Most MVPs fail before design starts — they try to ship the vision instead of a slice of it. After 140 launches, our rule is simple: the first release must let one user complete one valuable task end to end.
Start from the riskiest assumption
List everything the product must prove: will users pay, will the data flow, will the integration hold? Rank by risk, not by excitement. Your MVP tests the top two and fakes or defers the rest.
Slice vertically, not horizontally
A login page plus a settings page plus half a dashboard is not a product. One complete workflow — signup to value to notification — teaches more than five disconnected screens.
If the scope does not fit in eight weeks with one squad, it is not an MVP — it is phase one wearing a costume.
What we defer on purpose
Admin panels become database queries. Notifications become manual emails. Multi-role permissions become one trusted role. Each deferral is logged with a trigger — the metric that earns it real engineering later.
Run this exercise in discovery and your roadmap stops being a wishlist. It becomes a sequence of funded bets, each one shippable in weeks.