Create five hypotheses and five ways each could be wrong; conduct suggested 5–10 focused interviews or product observations, with the size adapted to the decision rather than sold as statistical proof. Prioritise recent buyers, non-buyers and churned users; ask about real behaviour and alternatives, not whether a speculative feature “sounds useful”. YC's interview guidance is a practical reference. How To Talk To Users — Y Combinator
Deliverables: interview guide, redacted notes, demand-versus-politeness map, observation log. Pitfall: selecting only fans or treating interview counts as representative market measurement.