assets/example-outputs/insight-cards.md
Output: Insight cards
When: the goal is to extract cross-interview insights. The default synthesis output (S5–S7).
Input: nuggets with clusters → score_insights.py. Write cards only for insight/watchlist.
Card format
### [ID] Title — one sentence
Statement: [observation + interpretation + implication — not a retelling]
Holds for: [roles/segments] · NOT for: [where it doesn't work]
Prevalence: X of N interviews (roles: …) · Criticality: high/medium/low
Type: CONSENSUS | TENSION ([role A] vs [role B] — on what)
Evidence (verified + support=yes only):
- [Interview, role] «verbatim quote» (Lxx)
- [Interview, role] «…» (Lxx)
Counter-evidence: [who didn't confirm / what contradicts / where there's silence]
Confidence: high/medium/low (based on triangulation + grounding)
Implication / recommendation: [what to do; for whom]
Rules
- A pattern = ≥k different interviews with a verified quote (not k quotes from one person).
- Every card goes through counter-evidence.
- Scope by role: "for collections staff", not "for employees".
- Sorting:
insightby frequency×criticality →watchlist(tensions, rare-but-critical).weak— into the appendix. - Optionally → HTML board:
render_board.py provenance.json.