Should we open our platform to outsiders?

The question behind the question

“Should we open up?” is usually asked as if openness were a single switch. In practice there are several decisions hidden inside it. Who may build on the product, and on what terms? Who may see the code, the data or the interface? Who sets and changes the rules? And who may compete with you on top of what you opened? Every successful case in the record answers these questions differently, and most of the failures answered one of them carelessly.

The underlying trade is between the size of the pie and the share of it you keep. Opening invites others to add value you could never add yourself, which makes the whole system more attractive. It also lets them capture some of that value, and sometimes lets them replace you. Carl Shapiro and Hal Varian put it plainly in Information Rules (1998): a firm’s reward is its share of the industry’s value multiplied by the total value of the industry, and openness raises the second number while lowering the first.

When opening worked

The clearest historical lesson comes from video recorders. In 1975 Sony launched Betamax on its own design and looked for partners afterwards. JVC launched VHS in 1976 with longer recording time and licensed it widely to other manufacturers. Many brands of recorder created a large installed base, which persuaded video shops to stock VHS tapes, which made VHS recorders more attractive still. Sony’s picture quality did not matter once the shops had chosen. JVC gave up exclusive control and won the market.

The same logic ran through the most successful platform decisions of the smartphone era, but with an important refinement: none of them opened everything. Google released Android as open source and let any phone maker use it, while keeping its app store, maps and other services proprietary and licensed under contract. Apple opened the iPhone to outside developers in 2008, but only through a store in which it reviewed every app and took a commission. Amazon opened its product pages to independent sellers in 2000, letting them compete with its own retail offers, while keeping the customer relationship, the payments and the data. In each case the company opened the layer that others needed and kept the layer from which it earned its money.

A third kind of success involved giving up control of the rules altogether. Dee Hock persuaded the banks issuing BankAmericard to turn it into an organisation they owned jointly, which became Visa; competitors would cooperate on shared rails because no single competitor controlled them. India’s UPI applied the same principle as public policy, building an instant payment system owned by a not-for-profit company of banks and open to any bank and to third-party apps. In 2011 Google, Microsoft and Yahoo created schema.org, a shared vocabulary for structured data, which succeeded where the more ambitious Semantic Web had struggled, partly because the companies that would reward its use agreed on it together.

When it did not

The failures in the record fall into two groups. The first kept too much closed when a rival offered an open alternative: Sony with Betamax, and in a different way Lotus, which sued makers of compatible spreadsheets for copying the look and feel of 1-2-3, spending years in court while Microsoft Excel took the market on a different platform.

The second group opened without rules and had to close later. Facebook opened its platform in 2007 with broad access to users’ data and their friends’ data, which attracted developers quickly and produced the access that later enabled the Cambridge Analytica affair; it restricted access progressively from 2014. Open-source infrastructure companies such as Elastic and MongoDB built businesses on permissive licences, then between 2018 and 2024 changed to source-available terms to stop cloud providers offering their software as a service. The record judges those relicensings to have backfired: communities forked the projects, and the cloud providers were not stopped. Closing what was open costs far more trust than never opening it.

A diagnosis to start from

Richard Rumelt’s Good Strategy/Bad Strategy (2011) argues that strategy begins with a diagnosis of the critical challenge. For the question of openness, the record suggests a diagnosis built on three facts. First, whether complements, the apps, accessories, content or services others build, make your product more valuable to its users; if they do not, openness brings little. Second, whether a credible open alternative exists or could be created quickly; if it does, a closed system has a limited time to stay ahead on integration alone. Third, what outsiders must trust you not to do once they depend on you.

From that diagnosis follows a guiding policy that most of the successes share: open the layer where outside contributions create value, keep a layer that is hard to replicate and that customers pay for, and make the rules credible, either by binding yourself, as TSMC did by promising never to make its own chips, or by putting governance in shared or public hands. The coherent actions are then specific: what to license, under which terms, what to publish, who sits on which committee, and what the company promises in writing about how it will and will not change the rules.

The strongest counter-case

The strongest objection is Apple. For most of its history Apple has kept tight control of hardware, software and distribution, and it has been among the most profitable companies in the world. Integration lets it make products that work together, decide quality and capture most of the value its ecosystem creates. Closed systems often win early, when integration matters most and standards are immature, which is the pattern this site calls open standards out-reach closed gardens: open systems tend to reach further over time, while closed ones win first. And Apple’s App Store, counted here as an opening, is a closed garden in most respects.

The counter-case is real, and it narrows the advice rather than reversing it. Integration wins when it is what customers are paying for, and when the integrator is clearly ahead. It is vulnerable when the market’s centre shifts to complements the integrator does not control, and it attracts regulators once the gate becomes the only way in, as Apple’s later disputes over its commission show.

What the wider research says

Studies of platforms broadly support the record. Kevin Boudreau, studying handheld computer platforms, found in 2010 that granting outside developers access raised the rate of new devices substantially, while giving up control entirely had smaller and less consistent effects. Thomas Eisenmann, Geoffrey Parker and Marshall Van Alstyne argued in the Harvard Business Review in 2006 that platform owners must decide which side to subsidise and how much to open, and that the answers depend on network effects on each side. The literature on standards, beginning with Joseph Farrell and Garth Saloner in the 1980s, shows that once an installed base forms, users find it hard to move even to a better alternative, which is why decisions about openness made early are so hard to reverse later; the site’s idea Path Dependence and Lock-In explores why.