Turn a pile of user interview notes into findings that change the roadmap
Get past the quote collection to the pattern nobody named out loud.
- Use it for
- PMs, researchers, and founders who just ran 6-10 user interviews and have pages of raw notes.
The default output of interview synthesis is a theme list that restates what people said: "users want faster onboarding," "users find pricing confusing." True, and useless — you already half-knew that from doing the interviews, and it doesn't tell you what to build. The value in synthesis isn't collecting what people said; it's finding what's true across people that no single interview stated directly, because no single person has visibility into the pattern their own answer is part of.
This prompt is built to resist the easy output. It asks for contradictions, for the thing said by the fewest people that matters most, and for what you'd have to be wrong about for each finding to be wrong — which is the part that makes a finding actionable instead of just plausible.
When not to use this
Needs real notes, not a memory of the interviews — vague recall produces vague synthesis. And it's not a substitute for going back to a transcript when a finding is contested; this surfaces candidates for patterns, it doesn't replace checking them against what was actually said.
Did this work?
At least one finding surprises you — something that wasn't obvious from any single interview but is obvious once it's named across all of them.
Tested on claude-opus-5. Evidence status is draft; it moves to battle-tested only on recorded runs, never by hand.