UX Glossary | A

Affinity diagram

Also: Affinity map, Affinity mapping, KJ method

Grouping individual research observations by similarity until the themes emerge from the data.

Finding things Searched for invoice P1 Gave up, rang support P2 Lost track of orders P4 Trust the total Checked total twice P2 Unsure card was saved P4 Doubted the free label P5 Too much typing Typed address by hand P1 Too many form fields P2 Wanted autofill P5 One voice: a quote, not a finding. Wants dark mode P3 Finding things Searched for invoice P1 Gave up, rang support P2 Lost track of orders P4 Trust the total Checked total twice P2 Unsure card was saved P4 Doubted the free label P5 Too much typing Typed address by hand P1 Too many form fields P2 Wanted autofill P5 One voice: a quote, not a finding. Wants dark mode P3
Affinity diagram The notes gather first and the names arrive last, so the themes come from the data. The note nobody else echoed stays a quote.

An affinity diagram turns a pile of individual observations into a set of themes. Each observation goes on its own note, the notes are grouped by similarity, and the groups are named only after they have formed.

The method comes from the Japanese anthropologist Jiro Kawakita, who developed it in the 1960s. It is still sometimes called the KJ method after his initials.

Why the order matters

The discipline is in naming the groups last. Deciding the categories first and sorting observations into them tests a hypothesis you already had. Letting groups form from similarity lets the data say something you did not expect, which is the entire point of qualitative research.

In practice

After ten interviews you might have 300 notes. Grouping them typically yields somewhere between five and ten themes, each supported by observations from several different participants. A theme carried by a single participant is a quote, not a finding, and labelling it as such keeps the synthesis honest.

Doing this as a team also has a side effect worth having: everyone who sorts the notes ends up having read the raw data, which makes the findings much harder to dismiss later.

Knowing the words is the easy part.

If you'd rather have someone apply them to your product, that's what I do.