Affinity mapping: how to synthesize interview notes into themes and decisions

Affinity mapping is a way to make sense of a pile of research notes: you write each observation on its own sticky note, then group related notes into clusters and work out what each cluster tells you. Nielsen Norman Group defines affinity diagramming as organizing related observations, ideas or findings into distinct clusters, and notes it is also called affinity mapping, collaborative sorting or snowballing (Nielsen Norman Group). The American Society for Quality says the method, also known as the KJ method, was created in the 1960s by the Japanese anthropologist Jiro Kawakita (ASQ).

For interview research, an affinity map has four passes: pull observations out of your notes, sort them into groups, decide what each group tells you, and agree what to do next. It works best as a team exercise soon after a round of interviews.

When to use it

Use affinity mapping when you have more notes than anyone can hold in their head and the team needs to agree on what was learned. ASQ recommends it when facts or ideas seem chaotic, when an issue is too large to grasp and when group consensus matters. It says affinity diagrams tend to have 40 to 60 items, though 100 to 200 isn't unusual.

GOV.UK's service manual recommends analyzing as soon as you can after each round of research, inviting everyone who observed and allowing about one hour of analysis for every two hours of research (GOV.UK).

A worked example

This example is hypothetical. A startup building returns software for specialty retail chains interviewed eight store operations managers about how stores handle in-store returns of online orders. The three people who observed the calls now have about 120 observations between them. The steps below follow their session.

Step 1: Write one observation per note

Each person goes through their notes and writes down anything relevant, one observation per note. GOV.UK's advice is to use a single color at this stage and to write exactly what you saw or heard, such as a quote or a behavior, not what you think it means. Check recordings for exact wording where you can.

Add the participant's ID to every note, such as "P3." You will need it later. How to take customer interview notes covers keeping what people said separate from your interpretation while you take notes.

The right-hand column holds conclusions. They might turn out to be correct, but they belong in a later step, after you have compared notes.

Step 2: Sort the notes into groups

Put all the notes on a wall or a digital whiteboard in no particular order. Then everyone moves notes so that related ones sit together. ASQ recommends sorting silently at first, without deciding categories in advance, and allows anyone to move a note someone else has placed. A note that fits nowhere can stay on its own, and a note that belongs in two groups can be copied.

Expect people to disagree about where notes belong. NN/g suggests grouping notes by content before trying to name the groups. It also warns against forcing notes into clusters and against dismissing small clusters, which can hold a view only a few participants raised.

GOV.UK lists common ways groups form: by topic, by stage in a process, by step in a transaction or by type of user. In the returns example, the first groups might be "finding the online order," "issuing the refund," "what happens to the item" and "month-end reconciliation."

Step 3: Name the groups and write findings

Once the groups settle, give each one a title. ASQ suggests looking for a note in the group that captures its meaning and using it as the header, or writing a new one if none fits. GOV.UK suggests splitting large groups into smaller themes where the notes allow it.

Then decide what each group tells you. GOV.UK recommends writing each finding as a short statement on a different color note. For example:

• Finding: Store staff can't look up an online order without the customer's email or a call to the online team. (P1, P2, P3, P6, P8)
• Finding: Some stores keep their own paper or spreadsheet logs of online returns to reconcile later. (P5, P7)

List the participant IDs behind every finding. That is the check that keeps the map honest, because a cluster of 15 notes might come from one talkative person. As how to analyze customer interviews puts it, ten comments from one person still represent one person's experience. Here, the first finding rests on five of the eight managers and the second on two.

Step 4: Agree on actions

GOV.UK suggests a third color for actions, such as design ideas to work on, new questions for the next round, changes to a prototype or items for the product backlog. NN/g suggests prioritizing the clusters that matter most, using dot voting if discussion stalls, and nominating a point person to see the actions through.

For the returns startup, actions might be: test whether staff can find an online order by phone number or card; ask the next round of interviewees how they reconcile online returns; and interview someone on a retailer's online team, whose side of the process nobody has heard yet.

Then photograph or export the map and write a short summary. How to turn customer interviews into a testable product decision and running a readout with skeptical stakeholders cover what to do with it next.

The paper version still works

You don't need software. Richard Krueger's 2002 guide to focus groups describes the same idea using printed transcripts: cut out quotes, tape them to a large sheet of paper for each question, mark each quote's source with a color or a code, then write a paragraph summarizing each sheet (Krueger, University of Minnesota). His advice to take a break and look at the big picture before writing up applies to any affinity map.

Common mistakes

• Writing interpretations on the first notes. You lose the evidence and end up debating opinions.
• Deciding the categories first. You will file notes into the groups you expected and miss the ones you didn't.
• Counting notes instead of people. Always check how many participants support a finding.
• Stopping at the clusters. A wall of labeled groups isn't a set of findings. Finish with statements and actions.
• Mixing kinds of participant without marking them. If you interviewed both store managers and head-office staff, make sure each note shows which.

Your next step

Within a day or two of your last interview, book a session with everyone who observed, allowing about an hour for every two hours of interviews. Ask each person to bring their observations written one per note, each tagged with a participant ID. Sort silently for the first 15 minutes, then name the groups, write findings with the IDs behind them, and agree on no more than three actions.

If the map shows a gap, such as a role nobody has interviewed yet, that is your next recruit. Instant Expert can find people who match a description you write, such as "store operations managers at specialty retail chains." You review who it finds, it sends your invitations, and you pay only for calls that get booked. The operations professionals in retail directory page is one place to start.