user-interview-synthesis

Start by extracting verbatim, participant-tagged observations with no interpretation, then cluster them bottom-up into themes. Grade each finding by evidence strength (behavior outweighs stated opinion) and attach a so-what implication linking it to a decision or follow-up question.

The workflow targets batches of 5-12 interviews, which is enough to establish that a pain exists and runs deep. Small-n findings must ship graded with n/N notation and never be presented as market-share percentages, which require survey-based quantification.

Actively list observations that contradict the team's hypothesis and give them their own section in the readout. Check whether repeated quotes are representative or merely quotable, and treat a readout that confirms everything the team believed as a warning sign requiring evidence.

Use surveys when you need to quantify how widespread a finding is, since interviews with small samples cannot support percentage claims. Interview synthesis establishes that a problem exists and its depth; survey design handles the quantification step downstream.

Keeping the verbatim observation layer interpretation-free prevents downstream synthesis failures, since every major distortion comes from smuggled interpretation. Tagging observations to participant IDs also makes it mechanically visible when one vivid participant dominates a theme.