MVP: Not the Smallest Product — the Smallest Reliable Learning
"Minimum Viable Product" gets misread constantly — as a low-quality first draft, or as an excuse to randomly shrink a wishlist. It's neither. An MVP is the smallest thing you can build that reliably tests your riskiest assumption.
A good first slice
Targets a specific user problem, produces genuinely usable value end-to-end, tests the riskiest assumption, and generates a measurable feedback signal that informs the next decision. What it is not: a low-quality product, a randomly shrunk scope, a purely technical layer, a first version with no learning question attached, or the small-looking first phase of a big plan you've already fully decided on.
Vertical, not horizontal
Horizontal slice
First all the analysis, then all the interface, then all the backend, then integration last. At no point can a real user actually complete anything — there's no usable value until everything is finished.
Vertical slice
One narrow path cut all the way through every layer — interface, logic, data — so it's genuinely usable end-to-end, even if it only covers one user type and one scenario.
Pick the assumption before the scope
The order matters. Teams that start with 'what's the smallest scope we can build' end up cutting the easiest features, not the riskiest ones. Start instead with: which untested assumption would cause the most damage if it turned out wrong? Then design the narrowest slice that puts that specific assumption in front of reality.
Close the loop with a decision rule
A slice without a pre-agreed decision rule just produces data nobody acts on. Before you ship it, agree on what result means 'continue,' what means 'change the approach,' and what means 'stop and revisit the assumption.' Deciding this after you see the data is exactly when it gets rationalized away.
Frequently asked questions
Is an MVP just a low-quality first version?
No. Scope can be minimal, but the quality and safety bar for what's included should not be lowered.
How do I pick the first slice?
Pick the riskiest, least-tested assumption first, then design the narrowest end-to-end scope that would reliably test it.
What if the first slice fails?
That's the point of testing it small and early — far cheaper than learning it after building the whole wishlist.
Now practice it
Design a first value slice from a real wishlist scenario and get AI feedback.
Open the PracticeLab →