Raw data (ungrouped stickies)
All observations, quotes, and ideas from the research — one per sticky. Everything goes on the board first, ungrouped. Don't organize during the dump phase. Each sticky should contain one atomic observation or idea.
Free template
Free affinity diagram template — cluster the noise into signal.
An affinity diagram turns a wall of research notes, observations, and ideas into named clusters with clear themes. The most important pattern-finding tool in a researcher's kit. Snap it when you're done.
Download on the
App Store
Free to start. Pro from $9.99/mo or $69.99/yr.
When to run this
Build an affinity diagram after customer discovery interviews, usability tests, brainstorming sessions, or any exercise that produces a large volume of unstructured observations or ideas. The diagram is the synthesis step that turns raw data into insight.
Affinity diagramming takes 60–90 minutes for a typical research set of 100–150 data points. Bring everyone who conducted the research — the clustering argument is the synthesis.
The structure
Emergent clusters
Naturally forming groups of related stickies, moved together during the sorting phase. Clusters form bottom-up — they're not predefined categories. The cluster is the unit of analysis: if two stickies belong together, they share a theme the team hasn't named yet.
Cluster headers
The name given to each cluster after it forms. Good headers are specific nouns or phrases that describe the insight, not generic categories. 'Users distrust action items that weren't created by someone in the room' is a header. 'Trust issues' is not — it's too vague to act on.
Super-clusters (second level)
Groups of related clusters at the next level of abstraction. Optional — use when you have many clusters (10+) and need to identify the major themes across the full dataset. Super-cluster headers become the section headings in the research report.
How to run it
- Dump all data (15 min)
Write every observation, quote, and idea on a sticky — one per note. Post them randomly on the board. No organization yet. If you have interview transcripts, pull notable quotes directly. Quantity matters at this stage.
- Read and silently cluster (20 min)
Everyone reads all the stickies individually. Begin moving stickies together that feel related. Do this in silence — the clustering should be intuitive, not argued. If two people move the same sticky to different places, that's data about an ambiguous observation.
- Debate outliers and overlaps (15 min)
Discuss stickies that are between clusters, stickies that don't fit anywhere, and clusters that may be too large (split them) or too small (merge them into adjacent clusters). Every sticky should end up in a cluster.
- Name the clusters (10 min)
Write a header sticky for each cluster. The header names the insight the cluster represents. This is the hardest step — naming forces you to commit to what the cluster actually means.
- Form super-clusters if needed
If you have 10+ clusters, group related clusters into 3–5 super-clusters with their own headers. Super-clusters become the major themes in the research debrief.
- Snap and report
Snap the affinity diagram with BoardSnap. The AI reads the cluster structure — headers and individual stickies — and outputs a structured theme summary that becomes the foundation of the research report.
Why affinity diagrams on a whiteboard + BoardSnap is better than digital
Affinity diagramming is the use case that makes physical whiteboards irreplaceable. The ability to move 150 stickies around a large surface, to cluster them, to step back and see the whole picture, to rearrange clusters as patterns emerge — this is an inherently spatial, physical activity. Digital tools impose grid constraints and load times that interrupt the flow of the clustering process.
The capture challenge for affinity diagrams is real — a full board of 150 clustered stickies is a dense artifact that's hard to photograph legibly. BoardSnap's VisionKit perspective correction flattens the board and the AI reads the cluster structure, preserving the theme names and the key observations within each cluster as a structured synthesis.
Frequently asked
How many stickies can an affinity diagram handle?
50–200 is the practical range. Below 50, the exercise is usually overkill — just read the notes. Above 200, the clustering phase takes too long and the clusters become too granular to be useful. For very large datasets, cluster in batches by interview or by research question, then merge the clusters.
What's the KJ Method?
The KJ Method (named for Jiro Kawakita, who developed it in the 1960s) is the formal name for affinity diagramming. The method specifies: write observations on cards, spread them randomly, cluster silently, name clusters. Affinity diagramming as used in UX and agile practice is a direct descendent of the KJ Method.
Should clusters be predefined before the sorting session?
No. Predefined categories turn affinity diagramming into card sorting — a related but different technique. The power of affinity diagramming is that clusters emerge bottom-up from the data, not top-down from a framework. Predefined categories mean you can only find patterns you already expect.
How many people should do the clustering?
Two to five. Too few and you miss interpretive diversity; too many and the session becomes chaotic. The ideal group includes everyone who conducted the research — they can arbitrate ambiguous stickies from memory of the interview context.
Can I use affinity diagramming for brainstorming ideas, not just research data?
Yes. After a brainstorm session, use affinity diagramming to cluster the ideas into themes. The technique works equally well for any large set of unstructured data — the input just shifts from observations to ideas.
Run your next affinity diagram and BoardSnap will summarize it.
No exporting, no transcription. Snap the board, get the action plan.
Free · 1 project, 30 boards
Pro $9.99/mo · everything unlimited
Pro $69.99/yr · save 42%