The Platform Bet

What this pack runs.

A platform, as this pack uses the term, is a product other people build on and sell through. A product becomes one at the moment a third party can earn more by building on it than by building alone, and it stays one only while that holds. So the decision is arithmetic on two sides, and most rooms price one: they choose a rate from a phone platform and argue about whether developers will accept it, without the two numbers that would settle it. The rate the layer needs comes from the owner's own costs, and the layer sits inside an engineering budget that exists for the owner's own product, so the figure in the room is an allocation that makes the platform look impossible. The most a builder will pay is a fact about somebody else's business, and nobody has asked it. This pack turns it into two rates and the distance between them. The rate the layer needs is the running cost, the build spread over the years, and the contribution on the sales opening displaces, over what third parties would sell through you. The break-even take rate is the builder's margin, less what it earns alone plus what building on you costs, over what it would sell through you. Between them is the corridor, and the two distances from the intended rate to its edges are what a board watches: how far the rate can rise before builders stop, and how far it can fall before the layer stops paying. Beside the arithmetic it sets out what a third party needs before it builds, derived from nine sourced platform bets and labeled as analysis: a waiting market it cannot reach alone, a layer that works without you, a rate that leaves it better off, and a promise you will not compete with it or close the door. It stops on two conditions almost no company has met: the platform layer costed on its own, and a builder asked what it earns without you.

The casebooks these cases come from.

The pack reads nine sourced decisions through the framework. They are drawn from these casebooks, each one a set of verified records on a single decision type.

A verified record has passed a mechanical claim-versus-source check, a second model's read, and human approval. The count is what exists today, not a target.