Affinity Diagramming
A collaborative method for clustering individual observations or ideas by similarity to reveal patterns, disagreements, and next steps.
Download Markdown (.md)Soroush’s Product Wiki · https://wiki.soroush.uk/topic/affinity-diagramming/
01What it means
Affinity diagramming organizes discrete notes into emergent groups based on perceived relationships, then labels and examines those groups. In research synthesis it helps a team externalize observations and discuss patterns; it is not the same as card sorting with participants to study an information architecture.[1][2]
- 01Capture observationsOne evidence-bearing idea per note
- 02Cluster silentlyGroup by perceived relationship, not preset categories
- 03Discuss movementSurface different interpretations and context
- 04Name themesUse specific labels grounded in the contents
- 05Review exceptionsOutliers, contradictions, and missing evidence
- 06Choose next actionQuestion, synthesis, or decision the grouping supports
Clusters are interpretations rather than findings by themselves; preserve source context and avoid forcing every item into a tidy theme.
02Why it matters
Collaborative clustering can make a large set of observations manageable and expose how team members interpret evidence. It builds shared understanding and provides a bridge from session notes to questions that deserve deeper analysis.
03When to use it
Use it after research sessions or in workshops when the material can be represented as atomic notes and collective interpretation is useful. Use a more rigorous analytic method for large, sensitive, or theory-building datasets where traceability and sustained interpretation are essential.
04How it works
- Prepare atomic notes that retain source identifiers and distinguish observation from interpretation.
- Have participants review and initially sort related notes without forcing preset categories.
- Discuss ambiguous placement, duplicate notes when relationships cross groups, and preserve outliers.
- Name clusters with specific evidence-based labels and examine relationships or gaps.
- Trace candidate findings back to raw data before prioritizing actions or communicating conclusions.
05A practical example
After six onboarding sessions, a team clusters observations about delayed permissions, unclear ownership, and repeated approvals. It labels a broader coordination pattern, then checks recordings before changing the administrator workflow.
06Where it breaks
Dominant voices, premature labels, and selective notes can manufacture consensus. Clusters show perceived similarity, not causality or prevalence, and a workshop board can lose the context of the original data.
Common mistakes
- Writing conclusions rather than atomic observations on the notes.
- Forcing every note into a neat cluster and discarding exceptions.
- Treating the largest cluster as automatically the most important finding.
07Sources
- Affinity Diagramming for Collaboratively Sorting UX Findings and Design IdeasNielsen Norman Group · Professional guidanceAffinity Diagramming for Collaboratively Sorting UX Findings and Design Ideas
- Analyse a research sessionGOV.UK Service Manual · Government service guidanceAnalyse a research session