Affinity chart
Silent clustering of notes until the themes name themselves.
Groups after the notes, not before
An affinity chart, or affinity diagram, takes many separate observations and clusters them by kinship. Jiro Kawakita's KJ method is the source of the practice most workshops use. The rule that makes it different from a fishbone diagram is the order: you do not start with category headers. Headers are written only after a cluster exists. Starting with headers turns the exercise into sorting mail into bins you already believed in.
It pairs well with brainstorming. Brainstorming produces the notes. The affinity chart organizes them. A stakeholder map is a different sort, because its axes are chosen in advance.
Example
Twenty notes from five people about a signup form: 'date field accepts last year', 'button label says Submit', 'error text is red on red', 'no confirmation email', 'phone number required for a club'. In silence, the group clusters. One cluster is about validation, one about the submit button's wording, one about what happens after success. Name the clusters only when moving a note would feel wrong. If a note fits two clusters, duplicate it. Do not average it.
Steps
- One fact per note. A note with two ideas will sit between clusters forever.
- Sort in silence first. Talking early lets the fastest voice name the bins.
- Name the cluster with a noun phrase. Use words from the notes, not a slogan from outside.
- Stop at a handful of clusters. Seven clusters from twenty notes means the sort is not finished, or the question was two questions.
Questions
- What is an affinity chart?
- A board where individual notes are clustered by similarity, and the cluster is named only after the group forms. It finds themes in unsorted observations.
- Why sort in silence?
- Silence keeps the first suggested category from pulling every note toward it. Discussion is useful after clusters exist.
- How is an affinity diagram different from themes I write first?
- Headers written first are hypotheses. An affinity chart tests whether the notes actually clump. If you already know the categories, you may want a fishbone instead.