"MVP" has been used to mean everything from a Figma file to a full product with one feature missing. The useful definition is narrower and it is about risk.
The definition worth using
The smallest build that tests your riskiest assumption with real users. Not the cheapest build, and not a subset of the roadmap — the one that answers the question you are most likely to be wrong about.
MVP, prototype and demo are different things
- Prototype — clickable, not real, answers "does this flow make sense?"
- Demo — real enough to show, not real enough to use
- MVP — a real app, in a real store, with real accounts and real data, just far narrower than the eventual product
Confusing these is how teams end up shipping a demo to paying users.
What gets cut
One platform. One core loop. Settings, admin surfaces, onboarding polish and the second persona all wait. Two stores doubles QA, review and support before you know whether anyone wants it.
What must not get cut
Proper auth, CI, crash reporting and a sane data model. These are invisible to users and they are exactly what makes version two a rewrite when they are missing. Cutting them is not scope reduction, it is debt.
How you know it worked
Instrumentation you decided on before launch, answering a question you wrote down before launch. An MVP that ships without that is just a small app.
In practice
Our MVP sprints run $15k–$40k over 6–10 weeks on one platform. The cost page breaks down what moves that number.