Hundreds of raw observations — quotes, behaviours, anomalies — clustered bottom-up into themes the team can act on. The bridge from a pile of research to a pattern.
▸ Try the interactive toolAffinity mapping is a collaborative technique for organising raw research observations into themes and patterns. It takes the hundreds of individual data points a discovery effort produces — interview quotes, usability observations, analytics anomalies, support verbatims — and transforms them into a structured map the team can actually reason from.
The defining discipline is working bottom-up: you start with individual observations and let the themes emerge from clustering them, rather than starting with categories and sorting observations into them. Top-down sorting just confirms the team's existing mental model; bottom-up clustering surfaces patterns nobody expected. Done as a team, it also builds shared understanding — everyone sees the same evidence resolve into the same themes.
Write each observation on its own note → cluster notes that feel related, without pre-set categories → let themes emerge from the clusters → name the themes last. Patterns are discovered, not imposed.
A discovery effort produces a flood of raw data that's useless until it's synthesised — and the synthesis method determines whether you find real patterns or just confirm your assumptions.
| The shortcut | What it costs | What it gives you instead |
|---|---|---|
| Drowning in raw data | Hundreds of observations with no structure are unusable. | Affinity mapping clusters them into a handful of actionable themes. |
| Top-down sorting | Pre-set categories just confirm what you already believed. | Bottom-up clustering lets unexpected patterns emerge. |
| Cherry-picking quotes | Without synthesis, teams pick the data points that fit their view. | Mapping all observations counters selective evidence. |
| Solo interpretation | One person's synthesis carries their bias. | A team affinity session builds shared, debated understanding. |
Break the raw research into individual data points — one quote, behaviour, or anomaly per note. Granularity matters; a note that bundles three ideas can't be clustered cleanly.
Group notes that feel related, working bottom-up. Resist the urge to start with named buckets — the whole point is to let the grouping reveal structure you didn't expect.
Once clusters form, name them — last, not first. The name describes what actually grouped, rather than a category you imposed before looking.
Cluster collaboratively so the interpretation is debated and shared. The disagreements about where a note belongs are often where the real insight surfaces.
Each robust theme becomes the basis for a structured insight (Tool 21) — the bridge from “we noticed a pattern” to “here's what it means and what to do.”
A team finished a round of research and started synthesising the obvious way — sorting observations into the categories they'd expected to find. The result neatly confirmed what they already believed, because the categories had been chosen before looking.
Re-running it as a bottom-up affinity map — clustering individual notes with no pre-set buckets — a theme emerged that none of their original categories would have captured: a recurring frustration that cut across several supposedly-separate problems. It had been invisible under top-down sorting because no one had a box for it. The bottom-up method found the pattern precisely because it didn't start from the team's assumptions.
The deliverable is the clustered map — raw observations grouped into named, emergent themes — ready to feed insight statements.
| Top-down (avoid) | Bottom-up (use) |
|---|---|
| Start with categories | Start with observations |
| Sort data into boxes | Let clusters form freely |
| Name themes first | Name themes last |
| Confirms assumptions | Surfaces the unexpected |
Raw research is inert until it's synthesised, and the direction of synthesis decides the outcome: top-down confirms what you knew, bottom-up reveals what you didn't. Affinity mapping is bottom-up by design.
The reason the bottom-up rule matters so much is that top-down sorting is a confirmation-bias machine wearing the costume of analysis. When you start with categories, every observation gets filed under something you already believed, and the synthesis feels rigorous while teaching you nothing new. Letting themes emerge from the clustering — naming them only after they've formed — is what allows the surprising pattern, the cross-cutting frustration, the thing nobody had a box for, to become visible. And doing it as a team turns synthesis from one person's interpretation into a shared, debated understanding the whole group will actually act on.
Top-down sorting just confirms your assumptions. Cluster bottom-up and name themes last.
One note per observation; bundled notes can't cluster cleanly.
Solo interpretation carries solo bias. Map as a team.
Some observations are outliers. Don't distort clusters to make everything fit.
Interviews, usability tests, analytics, and diary studies all produce the observations affinity mapping clusters.
Emergent themes become structured insights (Tool 21) — the next link in the synthesis chain.
Themes map to opportunities on the OST (Tool 03).
Bottom-up clustering counters the confirmation bias catalogued in Tool 25.
Affinity mapping — clustering many observations into themes — is exactly the kind of synthesis AI can accelerate, carefully.
The judgment that stays yours: The point of affinity mapping is the thinking that happens while you cluster — the felt sense of a pattern forming. Letting AI hand you finished clusters skips the insight. Use it to seed and challenge your map, then do the sense-making yourself.
Take 10–15 real observations from any source (reviews, support tickets, your own notes about a product). Write each on a separate note. Cluster them without deciding categories first.
Name the clusters only after they've formed. Notice whether any theme emerged that you wouldn't have predicted.
If a pattern appeared that you'd never have created a category for in advance, you've felt the difference between bottom-up discovery and top-down confirmation.