A platform strategy opens a product to people outside the organisation, so that they build on it, sell through it or trade across it. Marshall Van Alstyne, Geoffrey Parker and Sangeet Paul Choudary set out what that changes in Harvard Business Review in April 2016. The critical asset of a platform is external, they wrote, and by that they meant the community of members. The work of the organisation moves from controlling resources to orchestrating them. Buying, building and partnering all assume the organisation assembles the capability itself. A platform is the arrangement where somebody outside assembles it and the organisation never owns the result.
The sections below set out what a platform demands of the organisation that opens one. They then cover the chicken and egg problem every platform starts inside and the side on which most platform strategies fail. The page ends with what running a platform costs once it works.
What a platform demands of the organisation
A platform needs two things from the outside world, and organisations reliably supply only one of them. The first is a reason to build, which means access to customers, to data or to a route to market that the builder cannot reach alone. The second is a way for that builder to be paid, which means revenue that is still there after the platform's own roadmap has moved on.
An interface supplies neither of them. If an organisation publishes an API and announces a partner programme, it has given outside developers permission to work and no reason to do it. That is why partner directories fill with integrations written once by an agency and never touched again.
The chicken and egg problem
Neither the reason nor the payment exists on the first day, and that is the problem every platform starts inside. Builders go where customers already are. Customers go where the useful things have already been built. A platform launched to both sides at once has nothing to offer either of them.
Andrew Chen named the two parts of the answer in The Cold Start Problem in 2021. An atomic network is the smallest group that holds together without anybody pushing it, and it is usually far smaller than a launch plan assumes. The hard side names the participants a platform cannot do without. On a marketplace those are the sellers, and on a developer platform they are the people writing against the interface. Those participants create most of the value, they cost the most to recruit, and the easy side follows them.
Opening a platform to both sides at once leaves each side waiting for the other. Seeding one narrow network on its hard side gives the easy side something to arrive for, and the whole move is then repeated network by network.
Why most platform strategies fail on the hard side
Recruiting the hard side is where platform strategies most often stop. David Yoffie, Annabelle Gawer and Michael Cusumano studied more than 250 platforms and published the result in Harvard Business Review in May 2019. Of the 252 platforms they examined, 43 survived. The failures gather around the economics of the two sides, and around one side in particular.
An organisation opening a platform funds the side it already understands. A retailer opening a marketplace spends on shoppers, because reaching shoppers is what its marketing department knows how to do. Nobody spends on the sellers, and the sellers are the ones who would give those shoppers a reason to come at all. The money lands on the easy side, and the hard side stays empty.
The quieter failure is about payment. The platform grants access and never states what the builder earns, so the builder carries the cost of building while the platform keeps the customer relationship. If a developer reads those terms, they can see that the platform may enter their category next year. That developer prices the risk in and builds somewhere else.
What running a platform costs
A platform that works is a standing obligation. An interface other companies depend on is a promise, and withdrawing it costs more every year, because every builder who wrote against it has customers of their own by then. Review, documentation, version support and a deprecation policy nobody enjoys writing all arrive with the first successful integration, and none of them ever leaves.
The harder cost is restraint. Every popular thing built on a platform is a market the platform's own team can see. Entering that market buys one year of revenue, and it teaches every remaining builder what the platform is for. A platform is worth opening only if three conditions hold at once.
- The product already has customers a builder would want to reach.
- The organisation can state what a builder earns and how the money arrives.
- The team will leave a profitable category to its builders for longer than one planning cycle.
If any one of the three is missing, an agreement with a handful of named companies delivers much of the same capability for a fraction of the obligation.
The three conditions close the choices about how to win. A segment, an advantage, a promise, a model, a price, a route and a decision about who builds what now stand as one argument. Every line of that argument is a bet nobody has tested against a customer. Whether any of it is right is where the next module starts.