Customer Interview Insight Synthesis

The Prompt

You are a product research analyst. Synthesize multiple customer interviews into reliable, decision-ready insights.

INTERVIEW NOTES:
{{interview_notes}}

RESEARCH GOAL:
{{research_goal}}

TARGET USER SEGMENT:
{{target_user_segment}}

DECISION CONTEXT:
{{decision_context}}

Return exactly:
1) Evidence quality check
   - what is strong enough to use
   - where the notes are too thin, uneven, or ambiguous
   - where the sample is skewed
2) Evidence-backed findings (5 to 8)
   - finding statement
   - confidence
   - evidence snippets or paraphrases
3) Tension map
   - where users disagree
   - likely segment or context explanation
4) Opportunity areas
   - ranked by impact x feasibility
   - note what evidence supports each opportunity
5) Decision implications
   - what this suggests for the named decision context
   - what still cannot be decided yet
6) Do-not-conclude section
   - what the current data does not justify
7) Recommended next-step experiments or research actions (3)

Rules:
- Separate observed behavior from user opinion.
- Avoid generic "users want simplicity" statements unless evidenced.
- Explicitly call out sample-size limitations.
- If multiple interviewees are represented unevenly, say so.
- If `decision_context` is vague, explain how that limits prioritization quality.

When to Use

Use this after running roughly 5-20 discovery interviews and before roadmap or feature-priority meetings. It is especially useful when several stakeholders heard different stories and need one evidence-led synthesis before a decision gets made.

Variables

Variable	Description	Example
interview_notes	Combined raw notes, transcript excerpts, tagged summaries, or coded observations	”10 interview summaries from Notion with quotes and researcher notes”
research_goal	The question the research is meant to answer	”Why activation drops after week 1”
target_user_segment	Cohort definition for these interviews	”Solo consultants with <10 employees”
decision_context	The upcoming product, design, or prioritization decision tied to the research	”Prioritize Q2 onboarding improvements”, “Decide whether to simplify setup or improve education first”

Tips & Variations

• Include interview counts or simple identifiers inside interview_notes when possible. That makes confidence judgments much more useful.
• Ask the model to separate “what users said” from “what they actually did” when notes mix direct quotes with researcher interpretation.
• If the first output feels generic, the usual fix is stronger notes with concrete evidence snippets, not just asking for “deeper insights.”
• For regulated products, include “flag claims requiring legal/compliance validation.”
• If the goal is prioritization, ask for one follow-up section that compares opportunity size versus evidence strength so the loudest quote does not win by default.

Example Output

Evidence quality check: The sample covers only new users from one segment, so onboarding insights are stronger than long-term retention conclusions.

Finding: Users trust automated suggestions only after manual preview.

Evidence: 7 of 10 interviews described fear of publishing wrong information without preview.