The first 3 interviews are free · Start free or see the prices

Affinity Mapping Template: How to Do an Affinity Diagram

Use this template to sort interview notes into clusters from the bottom up. The rules say what goes on a note, when notes belong together and how to write a useful label.

Analyze a transcript free

Use it for

  • Notes, quotes or clips from 5 to 15 interviews, with the participant ID on each one.
  • A whiteboard, a wall or a digital board with space for 100 to 400 notes.
  • 2 to 5 people who took part in the research or read the transcripts.

Steps

  1. Write one observation per note

    Each note holds one fact, quote or behavior from one participant. Add the participant ID. Write in short sentences that a reader can understand without the transcript.

  2. Sort without categories

    Put notes up one at a time. Move a note next to another if the two feel related. Do not name clusters yet. Names too early force notes into your existing ideas.

  3. Sort in silence first

    For the first 20 minutes, the team sorts without talking. Anyone can move any note. If a note moves back and forth, copy it and put it in both places.

  4. Label each cluster

    When a cluster has 3 or more notes, write a label for it. Write the label as a statement in the participant's voice, such as "I do not trust numbers I did not calculate".

  5. Group the clusters

    Put related clusters under a higher-level group with its own label. Most maps end with 4 to 8 groups. Then write one insight for each group.

  6. Check the map

    Count the participants in each cluster. A cluster with notes from only one person is not a pattern. Look at the outliers and note the gaps, then decide what to research next.

Codebook

Copy the codebook as a table. It pastes into a spreadsheet with one row for each code.

Observation note

A note with one fact, quote or behavior from one participant.

Include
The note holds what a participant said or did, with their ID, in words close to theirs.
Exclude
The note holds your interpretation or a design idea. Put interpretations in labels and ideas in a separate area.
Example
“P4: Prints the weekly report and highlights the numbers that changed by hand.”

Cluster

A set of 3 or more notes that share one idea.

Include
The notes describe the same need, behavior or problem, from at least two participants.
Exclude
The notes share only a topic word, such as "pricing", but describe different ideas. Split them.
Example
“P4 prints the report. P7 copies numbers into a notebook. P9 asks a colleague to check each total.”

Cluster label

A short statement that says what the notes in a cluster have in common.

Include
The label is a full sentence in the participant's voice that a reader can agree or disagree with.
Exclude
The label is a single topic word, such as "Reports". A topic does not say anything. Rewrite it as a statement.
Example
“I check every number myself before I share it.”

Group

A higher-level set of clusters that share a broader idea.

Include
Two or more cluster labels point to the same broader need or problem.
Exclude
The clusters only sit next to each other on the board. Put them in a group only if their labels support one statement.
Example
“Group: I do not trust the data until I have touched it. [clusters on checking numbers and on manual edits]”

Insight

A statement that explains why a group of behaviors happens and what it means for the product.

Include
The statement links the behavior to a cause and points to a decision, written in your own voice.
Exclude
The statement only repeats the group label. An insight must add the why or the so what.
Example
“Analysts print and hand-check reports because the tool hides how numbers are calculated. Showing the formula could cut the checks.”

Outlier

A note that does not fit any cluster but could matter.

Include
The note describes a behavior that is different from the rest, or a strong reaction from one participant.
Exclude
The note fits a cluster but the sorter did not see it. Move it to that cluster.
Example
“P11: Has never opened the report. Gets the numbers by phone from the analyst.”

Gap

A question that the map raises but that the notes cannot answer.

Include
The team needs information that no participant gave, such as how a group outside the sample behaves.
Exclude
The answer is in the transcripts but no one wrote a note for it. Go back and add the notes.
Example
“Gap: Do managers who receive the printed reports check them again?”

Report outline

  1. Who took part, how many interviews the notes came from and when the session took place.
  2. A photo or export of the final map.
  3. The groups, each with its label, its clusters and the number of participants behind each cluster.
  4. One insight per group, with 2 or 3 example notes.
  5. Outliers and gaps, and the research that could answer each gap.
  6. Decisions and next steps from the session.

Mistakes to avoid

  • Do not start with category headings such as "Pain points" and "Features". Top-down headings make the map confirm what you already thought.
  • Do not write labels as single words. "Onboarding" says nothing. A statement says what you learned.
  • Do not let one person sort all the notes. The value of the method comes from different people seeing different links.
  • Do not count notes as evidence. Count participants. One talkative person can fill a cluster alone.

Guides that use it

Questions

Affinity mapping is a method to sort qualitative notes into groups by similarity. The groups come from the data, not from categories you set before.

Affinity mapping is a fast, visual group activity. Thematic analysis uses a written codebook and is slower but easier to repeat and check. Many teams use affinity mapping first, then a codebook.

Most maps from 5 to 15 interviews have 100 to 400 notes. More notes take longer to sort. If you have more, pick the notes that answer the research question.

The Board tab shows a study's clips as cards in lanes. You can add, rename and merge lanes and move cards between them. Export CSV gives one row per clip to import as sticky notes.

Ask real people your question

The AI moderator interviews screened people by voice or text. Each claim in the study report links to its clip. The first 3 interviews are free.

Start free