The most expensive way to test an idea is to build it. Yet that's the default for most teams: months of development, a launch, and then, finally, the market's verdict. By the time the answer arrives, the budget is spent.
Validation flips the order. You buy the market's verdict first, cheaply, and only then commit to building. It isn't about killing ideas, it's about finding the version of the idea that deserves the investment.
Start with the problem, not the solution
Every validation effort we run starts by separating the problem hypothesis from the solution hypothesis. 'Restaurant managers waste hours reconciling rosters' is a problem you can test with ten conversations. 'An AI rostering app' is a solution you can't test until you know the problem is real, painful and frequent.
If the problem isn't worth solving, no execution can save the solution.
Evidence beats opinions, including yours
Founders are the worst-placed people to judge their own idea, and that's fine: conviction is fuel. The discipline is in pairing conviction with evidence: interviews that probe behaviour rather than intent, landing pages that measure real demand, and concierge tests that deliver the value manually before any software exists.
- Ten problem interviews before any design work
- A demand test that asks for something real, an email is weak, a deposit is strong
- A concierge or manual version of the service to test willingness to pay
- A clickable prototype tested with people who match your actual customer
The prototype is the pivot point
A high-fidelity prototype is where validation gets serious. It's real enough that users react honestly, and cheap enough that changing course costs days, not quarters. This is the core of our Prototype to Proof engagement: two to three weeks from concept to something testable, with evidence attached.
The output of good validation isn't a yes or a no. It's a sharper product: a clearer customer, a tighter scope, and a roadmap where the risky assumptions are at the front, not buried in month four of the build.
