| name | customer-research-synthesis |
| description | Use when raw qualitative research — interview notes, survey responses, session recordings — needs to become structured insights that drive decisions. Triggers on: "인터뷰 정리", "고객 인터뷰 분석", "리서치 결과 정리", "설문 분석", "VOC 정리", UX research synthesis", |
| license | MIT |
| metadata | {"author":"wondelai","version":"1.1.0"} |
| scenarios | ["I have 10 interview transcripts — help me find the patterns.","고객 인터뷰 결과를 정리해서 PRD에 넣을 인사이트로 만들어줘.","Synthesize these survey verbatims into themes with evidence counts.","인터뷰 노트에서 Jobs-to-be-Done을 뽑아줘.","Which of these research findings are real patterns vs. one-off anecdotes?","VOC 데이터를 우선순위 결정에 쓸 수 있게 정리해줘."] |
| compatibility | {"recommended":["think-tool","sequential-thinking"],"optional":["mcp-reasoner"],"remote_mcp_note":"think-tool이 있으면 패턴/일화 판정과 JTBD 추출의 논리를 검증하는 데 도움이 됩니다. sequential-thinking은 6단계 체인 전체 흐름 유지에 유효합니다. Claude 설정 → MCP Servers에서 remote SSE 엔드포인트를 추가하세요."} |
Customer Research Synthesis
Raw quotes are not insights. An insight requires an observation, supporting evidence, and an implication for what to build or change.
Synthesis chain: Raw data → affinity clusters → Jobs-to-be-Done → insight cards → hypotheses
For the full step-by-step process with examples, load references/synthesis-process.md.
When to Use / When Not to Use
| Use | Do Not Use |
|---|
| Raw interview notes needing thematic structure | Quantitative survey with closed-ended stats |
| Verbatim quotes needing n= pattern analysis | Competitive research (use competitive-analysis) |
| Discovery research feeding a PRD or roadmap | Product analytics / behavioral data (use metrics-interpretation) |
| Post-research readout that needs insight cards | Research planning or question design |
Mode 1: Synthesize Raw Data
When the user provides raw interview notes, survey verbatims, or usability observations, produce:
- Affinity clusters — grouped themes with n= counts (e.g., "Pain: slow onboarding — 6/10 participants")
- Pattern vs. anecdote verdict for each cluster (see thresholds in
references/synthesis-process.md Step 3)
- Top 3 Jobs-to-be-Done extracted from the data
Run Step 3 (pattern judgment) and Step 4 (JTBD extraction) in parallel after affinity mapping — they are independent.
- Insight summary cards — one per major finding (observation + evidence + implication)
- Recommended hypotheses — feeds
../hypothesis-driven-dev/SKILL.md
Mode 2: Evaluate Synthesis Practice
When the user asks "how good is our research process?" or shares their existing synthesis output for review, rate 0–10:
| Score | State |
|---|
| 9–10 | Patterns documented with n= counts; JTBD extracted; insight cards connect to decisions; hypotheses written before building |
| 7–8 | Themes exist but evidence counts missing; insights stop at observation without implication |
| 5–6 | Research summarized as a quote list; no pattern analysis |
| 3–4 | "We talked to users" treated as sufficient; no structured output |
| 1–2 | No synthesis practice; decisions made without reference to research |
What Claude Does / What You Do
| Claude | You |
|---|
| Groups observations into affinity clusters with verb phrases | Conducts interviews and recordings |
| Counts n= occurrences and flags pattern vs. anecdote | Judges whether participants are representative |
| Extracts Jobs-to-be-Done using the JTBD template | Validates JTBD accuracy against live customer context |
| Writes insight cards (observation + evidence + implication) | Makes final decisions on what to build |
| Chains insights to testable hypotheses | Decides experiment priority and runs experiments |
Related Skills
../feature-prioritization/SKILL.md — insight cards as prioritization evidence../prd-development/SKILL.md — problem statement, personas, and Jobs-to-be-Done../hypothesis-driven-dev/SKILL.md — validated hypotheses ready for experiment design../pricing-monetization-strategy/SKILL.md — WTP signals from research../product-discovery/SKILL.md — triggered after customer research phase completes
