Affinity mapping for user interviews: a step-by-step guide

You have finished the interviews. Now you have transcripts, scattered notes, memorable quotes, and a product decision hiding somewhere inside them.

Affinity mapping helps you make that material discussable. You put one observation on each movable note, group related notes without deciding the categories in advance, name the patterns, and connect those patterns to the question you need to answer.

The wall is not the finding. It is a way to inspect the evidence together.

This guide gives you a practical affinity mapping process for user interviews, including a note template, a hypothetical worked example, and the checks that stop a neat cluster from becoming an unsupported conclusion.

What is affinity mapping in user research?

An affinity map, also called an affinity diagram, organises individual pieces of information into related groups. In user research, those pieces might be:

• a participant’s action;
• a short verbatim quote;
• an observed difficulty;
• a workaround;
• a support-ticket excerpt; or
• a relevant behavioural data point.

The method is associated with Japanese ethnologist Jiro Kawakita and is also known as the KJ method. The American Society for Quality’s account of the method describes writing one idea per card, arranging related cards together, and adding descriptive headers only after the groups take shape.

That order matters. If you begin with folders called Onboarding, Pricing, and Reporting, you are classifying evidence against a structure you already chose. Affinity mapping is most useful when you let relationships in the material suggest the first structure.

It is a simple method, but not an automatic one. Researchers still choose what becomes a note, which notes belong together, and what each cluster means. Braun and Clarke’s guidance on thematic analysis makes the same point more broadly: themes are developed through active interpretation; they do not simply emerge untouched from the data.

When to use an affinity map instead of a spreadsheet

Affinity maps and coding spreadsheets solve different problems. Neither is universally better.

A useful workflow can use both. Keep the transcript or spreadsheet as the source of record, then move selected, traceable observations onto the map. After the workshop, record the final themes and decisions back in a durable document.

If you need the wider synthesis workflow first, see Maren’s guide to synthesising user interviews into themes. Affinity mapping is the collaborative sorting surface inside that larger process.

Prepare the evidence before the workshop

The quality of the map depends on the quality of the notes. Do not paste whole transcripts onto a whiteboard and expect the workshop to perform the analysis for you.

Start with one research question

Write the question at the top of the board. It should be narrow enough to influence how you sort.

Too broad:

What did users say about onboarding?

Better:

Where do new account owners stall between signing up and inviting a teammate?

The same evidence could be grouped by journey stage, unmet need, role, or cause. Nielsen Norman Group recommends using a visible focus question because several groupings may be reasonable, but only some help with the decision in front of you.

Create a snapshot for each interview

Summarise each conversation while it is still fresh. Teresa Torres’s interview snapshot captures the participant context, a memorable quote, opportunities, insights, and a simple map of the story. It gives you a manageable input for cross-interview work without replacing the transcript.

Keep enough metadata to trace each note back to its source:

• participant ID, not necessarily the participant’s name;
• interview or evidence source;
• segment or role when relevant;
• transcript timestamp or message reference; and
• whether the note is a quote, observation, or interpretation.

Traceability lets you recover context when a short note becomes ambiguous.

Put one concrete observation on each note

Write what happened, not your conclusion about it.

Weak note:

Onboarding is confusing.

Better notes:

P04 asked support where to upload the first CSV.

P07 invited an engineer to complete the data connection.

P11 left the empty dashboard after two minutes.

The better notes can be checked against the source. They may eventually support a finding about onboarding, but they do not smuggle that finding in before the sort.

Use this compact affinity mapping template:

[Participant ID] [Evidence type] [Source reference]
One action, quote, need, workaround, or observation

Example:
[P07] [Observed action] [18:42]
Invited an engineer to connect the data source

Short notes are easier to scan, but clarity matters more than an arbitrary word limit. Preserve the actor, action, and relevant context.

How to create an affinity map from user interviews

The following process works on a physical wall or a shared digital whiteboard.

1. Put every relevant note on the board

Place the notes randomly, with the focus question visible. Colour can distinguish evidence types or participant segments, but do not use colour to announce the categories you expect to find.

If the board is too large to read, reduce the scope or divide the work by a defensible boundary, such as one journey stage. Do not quietly remove inconvenient evidence to make the workshop easier.

2. Sort related notes in silence

Ask everyone to move notes into rough groups without talking or naming the groups. Both ASQ and Nielsen Norman Group’s facilitation guidance recommend silence during the early sort.

Silence gives competing interpretations room to develop before the most confident person explains what the board is supposed to mean. People can move a note someone else placed. Notes that do not fit can remain alone.

If one observation genuinely contributes to two patterns, duplicate the note and mark the copies. Nielsen Norman Group explicitly allows duplicating a note across clusters. Do not count the copies as separate evidence later.

3. Discuss the groups and challenge the edges

Once rough groups exist, start the conversation:

• What relationship holds this group together?
• Which note fits least well?
• Are we grouping by shared meaning or by a repeated word?
• Does one participant account for most of the notes?
• What contradicts this apparent pattern?
• Would a different focus question change the grouping?

Move, split, combine, or dissolve groups. A lonely note is not a failed note. It may be an outlier, a different segment, an error, or the start of the next research question.

4. Name clusters with a claim, not a topic

Reporting, Integrations, and Pricing feedback are filing labels. They do not say what you learned.

Use a short sentence with a verb:

• New account owners need engineering help before they can see their own data.
• Teammates are invited to delegate setup, not to collaborate.
• An empty dashboard looks like a broken product.

This follows Nielsen Norman Group’s advice to use descriptive labels rather than broad nouns. It also exposes weak groups. If the notes cannot support a specific sentence, you may have collected a topic rather than found a pattern.

5. Test each candidate theme against the evidence

A cluster is a group of related notes. A theme is a recurring pattern organised around a central idea and relevant to the research question.

Before promoting a cluster to a theme, check:

• Coverage: How many distinct participants or sources support it?
• Concentration: Did one talkative participant create most of the notes?
• Variation: Is the pattern stronger in a particular role, plan, or behaviour?
• Contradictions: Which evidence does not fit?
• Traceability: Can every important claim be checked against its source?
• Relevance: Does the pattern help answer the focus question?

Do not treat note count as prevalence. Ten notes from one participant are still evidence from one participant. A small cluster may describe a severe problem; a large cluster may contain repeated detail about a minor one.

Braun and Clarke’s original paper presents thematic analysis as a deliberate and rigorous interpretive process. Affinity mapping can support that process, but a wall of grouped notes is not proof that a theme is common, causal, or important.

6. Turn themes into decisions and open questions

For each defensible theme, write:

Pattern:
Supporting participants or sources:
Contradictory evidence:
Segments where it appears:
Decision or implication:
What still needs to be learned:

Prioritise with the decision criteria your team already uses, such as severity, strategic relevance, confidence, and cost to investigate or address. Dot voting can surface team judgement; it cannot convert opinions into research evidence.

Finish with an owner and next step. That next step might be a product change, a small test, a closer behavioural-data check, or five targeted follow-up interviews.

Then export the board and save the theme summary. Steve Portigal’s second edition of Interviewing Users treats making sense of data and creating research impact as explicit parts of the work. The workshop only matters if its reasoning survives after the sticky notes disappear.

A hypothetical affinity mapping example

Suppose a B2B SaaS team interviews 14 account owners about onboarding. Its focus question is:

Where do new account owners stall between signing up and inviting a teammate?

After preparing traceable notes and sorting them, the team develops three candidate themes:

1. New account owners need technical help before they can see their own data.
2. People invite teammates to hand off setup, not because they have reached a collaborative moment.
3. The empty dashboard does not explain what must happen before results appear.

The team checks each theme by distinct participant, role, account size, and contradictory evidence. The first theme is well supported among smaller accounts. The second is interesting but based on only a few participants, so the team schedules targeted follow-up conversations. The third appears less often but points to a low-cost copy and design test.

Those are different decisions because cluster size is only one input. The example is deliberately hypothetical: the participant count, themes, and actions illustrate the method, not Maren customer findings or a general onboarding benchmark.

Common affinity mapping mistakes

Starting with predefined buckets

If every note must fit your existing feature areas, the map can only reproduce your product structure. Let early groups remain provisional.

Grouping by keywords

“Pricing page confused me”, “the alternative priced us out”, and “I would pay more for this” all contain price language but describe different experiences. Nielsen Norman Group calls this keyword matching. Ask what the notes mean in relation to the focus question.

Letting one person narrate the sort

An early explanation can become the team’s answer before alternatives have had a chance to form. Protect the silent phase, especially when a founder or senior leader has a strong prior.

Losing the participant behind the notes

Without participant IDs, a dense cluster can look widespread even when it came from one long interview. Keep source labels visible and count distinct contributors, not sticky notes.

Treating themes as a vote

Qualitative interviews help explain experiences and generate grounded hypotheses. They do not estimate how common a pattern is in your full customer base. Use product data or an appropriately designed quantitative study when prevalence matters.

How to run affinity mapping solo or remotely

For a remote team, use a shared whiteboard and the same sequence: prepare notes, sort silently, discuss, test, then document. Ask participants to zoom out periodically so local clusters do not hide the shape of the whole board.

Working solo removes groupthink but not personal bias. Sort the same evidence again on another day, begin from a different part of the board, and actively search for notes that contradict your preferred explanation. Keep a short decision log explaining why you combined, split, or rejected candidate themes.

Solo or together, resist polishing the board for presentation. Nielsen Norman Group notes that the discussion and shared reasoning can matter more than the final diagram. Preserve the evidence and decisions, not workshop theatre.

Where Maren fits

Maren runs adaptive user interviews, asks follow-up questions, and gives you transcripts, per-interview findings, and cross-interview synthesis to review.

An affinity map can complement that synthesis when your team wants to inspect the evidence together, combine interview findings with support or behavioural data, challenge a preliminary theme, or build shared understanding before a consequential decision.

Maren can make the conversations and first synthesis faster. Your team still owns the focus question, interpretation, trade-offs, and decision. That is the useful division of labour: Maren helps you hear more real stories; the map helps your team reason about what those stories mean.