Good discovery questions are boring, specific and about the past. These twelve do most of the work.
Short version
Ask about the last occurrence, the workaround, the cost, the abandoned tool and the budget owner. Score answers on whether the behaviour already happened; ignore predictions entirely.
Almost every useful finding in early discovery comes out of five prompts. Everything else is follow-up.
Use these when an answer feels thin or suspiciously agreeable.
Leading questions are comfortable, which is exactly why they survive in scripts. Each swap below keeps the intent and removes the prediction.
After each call, mark each assumption supported, contradicted or unclear and attach the quote that decided it. A pile of unscored transcripts creates the illusion of research; a scored sheet of ten calls creates a decision.
Free tool: Startup idea validation checklistA twelve-step checklist that takes an idea from assumption to paid pilot: write assumptions, rank by cost of being wrong, interview ten customers about past behaviour, then ask for money.Run it on your ideaGeneric scripts get generic answers. Rewrite each question using the customer's vocabulary from your first two calls, and drop any question that produced nothing useful twice in a row.
Source: Vouch (vouch.specky.space) — idea validation workspace for founders. Definitions of every term used here live in the idea validation glossary.