customer-discovery

# Customer discovery ## Usage ```text /customer-discovery plan <participant-or-segment> [area] [feature] /customer-discovery capture <participant-or-segment> [area] [feature] /customer-discovery debrief <participant-or-segment> [area] [feature] /customer-discovery synthesize <area> <feature> /customer-discovery publish <participant-or-segment> [area] [feature] ``` Store sessions under `workspace/<area>/projects/<feature>/discovery/`. Use a neutral participant identifier when the repository should not contain personal or customer data. ## Plan Read `context/templates/customer_discovery_template.md`. Establish: - participant role, segment, and relevant context; - research questions and testable hypotheses; - what is already known and its source; - recruiting or sample bias; - recording and consent constraints. Ask about past behavior and current workarounds. Avoid pitching the solution or asking users to predict hypothetical behavior. Good pattern: "Walk me through the last time [problem] happened." Weak pattern: "Would you use [proposed feature]?" Save the plan as `<participant>-plan.md`. ## Capture Save notes as `<participant>-notes-<date>.md`. Preserve: - verbatim quotes with context; - observed actions separately from participant claims; - pain severity, frequency, and current workaround; - triggers, desired outcomes, and constraints; - solution reactions only if a concept was actually shown; - follow-ups and permission limits. Do not infer willingness to pay from enthusiasm. Record concrete commitments, current spend, or active workarounds as distinct signals. ## Debrief Save `<participant>-debrief.md` promptly after the session. Include: - three or fewer key takeaways; - evidence for and against each hypothesis; - signal strength with rationale; - surprises and contradictions; - what is observation versus interpretation; - specific follow-ups and proposed artifact updates. Offer PRD updates and publishing; do not perform them automatically. ## Synthesize Read the underlying notes, not only prior debrief summaries. Use `context/templates/discovery_synthesis_template.md` and save `synthesis-<date>.md`. - State sample size, dates, method, segments, and selection bias. - Count supporting and contradicting sessions. - Preserve important minority or disconfirming evidence. - Separate validated learning, directional signal, and untested assumption. - Recommend build, explore, or stop with a confidence level and next evidence needed. - Do not imply statistical confidence the sample cannot support. ## Publish If a configured knowledge-base connector is available, prepare a preview with privacy-sensitive fields called out. Obtain explicit confirmation before uploading interview content. If not, keep the local draft and provide manual publishing guidance.