Affinity mapping, step by step
Last reviewed by The thinkdrop facilitation team
Affinity mapping means sorting a pile of individual notes into themes that emerge from the notes themselves rather than from categories you decided in advance. Group in silence first, name the clusters only after they exist, keep a miscellaneous pile, and timebox the whole thing to fifteen minutes or it will expand to fill the session.
Key facts
- Timebox
- 15 minutes for up to 60 notes
- Silent grouping
- First 5 minutes
- Target clusters
- 5–8, never more than 10
- Naming
- After grouping, never before
- Miscellaneous pile
- Always allowed
Step by step
- Get every idea onto its own card — One thought per card. A card containing two ideas will be argued about for five minutes and then split anyway, so split it at the start. Short beats complete: the card is a pointer to a conversation, not the conversation.
- Read everything before touching anything — Two minutes of silent reading. Groups that start moving cards immediately end up with themes that reflect the first ten notes read rather than the whole set.
- Group in silence — Five minutes where anyone can move any card and nobody explains why. Cards that get moved back and forth are the interesting ones — they belong to two themes, which is information rather than a problem. Silence stops the exercise turning into a debate about taxonomy.
- Name the clusters last — Only once the piles exist should you name them, and the name should describe what is in the pile rather than the category you hoped for. "Onboarding" is a category; "People do not know what to do after signing up" is a theme, and it is far more useful.
- Keep a miscellaneous pile and do not force it — There is always a set of cards that belong nowhere, and some of them are the most original ideas in the room. Forcing them into a theme is how good ideas get lost. Leave them visible.
- Vote on themes, not on cards — Once themes exist, dot-vote on themes. Voting on individual cards rewards whoever wrote the most cards on a topic rather than the topic's importance.
Why grouping eats the clock
Affinity mapping is the phase that overruns in nearly every session that overruns. The cause is almost always the same: the group is trying to produce a correct taxonomy rather than a workable one. There is no correct grouping of 60 subjective notes. Aim for good enough to decide on, and set a visible timer.
When it works and when it does not
It works on a pile of comparable items: research notes, retro observations, ideas from a brainstorm, support tickets. It works badly on items of different types — mixing a strategy statement, a bug and a feature request produces clusters organised by type rather than by meaning. Sort by type first if your pile is mixed.
- Good input: 30–80 comparable, one-thought notes.
- Bad input: fewer than 15 notes, where themes are already obvious.
- Bad input: mixed types — split those into separate maps.
Doing it remotely
Remotely, silent grouping is easier than in person because nobody can hover. What is harder is seeing the whole set at once, so keep the card count below 60 and give a longer reading window. Automatic clustering is useful as a first draft, provided the group treats it as something to correct rather than accept.
Frequently asked questions
How many themes should come out of it?
Five to eight. More than ten means you have re-listed the cards; fewer than four usually means the themes are so broad they cannot be acted on.
Who moves the cards?
Everyone, at the same time, in silence. A single person sorting on behalf of the group produces one person's mental model with an audience.
Can AI do the grouping?
It can produce a fast first draft, which is genuinely useful on 60 notes. Treat the output as a starting arrangement the group edits, because the disagreements about where a card belongs are where the insight is.