How to do affinity mapping — and avoid the clustering mistakes that kill the insight.

Affinity mapping, also called KJ Method (after its inventor Jiro Kawakita), is the standard technique for making sense of large volumes of qualitative data — user interviews, survey responses, field observations.

Phase 1 — Data generation. Each participant writes observations on sticky notes — one observation per note. Notes should be written as direct observations or quotes, not interpretations. "User opened the settings menu 4 times looking for the export button" is an observation. "Users find the UI confusing" is an interpretation. Observations make better affinity notes because they can cluster in unexpected ways; interpretations predetermine the categories.

Source your notes from research data: interview transcripts, session recordings, survey responses. If you're doing affinity mapping during a workshop, each person writes what they saw or heard in the research sessions they attended.

Phase 2 — Silent clustering. Post all notes on a large surface — a whiteboard, a wall, a digital canvas. Then cluster in silence. No talking. Each person moves notes toward similar notes. If someone moves a note you disagree with, move it back. Disagreements are resolved by repetition: if you move it back three times and someone else keeps moving it, leave a copy in both clusters. This physical, silent process prevents groupthink — louder voices don't dominate the clustering logic.

Clustering typically takes 15–30 minutes depending on the volume of notes. Stop when notes stop moving significantly.

Phase 3 — Cluster naming. Now break the silence. For each cluster, the group writes a header note — a phrase that captures what all the notes in the cluster have in common. Good cluster names are specific: "Users can't find export" is better than "Navigation issues." Headers should be in the same vocabulary as the notes — phrases users would use, not category labels that reflect your internal taxonomy.

Phase 4 — Prioritization. Give each participant 3–5 votes (dot stickers) to mark the most important clusters — important meaning highest impact on the user or the product. The top clusters become the focus of design decisions, backlog prioritization, or further research.

Common mistakes: Writing interpretations instead of observations. Talking during the clustering phase. Naming clusters before all notes are placed. Stopping at one level — with enough notes, sub-clusters within clusters reveal important nuance.

Snap the completed affinity map with BoardSnap. The AI reads the cluster headers, note content, and vote counts and produces a structured summary with ranked themes — ready to paste into a research readout or product brief.