work, by its members' own account, is handling "people problems", visa letters, governance disputes, liaison with the CNCF, while staying out of technical decisions entirely. That is a coordination architecture sized to the uncertainty of the environment: no single vendor's roadmap risk gets imposed on the whole ecosystem, decisions move at the (slower, more contested) pace a multi-vendor standard requires, and legitimacy comes from the absence of a single controlling firm rather than the presence of one. A decade on, Kubernetes remains the shared substrate across AWS, Google Cloud, Microsoft Azure, and most of the independent cloud-native vendor landscape, a result that tight single-firm control over the same technology would have made far harder to sustain, precisely because no single vendor's ownership would have been acceptable to its rivals.

When platformless isn't enough: the EV charging standards war

The opposite failure mode shows up where Clough's theory would predict it: in a physically interdependent system that tried to run with too little coordination for too long. Through the early 2020s, electric vehicle charging in North America was close to "platformless" in the sense Clough's paper uses the term, a loose network of automakers and charging-network operators, each running its own connector standard (Tesla's proprietary design, the CCS connector pushed by most other automakers, and the older CHAdeMO standard), with no central owner coordinating which plug would fit which car at which charger. That looseness is fine for an environment with low interdependence. It is a serious liability for one where the entire value of the product, being able to charge your car at any station, depends on physical compatibility that a loose network, by construction, cannot guarantee.

The correction came fast once the mismatch became undeniable. Tesla opened its connector design in November 2022, renaming it the North American Charging System and noting, in its own announcement, that NACS vehicles already outnumbered CCS vehicles two-to-one and that its Supercharger network had 60 percent more charging posts than all CCS-equipped networks combined, a scale advantage that made continued fragmentation costly for everyone else. Ford was first to move, announcing in May 2023 that its EV customers would get access to more than 12,000 Tesla Superchargers, with native NACS ports replacing CCS starting with the 2025 model year; General Motors followed within weeks. By December 2023, SAE International had completed the technical report standardizing the connector as SAE J3400, developed, SAE said, "through broad-based industry consensus." A platformless ecosystem had run for a decade on an interdependence it could not actually support without a referee, and once the mismatch became obvious to everyone simultaneously, convergence happened in about thirteen months rather than through years of negotiated standard-setting.

If your own ecosystem looks platformless today, the test is not whether anyone likes the looseness. It is whether what participants are coordinating on is genuinely low-interdependence, where loose and uncoordinated is fine and even good for innovation speed, or only quietly high-interdependence, where the absence of a referee is a standing liability waiting for a trigger to force one into existence on someone else's timeline rather than yours.

Diagnose before you copy

None of these three ecosystems chose its architecture arbitrarily, and none of them is wrong in the abstract. Apple's control is justified exactly where it still solves a real coordination problem and overextended everywhere it does not. Kubernetes' coalition governance is justified because the uncertainty and multi-vendor distrust it was built for were real and durable. The pre-2022 EV charging network's looseness was justified only as long as nobody noticed that "which plug fits where" is not actually a low-interdependence problem. Clough's discriminating-alignment framework gives operators a diagnostic rather than a playbook: map your own coordination needs, your own uncertainty, and where control genuinely needs to sit, before reaching for whichever governance model is currently fashionable. Copying Apple's grip when your ecosystem does not need that much coordination is not caution, it is an expensive, avoidable misalignment. Staying platformless when your participants secretly depend on each other's compatibility is not flexibility, it is deferred coordination cost with interest. The paper's most useful instruction may be its last one: environmental parameters shift, so the governance choice that was aligned five years ago is a hypothesis to retest, not a settled answer.