Learn how to organize messy qualitative research into meaningful clusters using affinity diagrams. Step-by-step instructions with real examples.
You have just finished interviewing twelve users. You have pages of notes, audio recordings, screenshots of their workflows, and a growing sense that there are patterns hiding in the noise. But when someone asks you "so what did you learn?" you struggle to give a clear answer. That is exactly the moment when affinity diagrams earn their keep.
An affinity diagram is a bottom-up method for organizing qualitative data. You take individual observations, write each one on a separate note, then group them by natural similarity. The clusters that emerge become your themes, and those themes become the foundation for everything that follows in your design process.
Affinity diagrams are most useful at the transition between the Empathize stage and the Define stage. You have collected raw data and need to make sense of it. But they also work well after brainstorming sessions (to sort ideas), after usability tests (to categorize findings), or any time you have more than 20 individual data points that need structure.
They do not work well for quantitative data, for data sets smaller than about 15 items (just use a list), or when the categories are already known (in that case, just sort into the existing categories).
If you are working in person, you need sticky notes, markers, and a large wall or whiteboard. Each person on the team should have their own color of sticky notes so you can see whose observations are whose. If you are remote, use a digital whiteboard tool. The specific tool matters less than having enough space to spread things out.
One critical rule: one observation per note. Not a summary, not a conclusion, not a feeling. An observation. "User #4 checked her email three times during the checkout flow" is an observation. "Users are distracted during checkout" is a conclusion. Put the observation on the note and save the conclusions for later.
Each team member reviews their research data and writes observations on individual notes. Aim for 30 to 60 notes per person. Do not edit yourself. If you noticed it, write it down. Observations that seem trivial often become important when you see five other people noticed the same thing.
One person reads their note aloud and places it on the wall. The next person either places their note near a similar one or starts a new area. Keep going until all notes are placed. Do not discuss or debate during this phase. The goal is placement, not agreement. If two notes seem related, put them near each other. If you are not sure, leave space between them.
Everyone moves notes around silently. No talking. This forces you to think about relationships without being influenced by whoever speaks loudest. If someone moves a note you placed, let it happen. You can move it back if you genuinely disagree, but resist the urge to defend your placement.
Now you talk. Look at the groups that have formed and give each one a name. The name should describe what the notes in the group have in common. "Trust issues with payment" is a good cluster name. "Payments" is too vague. "Users do not trust our payment form because it looks different from what they are used to on Amazon" is too specific for a cluster name (though it might be a great observation within the cluster).
Some clusters will be related. "Trust issues with payment" and "Abandonment at checkout" probably connect. Draw lines between related clusters. Note which clusters have the most notes (volume signals importance) and which have the most emotional notes (intensity signals opportunity).
A good affinity diagram typically produces 5 to 10 clusters from 100 to 200 notes. If you have more than 15 clusters, some of them are probably too granular and should be combined. If you have fewer than 4, you probably grouped too aggressively.
Each cluster should have at least 3 notes. A "cluster" of 1 or 2 notes is really just an outlier. That does not mean it is unimportant. Outliers can be the most interesting findings. But they are not themes.
Watch out for "junk drawer" clusters. If you have a group called "Other" or "Miscellaneous" with 20 notes in it, that is a sign you need to spend more time sorting. There are probably two or three real themes hiding in that pile.
The clusters themselves are not insights. They are organized data. The insight comes when you ask "what does this cluster tell us about our users' needs?" For each cluster, write one sentence that captures the implication for your design work.
For example, a cluster labeled "Workarounds for missing features" might yield the insight: "Users are solving their own problems with duct-tape solutions, which means there is demand for functionality we have not built yet, and users are resourceful enough to adopt it if we build it right."
These insights feed directly into How Might We questions and problem statements. The affinity diagram is the bridge between raw empathy data and structured problem definition.
Running this exercise remotely requires a few adjustments. Give people more time for individual note generation (the async nature of remote work means people need more ramp-up time). Use a timer and keep the video call running even during silent sorting so people stay focused. Break the session into two parts if attention spans are short: notes and placement in one session, sorting and naming in another.
The biggest risk with remote affinity diagrams is that digital tools make it too easy to create perfectly organized grids. Resist the urge to make it neat. The messy overlap between clusters is where the interesting insights live.
Affinity diagrams sit at the heart of the Define stage. The clusters you create become the themes of your empathy maps. They feed into persona development. They shape the problem statements that guide your Ideate stage.
When done well, an affinity diagram gives your team a shared understanding of what the research revealed. Instead of twelve people having twelve different interpretations of twelve interviews, you have one coherent picture that everyone helped build.
Related guides: brainstorming techniques · crazy eights sketching · storyboarding techniques
Design Thinker Labs Home · All Guides · How It Works · Pricing