How to do affinity mapping, with a worked example

To do affinity mapping, write each finding on its own sticky note and put all the notes on one board. Sort the notes into groups of similar findings, name each group after what its notes have in common, then rank the groups and write a next step for the ones that matter.

Nielsen Norman Group, or NN/g, defines it as "organizing related observations, ideas, concepts, or findings into distinct clusters". Its three steps are to generate notes, organise them into clusters, and prioritise the clusters and next steps. This guide follows those steps and spends extra time on the names, because a group called "Trust" doesn't yet tell anyone what to do.

We make Prompt Board, and our affinity map template opens on the nine notes below. Sticky notes on a wall work too.

Write notes you can sort

A note holds one finding and says what someone did or said. If it holds your conclusion instead, there's nothing left to sort.

The middle row becomes two notes, because those two findings may belong in different groups.

NN/g has each person write notes independently, "to avoid being biased by other team members or conversations", and suggests a limit of 5 to 10 minutes for this step. It also mentions using one colour for direct quotes and another for observations. If the notes come from interviews, add the interview number to each one. You'll want to know later whether three notes came from three people or from one.

Nine notes, one board

Everything in this section is invented: the whiteboard start-up, its five interviews, the notes, the votes and the decisions. The start-up isn't Prompt Board, and nobody ran these interviews.

Three people are building a small online whiteboard. They interview five new users about their first week and come away with nine notes, posted in no order.

The first sort

All three move notes at the same time, without discussing them, and stop after five minutes. Four piles have formed.

Sorting in silence is our template's instruction, and it's a preference. NN/g says the team organises the notes together and doesn't ask for silence. We use it so that the first person to speak doesn't set the groups.

The note that moved

"Worried about guest costs" travelled between piles twice. One person read it as a money worry and put it beside "Asked where data lives", because both sounded like doubts about the company. Another moved it next to the share button.

They settled it by asking what the interviewee was trying to do at the time. She was trying to invite a client, so the note went to Inviting the team. That's a useful rule for any note that keeps moving: sort by what the person was trying to do, not by the word on the note.

Four names, rewritten

NN/g's advice is to "give each cluster a clear label representing all included sticky notes", and to start grouping before you decide on a label. The four names on the board are fine for finding your way around. They don't yet say anything, so the team rewrites each as a sentence.

A sentence can be wrong, and that's why it helps. Someone can now look at a group and say "those two notes don't support that", which nobody could say about the word "Trust".

Rank, then one next step each

Each person gets three dots, so nine in all. Inviting the team takes four, Getting started three, Trust two and Delight none.

Inviting the team has fewer notes than Getting started and still comes first, because a board nobody else can open doesn't get far in a team. Note count isn't importance, and NN/g warns against the same mistake: "Don't discount a small cluster just because it has few sticky notes."

NN/g suggests naming a point person for the action items, so each row gets one name. Nine notes from five interviews can't tell this team what all its users think. They tell it what to check next, which is why two of the four rows include another test or another round of questions.

An hour, in order

NN/g gives no fixed group size. It says you can make a map alone, that it's best made as a team, and that a large workshop should "break up participants into smaller groups". The timings below are our suggestion for three to six people and up to a few dozen notes.

1. Ten minutes to write. Everyone writes findings alone, one per note.
2. Ten minutes to sort. Everyone moves notes without talking, and notes can move more than once.
3. Fifteen minutes on the awkward notes. Talk through any note that keeps moving and any pile that has grown too large to name.
4. Ten minutes to name. Write a sentence for each group, starting with the groups people disagree on.
5. Fifteen minutes to rank and assign. Vote with dots, then write one next step and one name per group you'll act on.

Which board answers which question

Card sorting sounds similar and is a different job. NN/g describes it as "a specific research method used to determine the information architecture of a site or application", where affinity mapping groups broad ideas.

Where the method came from

We can't fully settle this one. Raymond Scupin's 1997 essay in Human Organization, going by its abstract, discusses the KJ method "developed by a Japanese ethnologist, Jiro Kawakita", and says it grew out of difficulties interpreting ethnographic data in Nepal. We read the abstract and not the full essay.

Lucid Meetings' glossary entry on the KJ technique lists silent note-writing, grouping, naming and voting, which is close to the steps above. Neither page uses the word "affinity", and NN/g's article gives no origin. The claim that affinity mapping descends from the KJ method is widely repeated, and we only found it stated outright on weaker sources, so treat it as likely and unproven.

Frequently asked questions

Is an affinity map the same as an affinity diagram?

Yes. NN/g lists affinity mapping, collaborative sorting and snowballing as other names for affinity diagramming.

What do you do after affinity mapping?

Rank the groups, write next steps for the ones you'll act on, and give someone responsibility for them. NN/g suggests dot voting if the team can't agree, and saving the finished diagram for future reference.

How many notes do you need?

NN/g sets no minimum or maximum. Our example has nine so that it fits on a page, and a real round of interviews can produce far more. With a very large pile, NN/g suggests taking a first pass through the data before the workshop.

Can you do affinity mapping alone?

Yes, NN/g says it's possible, and that it works best as a team because the map then prompts discussion and helps with decisions. If you sort alone, ask someone else to read your group names and challenge them.

Does the sorting have to be silent?

No. It's our template's instruction, and NN/g's steps don't require it. Try it if one person tends to set the direction in your meetings.

Sort your own notes

Paste your interview notes or other findings in, one per note, and give everyone ten minutes to sort. Open the affinity map template, clear the sample notes and share the board link with the team. When the groups have names, carry the top one into a how might we question or an empathy map.