UX Research · Step 6 · 7 min read
From notes to themes, affinity mapping
Affinity mapping turns a wall of research notes into a few themes a team agrees on. Where the method comes from, how to write notes that cluster well, why you need a focus question, how to sort without keyword matching or groupthink, and how to name themes so they say something.
Six interviews give you hundreds of things people said and did. Nobody can hold hundreds of things in their head. Affinity mapping is the oldest trick for that: write each thing on its own note, put them all where everyone can see, and move them around until groups appear. It looks like a craft activity. Done well, it's analysis that a whole team does together, and the team comes out believing the result because they built it.
What you'll be able to do after reading:
- Explain what affinity mapping is, and where it comes from
- Write research notes that can be sorted
- Run a sorting session with a focus question, in silence
- Turn clusters into themes that are sentences, not labels
1. What it is
NN/g defines affinity diagramming as organising related observations, ideas or findings into distinct clusters, usually with physical or digital sticky notes. It also goes by affinity mapping, collaborative sorting and snowballing (NN/g, Affinity Diagramming, 2024). It isn't card sorting: card sorting is a research method with participants, to learn how they group content (see Card sorting and tree testing); affinity mapping is how your team groups its own findings.
The method is usually credited to the Japanese researcher Jiro Kawakita, who devised it in the 1960s; that's why it's also called the KJ method (Wikipedia, Affinity diagram). The anthropologist Raymond Scupin described it, for other anthropologists, as a technique for analysing ethnographic data (Scupin, Human Organization, 1997). It's also frequently used in contextual inquiry to organise field notes (see Contextual inquiry and diary studies). The UX Lead course has a short facilitation page on it: Affinity mapping.
2. Notes that cluster well
Notes that make clustering hard
Users hate remindersA conclusion, not an observation, and nobody knows who said it
P3 talked about money, her flatmates, the app, her bank and a fight in JuneFive things on one note: it can't go in one group
Confusing!What was? Where? For whom?
Notes that cluster well
P3 waited three months before asking Tom for €140P3 · SAW
"Asking feels like I don't trust them."P5 · QUOTE
P1 keeps her own spreadsheet and checks the app's total against itP1 · SAW
The board can only be as good as its notes. Write them from your notes and transcripts, before the session, with four rules:
- One observation per note. A note that holds five things can't go in one group.
- What happened or was said, not what you concluded. "Users hate reminders" is a conclusion; it belongs at the end of the process, not on the wall at the start.
- Who it came from. A participant code, like P3, lets you count how many people a theme rests on, and go back to the source.
- Quote or observation, marked. NN/g suggests different colours for direct quotes and things you observed. Add a written label too, so colour isn't the only signal.
If you have a lot of data, NN/g suggests a first pass to choose the most relevant data, or splitting into smaller groups, to avoid session fatigue.
3. The focus question
1 · Notes on the wallOne observation per note, from six interviews
"The reminder sounds like me nagging" P4
P3 waited 3 months to ask Tom for €140
P5 keeps receipts as photos in the group chat
P1 checks the app against her own spreadsheet
P2's flat settles up over dinner on the 1st
"Asking feels like I don't trust them" P5
P1 settles up on rent day
P2 rounds amounts down so she doesn't look petty
P6 doesn't believe the total after edits
P4 turned reminders off
2 · ClustersFocus question: what gets in the way of settling up?
Waiting, so it doesn't feel rude
P3 waited 3 months to ask Tom for €140
"Asking feels like I don't trust them" P5
P2 rounds amounts down so she doesn't look petty
"The reminder sounds like me nagging" P4
P4 turned reminders off
Checking the app against my own record
P1 checks the app against her own spreadsheet
P6 doesn't believe the total after edits
P5 keeps receipts as photos in the group chat
Settling up on a fixed date
P2's flat settles up over dinner on the 1st
P1 settles up on rent day
3 · ThemesEach cluster becomes a sentence you could disagree with
Asking for money feels rude, so people wait and the debt grows5 notes · P2, P3, P4, P5
People trust their own record more than the app's total3 notes · P1, P5, P6
A fixed date takes the asking out of it2 notes · P1, P2
"Sort these by theme" isn't enough. Maddie Brown, at NN/g, learned this the hard way: teams stood in front of the wall unsure how to begin, because there are many valid ways to sort the same notes. Her fix is a focus question, written where everyone can see it, which decides what "belongs together" means (NN/g, Avoiding 3 Common Pitfalls of Affinity Diagramming, 2023).
The same notes sort differently under different questions. Under "What gets in the way of settling up?", P4 turning reminders off goes with the notes about waiting to ask. Under "How do people use the app's features?", it would go with feature usage. Both are valid; only one answers your study's research question. Take the focus question from the research plan.
4. Sort in silence, by meaning
The session itself is simple. Each person posts their notes; then everyone moves notes into groups. Two pitfalls ruin it, both from Brown's article:
- Keyword matching. Putting "Menu button", "Login button" and "Add to cart button" together because they share a word. You end up with clusters called "Button issues" that say nothing. Watch for clusters held together by a word instead of a meaning, and say so out loud.
- Groupthink. A senior person, an enthusiast or an expert suggests a way of sorting in the first minute, and everyone follows. Brown's remedy is silence, especially at the start, so different sorting strategies can compete before one wins.
Some practical habits from NN/g's guide:
- Group first, name later. It keeps people from forcing notes into a label they picked too early.
- Don't force clusters. There's no right number of groups.
- Don't dismiss small clusters. A group of two notes can be the most important one on the wall.
- Duplicate a note if it really belongs in two groups.
- Keep the outliers in a visible "doesn't fit" corner. Sometimes they're the start of a theme you haven't seen yet.
For a remote team, a digital whiteboard works the same way. Agree the colours, ask people to stay muted for the first round of sorting, and copy the finished board before anyone tidies it.
5. From clusters to themes
A cluster is a pile; a theme is a claim. Name each group as a sentence that someone could disagree with:
| Cluster label | Theme |
|---|---|
| Reminders | Asking for money feels rude, so people wait and the debt grows |
| Trust | People trust their own record more than the app's total |
| Timing | A fixed date takes the asking out of it |
The left column only says what the notes are about. The right column says what they mean, and you can check it: is it true for most of the notes in the group? How many participants does it rest on? Write the count next to each theme, as in the board above.
Then NN/g's last step: prioritise, with discussion or dot voting, and leave with next steps and an owner for each (NN/g, 2024). NN/g also notes that the conversations while building the board matter more than the final picture. That's the strongest reason to involve the team instead of doing it alone.
Affinity mapping is fast and social. It's less systematic than Thematic analysis, the next lesson, which codes every segment of every transcript. Many studies use both: affinity mapping with the team the day after the last session, thematic analysis to check and deepen the themes before the report.
6. Common mistakes
- Notes that are conclusions ("users are confused") instead of observations.
- No participant code on the notes, so nobody can count or check.
- Sorting without a focus question.
- Clusters held together by a shared word.
- The most senior person sorting first, out loud.
- Theme names of one word.
- Photographing the board and never looking at it again.
7. Vocabulary
| Term | Meaning |
|---|---|
| Affinity mapping | Sorting notes into groups by what they have in common |
| KJ method | Another name for it, after Jiro Kawakita |
| Focus question | The question that decides what "belongs together" means |
| Cluster | A group of notes on the board |
| Theme | A cluster named as a claim about what it means |
| Keyword matching | Grouping by shared words instead of shared meaning |
| Dot voting | Prioritising by giving each person a few votes to place |
8. Try it (no tools needed)
- Write ten things you heard about a product or service this week, from friends, colleagues or reviews, one per scrap of paper, with who said it.
- Write a focus question: "What makes ___ hard?"
- Group them in silence. Then name each group as a sentence.
Reflection questions: Would your groups change under a different focus question? Which note didn't fit anywhere, and what might it be telling you?
9. If you already practise
- Prepare notes before the session, from transcripts, so the session is sorting, not writing.
- Start every session with the focus question and the rule of silence, and name keyword matching as a thing you'll call out.
- Write the participant count on every theme, and carry it into the report.
- Try an AI tool to draft notes from transcripts, then check each one against the source; AI-assisted synthesis, step by step shows how.
Checklist
Learn more
- Kara Pernice, Affinity Diagramming for Collaboratively Sorting UX Findings and Design Ideas (NN/g, 2024)
- Maddie Brown, Avoiding 3 Common Pitfalls of Affinity Diagramming (NN/g, 2023)
- Raymond Scupin, The KJ Method: A Technique for Analyzing Data Derived from Japanese Ethnology (Human Organization, 1997)
Books, courses and videos to go deeper
Book links go to Amazon. As an Amazon Associate I earn from qualifying purchases, at no extra cost to you.
Related
Affinity mapping · Thematic analysis · Notes, recordings and transcripts · AI-assisted synthesis, step by step · Synthesise a study