Blog

Affinity mapping: turning interview noise into themes

July 12, 2026 · 6 min read

After a round of interviews you don't have findings — you have noise. Forty observations, some contradictory, some duplicated, a few brilliant. Affinity mapping is the classic method (born as the KJ method in 1960s Japan) for turning that wall of notes into a handful of themes you can actually defend.

Why bottom-up beats top-down

The discipline of affinity mapping is that you don't start with categories. You start with the evidence and let structure emerge: cluster notes that feel related, split clusters that grow vague, and only then name what you're looking at. Top-down sorting confirms what you already believed; bottom-up clustering is how research surprises you.

A practical remote workflow

• One observation per note — atomic, in the participant's words where possible, with the participant attached.
• Cluster by feel first. 'These belong together' is enough; you don't need to say why yet.
• Name clusters late. A premature label attracts notes that don't belong to it.
• Let small clusters live. Two strong notes can be a theme; twelve weak ones might be noise.
• Keep an unsorted pile. Forcing every note into a group manufactures false patterns.
• Merge ruthlessly at the end — near-duplicate themes are how taxonomies rot.

Where AI clustering fits

With sixty notes, the blank-wall moment is real. AI is good at the opening move: proposing a first grouping of notes into candidate themes, clearly labelled as suggestions. Treat it as a starting arrangement, not an answer — the value of affinity mapping is the arguing and rearranging, and that stays with the humans who heard the interviews. A suggested cluster you disagree with is doing its job: it made you articulate why.

From clusters to insights

A named cluster with evidence behind it is a candidate insight. Write it up as a claim — “technicians don't trust offline sync”, not “sync stuff” — attach the strongest quotes from the cluster, rate severity and confidence honestly, and it graduates from the board into the repository, where it can be found, cited, and eventually triaged toward the roadmap.

How Lens helps

Lens gives each project an affinity board: drag notes between clusters, rename, recolor, and merge groups, with every note carrying its source participant. One click asks AI to propose clusters across your notes — marked as suggested, rearrangeable like anything else. When a cluster hardens into a finding, capture it as an insight with its evidence attached. For the step after that, see Insight triage: from validated finding to product roadmap.

← All posts