What problem does it solve? Raw research data — interview transcripts, field notes, diary entries, open survey text — often gets reduced to cherry-picked quotes, invented percentages, or vague themes that cannot be traced back to evidence. This Skill turns raw qualitative (and mixed qual+quant) data into defensible findings by enforcing traceability, a mandatory disconfirming-case hunt, honest prevalence counts, and calibrated confidence levels. ## Core Features & Use Cases - Situational approach selection: Routes to thematic coding, affinity diagramming, framework-mapping (JTBD/journey/matrix), or mixed qual+quant synthesis based on the decision, data, and problem shape — never one fixed pipeline. - Rigor constants: Enforces source-tagged traceability, a mandatory disconfirming-case hunt before accepting any theme, prevalence as "X of Y" (never bare percentages), a High/Med/Low confidence level per finding, and reporting of negative cases. - Define-stage outputs: Produces linted user need statements and How-Might-We questions built only on High/Med-confidence insights, each tied to a decision and a metric. - Use Case: You have 11 interview transcripts from a chatbot beta test. The Skill atomizes the data, catches that a vivid "users love auto-booking" quote is only 2 of 11 participants, and produces a High-confidence finding (6 of 11 need a confirm-and-undo step) with a need statement and HMW question. ## Quick Start Synthesize these interview transcripts into themes, insights, need statements, and How-Might-We questions with evidence counts and confidence levels.
