Affinity Mapping in Qualitative Research: Step-by-Step

On this page

• What Is Affinity Mapping?
• When to Use Affinity Mapping vs Thematic Analysis
• What You Need Before You Start
• Step-by-Step: How to Run an Affinity Mapping Session
• Turning Your Affinity Map into Actionable Insights
• Common Mistakes and How to Avoid Them
• Frequently Asked Questions

What Is Affinity Mapping?

Affinity mapping is a bottom-up synthesis technique for organising raw qualitative observations into emergent themes. Rather than applying a pre-set framework to your data, you let categories form from the material itself — clusters arise because notes share something, not because a codebook told you where to put them.

This distinction matters. In top-down coding, you define a schema first and sort data into it. In affinity mapping, the labels come last. The approach originates in the KJ method, developed by Japanese anthropologist Kawakita Jiro in the 1960s, and it has since become a standard tool in UX and product research — valued precisely because it surfaces unexpected patterns rather than confirming existing ones.

The two most common inputs are notes from user interviews and observations from usability testing sessions. Both produce the kind of granular, behavioural data that affinity mapping handles well: discrete moments, verbatim quotes, and specific actions that resist neat categorisation before you look across them as a whole.

---

When to Use Affinity Mapping vs Thematic Analysis

Both methods produce themes from qualitative data. The practical difference is in who is in the room, how much time you have, and how much rigour the output needs.

Affinity mapping suits workshop settings where cross-functional stakeholders — designers, product managers, engineers — need to be part of the synthesis. The process is fast, visual, and tactile. Emergent grouping favours discovery: you find things you were not looking for. It works best with 3–15 participants’ worth of notes and a timeline measured in hours rather than days.

Thematic analysis is better for solo analysts working with larger datasets, or for outputs that will face academic or regulatory scrutiny. It involves an iterative codebook, explicit decisions about intercoder reliability, and a more structured audit trail. It takes longer and is harder to run collaboratively in real time.

There is genuine overlap: both are inductive by default, both produce named themes, and both require careful attention to how labels are worded. The practical decision rule is straightforward — if you need stakeholders in the room and synthesis needs to happen this week, affinity mapping is the right choice. If you are working alone on a large dataset and need a defensible codebook, thematic analysis is the better fit. For a fuller discussion of how these approaches fit into broader synthesis & analysis workflows, see our linked hub.

---

What You Need Before You Start

Good affinity mapping depends on preparation. Arriving at a session with poorly formed notes wastes everyone’s time and produces unreliable clusters.

Atomic notes. Each sticky note — physical or digital — should contain one observation only. “User said she always checks the price before clicking through” is atomic. “User seemed confused and kept going back” is not: it blends observation with interpretation and contains two separate behaviours.

Volume. Aim for 5–10 notes per participant. With five participants, expect 25–50 notes minimum before the session. Fewer than 20 notes and clustering becomes trivial; more than 80 with a single group risks cognitive overload — consider splitting into sub-groups.

Tooling. Choose based on team location:

Tool	Best for	Watch out for
Physical wall + Post-its	Co-located teams, tactile engagement	Hard to archive; no search later
Miro	Remote teams, persistent boards	Can feel abstract; harder to enforce silence
FigJam	Design-led teams already in Figma	Limited export options
Mural	Facilitated enterprise workshops	Learning curve for infrequent users

Team composition. A researcher facilitator plus one to three cross-functional observers is ideal. More than five people and the sorting phase becomes chaotic; fewer than two and you lose the challenge-and-debate step that improves cluster quality.

Pre-work. De-identify notes before the session. Strip interpretation. Where you have a verbatim quote, use it — the participant’s own words make for stronger cluster labels later. For guidance on capturing this quality of note during fieldwork, see how to conduct user interviews.

---

Step-by-Step: How to Run an Affinity Mapping Session

A well-run session with 40–60 notes and three to four participants takes 90–120 minutes. Here is the sequence.

Step 1 — Dump (10–15 minutes)

Everyone places their notes on the board simultaneously, in silence. No explanation, no commentary. The aim is to externalise everything before anyone starts making sense of it. Talking at this stage anchors people to their own interpretation before they have seen the full picture.

Step 2 — Sort (20–30 minutes)

Move notes into proximity clusters, still in silence and still simultaneously. If you think a note belongs near another, place it there. If someone else moves a note you placed, let it go — disagreements are surfaced in Step 5, not now. The silence prevents one confident voice from shaping everyone else’s groupings prematurely.

If you are working with notes from guerrilla usability testing findings, you may notice clusters forming around specific interface moments — that is expected and useful. Do not force them to merge with attitudinal clusters from interviews.

Step 3 — Name (10–15 minutes)

Draft a header label for each cluster. Use the language of the data. If several notes describe users checking a price before committing, a label like “Price verification before action” is grounded. “Trust issues” is an interpretation — avoid it at this stage. Good labels describe what the cluster contains, not what you think it means.

Step 4 — Group clusters into second-order themes (15–20 minutes)

Look for relationships between clusters. Two or three related clusters can be gathered under a higher-level theme. Aim for two to three levels of hierarchy at most: individual notes → clusters → themes. Flatter than this and you lose structure; deeper and you create a taxonomy that nobody will use.

Step 5 — Challenge and stabilise (20–30 minutes)

Open the floor. Question placements. Ask: “Why is this note here and not there?” Merge clusters that are genuinely the same thing. Split clusters that contain two distinct ideas. Agree on final labels as a group. This is the most analytically valuable part of the session — do not timebox it too aggressively, but do timebox it.

Step 6 — Capture and assign

Photograph the physical board or export the digital board. Assign one person to own the write-up. The map degrades in meaning quickly once people leave the room, so the owner should start the synthesis document the same day.

Facilitator tips:

• Post a visible timer for Steps 1 and 2 to keep momentum.
• If someone starts talking during the silent sort, quietly redirect — “Save it for Step 5.”
• Resist the urge to pre-sort notes by interview participant. Mixed provenance is the point.

---

Turning Your Affinity Map into Actionable Insights

The map itself is not the deliverable. A photograph of sticky notes on a wall communicates nothing to a stakeholder who was not in the room. Your job after the session is to translate clusters into insight statements.

A useful format: “Users struggle with X because Y.” This forces you to name both the observed behaviour and the underlying cause. A cluster full of notes about users re-reading confirmation screens becomes: “Users re-read confirmation screens because the transaction summary does not show the final amount until the last step.” That is actionable. A cluster label that reads “Confusion around costs” is not.

Frequency vs severity. A cluster with many notes is not automatically the most important one. A cluster with five notes describing a complete task failure may matter more than a cluster with twenty notes describing mild friction. Assess each cluster on both dimensions before prioritising.

Prioritisation bridge. For product teams, link clusters directly to opportunity areas or jobs-to-be-done. A cluster around price verification, for instance, maps to a specific user job — and naming that job makes it easier to evaluate potential solutions.

Outputs to produce:

• Insight repository cards (one per cluster, including supporting quotes)
• Opportunity statements for product backlog input
• A findings section for the broader research report

Avoid the wall-of-stickies trap: teams that export a board and share it as the final artefact are sharing process, not insight. For guidance on bridging research findings into product decisions, see our Insight to Impact hub.

---

Common Mistakes and How to Avoid Them

Writing interpretation instead of observation. “Users are confused” belongs in your analysis, not on a note. “User clicked the back button three times before completing the step” is observable and placeable. Keep notes descriptive.

Letting one voice dominate sorting. If the most senior person in the room starts moving notes and explaining their reasoning before the silent sort is done, everyone else follows. Enforce the silent first pass without exception.

Over-clustering or under-clustering. Too few broad themes and you lose the specific signals that should drive design decisions. Too many micro-clusters and the map becomes a flat list with no hierarchy. If you finish with fewer than four clusters or more than fifteen, revisit the groupings.

Skipping second-order grouping. Many teams stop at the first level of clusters and treat that as the output. Without the second-order step, you have a list of topics, not a structured understanding of the problem space.

Never revisiting the map. Affinity maps made in workshops and left unwritten for a week lose most of their meaning. The person assigned to write up findings should do so within 24 hours while the session is still fresh.

---

Frequently Asked Questions

How many participants do you need before affinity mapping is useful?

Affinity mapping adds value from as few as three to five participants. With fewer sessions you may have too few notes — under 15 — for meaningful clustering to emerge, since coincidental similarities can look like patterns. With more than 15 to 20 participants, the volume of notes risks overloading a single session; in that case, consider splitting your data into sub-groups and running separate passes before combining themes.

Can affinity mapping be done asynchronously?

Yes, particularly in digital tools such as Miro or FigJam. The most reliable approach is a hybrid: each team member adds their notes independently before the session, then the group meets synchronously to handle sorting, naming, and the challenge step. Fully asynchronous clustering is possible but tends to produce weaker results — the debate in Step 5 is where the most important analytical decisions are made, and it is difficult to replicate that quality in comments or annotations.

What is the difference between an affinity map and an affinity diagram?

The terms are used interchangeably in UX and product research. “Affinity diagram” is the original label from the KJ method; “affinity map” is the more common contemporary term in design and product contexts. The process, the inputs, and the output are identical — the difference is purely one of vocabulary across communities of practice.

How do I know when my affinity map is finished?

The map is stable when three conditions are met: moving any remaining ungrouped notes does not change the meaning of existing clusters; every cluster has a clear, single-sentence label that the whole team agrees on; and the hierarchy reflects what the data shows rather than assumptions the team brought into the room. If the team is still debating whether a note belongs in one cluster or another, the label for at least one of those clusters probably needs refining.