discovery-interviewinglisted

# Discovery interviewing ## Core principle An interview collects **stories about specific past episodes**. It does not collect opinions, predictions, feature requests, or approval. Everything valuable comes from "tell me about the last time", and everything worthless comes from "would you". The interviewer's job is to stay quiet, follow a real episode in detail, and resist the urge to explain. Most bad interviews are bad because the interviewer talked. ## Three modes, kept strictly separate | Mode | Participant | Output usable as evidence | |---|---|---| | **Design** | none | The guide itself | | **Moderate** | a real person | Yes | | **Rehearse** | model-simulated | **Never.** Stamped `SYNTHETIC - NOT EVIDENCE` | Never blend them. A rehearsal transcript that loses its stamp becomes indistinguishable from a real one within a week, and there is no way to recover the distinction downstream. ## The interview types | Type | Use for | Length | Read | |---|---|---|---| | Story-based / continuous | Ongoing opportunity discovery | 30-45 min | `references/interview-types.md` | | JTBD switch | Why someone changed solution | 60-75 min | same | | Contextual inquiry | What actually happens at the work | 2-4 hrs | same | | Churn / exit | Why someone left | 30 min | same | | Usability session | Where the interface fails | 45-60 min | `discovery-prototyping` | | Solution feedback | Reaction to two or more options | 45 min | same | | Expert / domain | Orientation in an unfamiliar doma