Affinity mapping: how to turn raw research into insights that stick
Affinity Mapping
✓ themes that survive being broken and rebuilt
On this page
- TL;DR
- What is affinity mapping?
- When to use affinity mapping
- How to run an affinity mapping session
- Affinity mapping vs. other analysis methods
- Digital, physical, and AI-assisted affinity mapping
- Running affinity mapping remotely
- What to capture after an affinity mapping session
- Connecting affinity mapping to your research repository
- Common affinity mapping mistakes
- Turn patterns into decisions
You have just finished 15 user interviews. You have pages of notes, a whiteboard full of sticky notes, and a vague sense that the real insight is somewhere in there. Then you spend three hours moving colored squares around, and two weeks later nobody can find the results.
That is the gap affinity mapping often falls into. Not because the method does not work, but because teams can treat it as a workshop exercise instead of part of a larger research and synthesis process.
TL;DR
Affinity mapping, also called affinity diagramming or the KJ method, is a bottom-up way to organize research observations into related groups and identify themes. It works especially well after interviews and usability studies, when you have a large amount of qualitative data to make sense of. The most useful affinity maps involve the team in analysis, stay grounded in the original evidence, and turn the resulting themes into findings that live somewhere people can find and reuse them.
What is affinity mapping?
Affinity mapping is a qualitative analysis technique that organizes large volumes of observations, quotes, and findings into clusters based on natural relationships.
The method grew out of the KJ method developed by Japanese anthropologist Jiro Kawakita in the 1960s. You may also hear it called an affinity diagram, particularly in service design and quality management. In UX research, affinity mapping is the more common term.
The basic idea is simple: write one observation per note, then group related notes together. As those groups take shape, patterns begin to emerge that can be difficult to spot when you are reading interview transcripts or research notes one at a time.
What makes affinity mapping particularly useful is that it is bottom-up. You do not begin with a predetermined set of categories. You let the categories emerge from the evidence.
That distinction matters. If you decide what all your themes are before looking closely at the research, you are more likely to find what you expected to find. Affinity mapping gives unexpected patterns room to surface.
When to use affinity mapping
Affinity mapping works best when you have enough qualitative material that patterns are difficult to see by simply reading through it. It is a good fit:
- After interviews or usability sessions, when you have many observations and need to find patterns.
- When multiple people collect the data and need to develop a shared understanding of what they learned.
- When your team disagrees about what the research is telling you.
- Early in a problem space where the important categories are not obvious yet.
When it may not be the right method
- When you already have established categories and primarily need to count how often they appear. Content analysis may be a better fit.
- When you only have a small amount of qualitative data that can be synthesized more directly.
- When nobody involved in the exercise has engaged with the raw research. Affinity mapping should help you analyze evidence, not replace reading or watching it.
How to run an affinity mapping session
1. Prepare the observations
Start with one observation per note. This simple rule makes the rest of the process much easier. A useful note captures one specific thing a participant said or did.
- Good: “Checked the export twice because she did not trust that the first one had worked.”
- Too vague: “Export issues.”
- Too broad: “P4 had several problems with export, dashboard, and permissions.”
The first can be compared meaningfully with other observations. The second has already compressed the evidence into a vague conclusion. The third contains several observations that should be separated.
Include the participant ID or another source reference on every note. You will need it later when you want to understand how many different people contributed to a theme rather than simply counting how many sticky notes happen to be in the cluster.
For a typical workshop, you might work with dozens to a couple hundred observations. If you have hundreds more, consider doing an initial digital pass or dividing the analysis into manageable sections before bringing the team together.
2. Bring the right people
Affinity mapping can be particularly valuable as a team exercise. Researchers may notice methodological patterns, while product managers, designers, and engineers bring different knowledge of the product and customer experience.
That diversity also creates a useful side effect: people develop a shared understanding of the research while analyzing it instead of simply receiving the findings afterward.
A small working group is usually easier to run than a large room. You want enough people to bring different perspectives without turning the session into crowd management.
Set expectations before you begin. This is an analysis session, not a presentation. People should expect to spend their time reading, moving, comparing, and discussing evidence.
3. Cluster in silence first
Start by having everyone move notes at the same time without discussing them. Place related observations near each other, and move notes someone else has placed if you think they belong somewhere else.
Silent clustering reduces the chance that the most senior or vocal person defines the categories before everyone else has had a chance to interpret the evidence.
Continue until movement begins to slow and natural groups start to form.
4. Discuss and refine
Now start talking.
Pay particular attention to notes that people moved repeatedly, observations that could belong in more than one place, and notes that do not seem to fit anywhere.
Split clusters that have become too broad. Merge groups that appear to describe the same underlying experience.
This conversation is part of the synthesis. Debating whether an observation belongs under “onboarding friction” or “unclear expectations,” for example, may reveal that the customer experience sits somewhere between the two.
5. Break it and rebuild it
This is the step many teams skip, and it is a useful way to pressure-test the patterns you found.
Take one or two of your largest or most important clusters, break them apart, and try grouping the observations again without relying on the original structure.
Do similar groups reappear? If they do, you have more reason to trust that the pattern reflects something meaningful in the data. If the notes reorganize completely, treat the original cluster more cautiously. It may be one reasonable interpretation rather than a stable finding.
The goal is not to prove that there is only one correct way to organize the data. It is to challenge the first convenient story before turning it into a research finding.
6. Name the clusters last
Wait until your groups are relatively stable before naming them. Then make the name describe what is actually happening rather than simply naming a topic.
- Weak: “Trust”
- Stronger: “People redo work manually because they do not trust that the system completed it.”
Likewise, “Users cannot find their history” is more useful than “Navigation.” The goal is to move from topics to findings. Those findings are what eventually make their way into product discussions, reports, tickets, and your research repository.
7. Count the evidence
For each theme, count distinct participants, not sticky notes.
Five observations from one particularly talkative participant still represent one person's experience. A theme that appeared across nine of 12 participants carries different weight from one that appeared in two.
That does not mean frequency automatically determines importance. A less common issue can still be critical. But keeping the participant count attached to the theme gives everyone useful context when interpreting it.
8. Interpret what you found
Clustering is not the end of analysis. Ask:
- Which themes are actionable?
- Which themes appear related?
- Do any findings contradict one another?
- Could those contradictions point to different customer segments or use cases?
- What did you expect to find that never appeared?
- What new questions did the analysis create?
That last question matters. Affinity mapping makes visible what is in the data. You still have to deliberately ask what might be missing.
Affinity mapping vs. other analysis methods
Affinity mapping is a bottom-up method. It works well for discovery because you allow patterns to emerge from the research instead of imposing categories in advance.
Content analysis is more structured. You generally begin with categories or codes and apply them consistently across the data, which makes it useful for tracking known themes and comparing them over time.
Framework analysis also uses a more structured analytical framework and can work particularly well when several researchers need to analyze qualitative data consistently.
Mind mapping vs. affinity diagrams
Mind mapping and affinity diagramming are easy to confuse because both create visual groups of related ideas. But they work in opposite directions.
A mind map starts with a central idea and branches outward into related concepts. You already know what sits at the center, so the exercise helps you explore or organize your thinking around it.
An affinity diagram starts with the evidence. There is no predetermined center. You group individual observations and discover the larger structure through the clustering process.
Put simply: mind mapping organizes ideas; affinity mapping organizes evidence to uncover patterns.
Card sorting is different again. In card sorting, research participants organize information, which helps you understand their mental models. In affinity mapping, your team organizes research data that has already been collected.
Digital, physical, and AI-assisted affinity mapping
Physical sticky notes make affinity mapping tactile and collaborative. People can quickly scan the room, move observations, and see groups taking shape.
The downside is what happens afterward. A wall of sticky notes is difficult to search, share with a distributed team, or connect to future research.
Digital whiteboards solve much of that problem and make remote affinity mapping possible. But they introduce another: the finished board can easily become one more link nobody opens again.
Whichever format you use, plan where the findings will live after the workshop. The board is the workspace. It should not be the final home for the research.
AI adds another option, particularly when you are working with hundreds of observations. It can generate an initial clustering or suggest possible themes much faster than a person sorting every note manually.
Use that output as a starting point to interrogate, not a finished analysis. Compare AI-generated groups against the raw observations, look closely at outliers, and ask whether alternative groupings tell a different story. Human judgment is still what turns clusters into defensible research findings.
Running affinity mapping remotely
Remote affinity mapping follows the same basic process, but a little preparation makes it much smoother.
Set up the research observations in your digital workspace before the session rather than asking people to transcribe notes while everyone waits.
Begin with silent clustering just as you would in person, then bring the group together to discuss the patterns that emerged. Use the conversation for the difficult interpretive work rather than spending the session on setup.
Timebox the discussion so the group does not spend 30 minutes trying to find the perfect home for one ambiguous observation. The goal is not unanimous agreement on every sticky note. It is a useful representation of the dominant patterns in the research.
Before anyone closes the board, document the cluster names and a short description of each finding somewhere more permanent.
What to capture after an affinity mapping session
The board itself is not the final deliverable. Capture enough context that someone who was not in the workshop can understand and trust what came out of it:
- Session context: the research question and data sources you analyzed.
- Themes: each theme's name, a short explanation, the number of distinct participants represented, and supporting evidence.
- Key findings: the three to five things a stakeholder actually needs to know.
- Connections and tensions: themes that reinforce or contradict one another.
- Open questions: what the analysis surfaced that you still need to understand.
- Source evidence: links back to relevant transcripts, recordings, quotes, and the original affinity map.
Aim for something people can understand quickly and return to later rather than documenting the workshop for its own sake.
Connecting affinity mapping to your research repository
The affinity map is where synthesis happens. The repository is where that synthesis becomes reusable.
If the findings stay on a whiteboard, their useful life is usually short. Months later, a product manager can be working on exactly the same problem without knowing that research already exists.
Move the findings into your repository while the context is still fresh. Each insight should connect back to the underlying evidence and include enough metadata to surface when someone searches for that product area or customer problem later.
AI-powered research repositories can make that easier. Instead of depending entirely on someone remembering the name of an old study or finding the right whiteboard, teams can search across previous research and surface relevant findings alongside their source evidence.
ServiceNow offers one example of what connected research infrastructure can do at scale. The company consolidated its research stack from 15 tools to seven and cut recruitment time from 118 days to six.
The larger point is not simply to use fewer tools. It is to make sure the work you put into synthesis compounds instead of expiring when everyone closes the whiteboard.
Common affinity mapping mistakes
- Clustering by topic instead of meaning. Putting every export-related observation together may tell you what people talked about, but not what they experienced. Look for patterns in behavior, motivation, friction, or need.
- Naming clusters too early. Once you call a group “trust,” people naturally start sorting observations according to that label. Let the evidence settle first.
- Letting one participant dominate. Count the number of distinct participants represented in a theme, not simply the number of observations.
- Stopping at the wall. A photograph or screenshot of sticky notes is not a research finding. Capture the insight and connect it to its evidence.
- Leaving the analysis to one person. Solo affinity mapping can work, but collaborative analysis gives other team members direct exposure to the evidence and creates shared understanding.
- Working from memory instead of raw research. Keep transcripts, recordings, or detailed notes available. You should be able to return to the source when a sticky note needs context.
Turn patterns into decisions
Affinity mapping is most useful when the patterns you uncover do not disappear into a workshop board. Keep those themes connected to the interviews, feedback, and other research behind them so your team can revisit the evidence and use it in future decisions. A mature ResearchOps practice and the 5 Cs of a research repository both help make that happen.
Frequently asked questions
What is affinity mapping?
Affinity mapping is a bottom-up qualitative analysis method that groups individual research observations based on their relationships until larger themes emerge. It is also called affinity diagramming or the KJ method.
What is the difference between affinity mapping and an affinity diagram?
They refer to the same basic technique. “Affinity diagram” is commonly used in disciplines such as service design and quality management, while “affinity mapping” is common in UX research.
How many notes do you need for affinity mapping?
There is no required number. You need enough observations for meaningful patterns to emerge without creating so much material that the session becomes unmanageable. For many research projects, dozens to a couple hundred observations can be a practical working range. Larger datasets may benefit from an initial digital or AI-assisted pass before collaborative analysis.
Is a mind map the same as an affinity diagram?
No. A mind map starts with a central idea and organizes related thoughts around it. An affinity diagram starts with individual pieces of evidence and groups them until a structure emerges.
How long does an affinity mapping session take?
It depends on the amount of data, the number of people involved, and how much preparation you do beforehand. A focused workshop may take a couple of hours, while a large dataset can require multiple rounds of clustering and synthesis.
How many clusters should you end up with?
There is no ideal number. You want enough clusters to capture meaningful differences in the research without creating so many that you have simply recreated the raw data in smaller piles. If your clusters are extremely broad, look for useful subthemes. If you have dozens of tiny clusters, look for relationships between them.
