TL;DR: Affinity Mapping
- Affinity mapping is a group analysis method that turns scattered qualitative notes into clusters of related themes.
- You write observations on sticky notes (physical or digital), group similar ones together, and name each cluster.
- It works best after user interviews, usability tests, and field research – any time you’re drowning in raw qualitative data.
- Card sorting organizes information architecture; affinity mapping organizes research findings – they’re not the same thing.
- Remote teams can run it in Miro, FigJam, or Lucidspark just as effectively as in a room with a wall full of Post-its.
What Is Affinity Mapping and How It Helps Organize Research Data
You’ve done the interviews. You have pages of notes, a handful of recordings, and a document full of quotes that don’t obviously connect to each other yet. Now what?
This is the exact problem affinity mapping was designed to solve.
Affinity mapping (also called an affinity diagram or KJ method, after its creator Jiro Kawakita) is a collaborative analysis technique where you group large volumes of qualitative data into clusters based on natural relationships. Each piece of data – an observation, a quote, a behavior – goes on its own sticky note. Then, as a team, you sort those notes into groups that feel related. No predetermined categories. The structure emerges from the data itself.
The method was originally developed in the 1960s for organizing complex ethnographic research, but product teams, UX researchers, and consultants picked it up because it works surprisingly well for making sense of messy interview data. The Nielsen Norman Group describes it as one of the most widely used sense-making tools in UX research.
What makes it useful isn’t the sticky notes. It’s the act of physically (or digitally) handling each piece of data individually, resisting the urge to jump to conclusions, and letting patterns surface through grouping rather than assumption.
A typical affinity map starts with individual notes and ends with three or four layers of organization: raw data at the bottom, grouped observations in the middle, and labeled themes at the top. Those themes become your insights – the things you can actually act on.
When to use it:
- After a round of user interviews when you need to find common pain points
- After a usability test session with multiple participants
- During a design sprint to align the team on what users actually said
- Any time you have 50+ observations and no clear picture of what they mean together
One note: affinity mapping is a qualitative method. It’s about finding themes, not counting frequencies. If you need statistical significance, this isn’t the tool. But if you need to understand why something is happening – the motivations, frustrations, and mental models behind user behavior – it’s one of the most direct paths there.
How to Create an Affinity Diagram Step by Step
Here’s the full process behind how to create an affinity diagram, including the steps that tend to go wrong.
Step 1: Gather your raw data
Before you touch a sticky note, you need your source material: interview transcripts, session notes, usability test observations, field research recordings. The more specific the data, the better the output. “Users were confused” is not a data point. “Three participants couldn’t find the export button on the first try and assumed the feature didn’t exist” is.
Go through your notes and pull out individual observations, quotes, and behaviors. Each one gets its own note. Don’t editorialize yet – write down what you saw or heard, not what you think it means.
Aim for 50–200 notes for a meaningful session. Fewer than 50 and the groupings will feel forced. More than 200 and the session becomes unwieldy.
Step 2: Set up your workspace
Physical: A blank wall, whiteboard, or large table. Use different colored sticky notes for different data sources or participants – it helps you spot if patterns are coming from one participant or many.
Digital: Miro, FigJam, or Lucidspark all work well. FigJam has a native sticky note feature that’s fast to use. Miro gives you more flexibility for large datasets. The American Society for Quality recommends digital tools for distributed teams running the process remotely.
Step 3: Silent sorting (the most important step most teams skip)
Put all notes up at once. Then – and this is critical – have each team member silently move notes into groups they feel belong together. No talking. No debating. Just moving.
Why silent? Because the moment someone explains their reasoning, the whole group anchors to that logic. Silent sorting surfaces more diverse groupings and prevents one person’s mental model from dominating the session.
This phase typically takes 20–40 minutes depending on data volume. Notes will get moved multiple times. That’s fine.
Step 4: Discuss and refine
Once movement slows down, open the floor. Walk through contested groupings – notes that kept getting moved back and forth. Discuss what the disagreement is really about. Often it reveals a genuinely ambiguous data point, or two valid but different ways of interpreting an observation.
Don’t force resolution. Some notes belong in two places. Make a copy.
Step 5: Name your clusters
This is harder than it sounds. Each cluster needs a name that captures the insight, not just the topic.
- Weak label: “Navigation”
- Stronger label: “Users rely on search because they don’t trust the navigation menu”
The label should be a sentence, or at least a phrase that could stand alone as a finding. If you can’t write a meaningful label, the cluster probably isn’t cohesive yet.
Step 6: Look for second-order themes
Once your primary clusters are named, step back. Are there clusters that relate to each other? Group them into higher-level themes. This second tier is where strategic insights live – the patterns beneath the patterns.
A cluster about “confusing error messages” and a cluster about “unclear pricing page” might both fall under a higher theme: “Users don’t trust the product to be honest with them.”
Step 7: Document everything
Take photos. Export the board. Write a synthesis document that captures each theme, supporting observations, and implications. The affinity map itself isn’t the deliverable – the documented insights are.
Affinity Mapping vs Card Sorting: What’s the Difference
These two methods get conflated constantly, and it’s worth being precise about the distinction.
Card sorting is a method for understanding how users categorize information. You give participants a set of cards (each representing a piece of content or a feature) and ask them to group them in whatever way makes sense to them. The output tells you how users mentally organize your content – which informs navigation structures, information architecture, and labeling.
Affinity mapping is a researcher-led method for organizing your data. You’re not asking users to do anything. You’re the one sorting – taking observations from research and grouping them to find patterns.
The confusion probably comes from the fact that both methods involve grouping things. But the inputs, the participants, and the outputs are entirely different.
| Affinity Mapping | Card Sorting | |
| Who does the sorting | Research team | Study participants |
| What gets sorted | Research observations and notes | Content items or features |
| Output | Themes and insights from research | User mental models for IA |
| Best used after | Interviews, usability tests, field research | Before designing navigation or content structure |
| Question it answers | What are users experiencing? | How do users expect content to be organized? |
Both are legitimate. They just answer different questions. If you’re trying to figure out what your nav should look like, use card sorting. If you’re trying to make sense of interview data, use affinity mapping.
Some teams use card sorting during early discovery research to understand user mental models, then use affinity mapping later to synthesize everything they’ve learned. That combination works well.
How Affinity Mapping Helps Teams Find User Pain Points
Most qualitative research produces more data than teams can act on. You do five interviews, get 200 observations, and suddenly you’re trying to hold all of it in your head while making product decisions. That’s how important findings get lost.
Affinity mapping forces you to handle each observation individually. When you write “participant mentioned she’d abandoned three checkout flows this month because she didn’t trust the site with her card details” on its own sticky note and then watch it cluster with four other trust-related observations from different participants – that pattern becomes hard to ignore.
Gamestorming’s guide to affinity maps describes this well: the process makes implicit patterns explicit. You’re not discovering new information – you’re making the information you already have legible.
For teams trying to find user pain points specifically, the method is useful because:
It prevents premature synthesis. Without a structured method, researchers often jump from “I heard this in interviews” to “users want X” too quickly. Affinity mapping slows that down. You sit with the raw data longer before drawing conclusions.
It surfaces minority voices. A pain point mentioned by one person might get buried in a meeting. On a sticky note wall, it sits alongside everything else and gets the same consideration. Sometimes the single outlier is the most important signal.
It creates shared understanding. When the whole team participates in the sorting, everyone ends up with the same mental model of what users are experiencing. You don’t have to translate findings to stakeholders who weren’t in the room – they built the map too.
It produces defensible insights. Each theme you name is backed by specific, traceable observations. When someone asks “why are we prioritizing this?” you can point to the cluster.
That said, affinity mapping has real limits. It works best with data from 5–15 participants. Below that, you don’t have enough data to find meaningful patterns. Above 15–20 participants, the volume of notes becomes overwhelming and you might need to pre-process data before the session.
It’s also not a substitute for analysis. Grouping observations isn’t the same as interpreting them. You still need to think critically about why those observations cluster together and what they imply for your product or design decisions.
Best Practices for Running an Affinity Mapping Session
A few things that actually make a difference:
Keep data atomic. One observation per sticky note, always. The temptation is to write “users struggled with navigation and also mentioned that the search wasn’t helpful” on one note. Don’t. Split it. You lose flexibility when notes are compound.
Use participant language. Write quotes or close paraphrases rather than your interpretation. “I always forget where the settings are” is better than “navigation is confusing.” The participant’s words carry more meaning and they’re more useful in the sorting phase.
Include the source. A small notation like “P3” (participant 3) or “session 2” on each note lets you trace findings back to their origin. This matters when someone questions a theme – you need to be able to pull the original observation.
Timebox the silent sort. Without a time limit, silent sorting drags. 20–30 minutes is usually enough for 100–150 notes. Set a timer and stick to it.
Don’t over-label. Aim for 5–10 primary clusters for a typical round of research. If you end up with 20 groups, you’ve probably split things too finely and lost the higher-level themes.
Run it in two rounds for large datasets. For 200+ notes, do a first pass where you create rough groupings, then refine in a second pass with the whole team. Trying to do everything in one session with a large dataset often results in fatigue and sloppy clusters near the end.
Photograph before you clean up. Physical boards get knocked. Digital boards get accidentally reorganized. Save the state of the map before you start moving toward the polished version.
One more thing that’s easy to miss: the conversation during the session is often as valuable as the map itself. Team members will surface context, challenge interpretations, and share observations that didn’t make it into the notes. If someone’s running the session remotely, consider recording it (with consent) so those discussions aren’t lost.
How Articos Can Help With the Research That Feeds Your Affinity Map
Affinity mapping is only as good as the research that goes into it. If your interviews were rushed, if you only talked to three people, or if your data is a mix of vague impressions and actual observations – the map will reflect that.
The hardest part of building a solid affinity map isn’t the mapping session itself. It’s generating enough quality research data to feed it. Recruiting the right participants, scheduling calls that don’t fall apart, conducting structured interviews – that process typically takes weeks.
Platforms like Articos have made synthetic research a practical option for teams who need research data faster. Articos generates AI-moderated interviews with synthetic personas and delivers structured research reports in about 30 minutes, with no participant recruitment involved. For agencies and product teams running continuous discovery sprints, it’s become a way to have research data ready before the affinity mapping session – rather than waiting weeks to get started.
If your team needs user research without the recruitment overhead, it’s worth understanding how synthetic research fits alongside traditional methods in your workflow.
Start a free trial of Articos →
FAQs: Affinity Mapping
Yes – it’s one of the most practical uses of the method. If you’ve collected feedback through surveys, support tickets, or NPS responses, affinity mapping helps you find the themes that appear across individual comments. The key is making sure each feedback item gets its own note rather than grouping pre-analysis.
Miro and FigJam are the most widely used for remote teams. Lucidspark and MURAL are solid alternatives. For very large datasets, some teams pre-process notes in Notion or Airtable before importing into a visual board. The tool matters less than the process – the silent sorting step works the same way on a screen as on a wall.
There’s no fixed number, but 5–10 primary clusters is a reasonable range for a standard round of qualitative research. Fewer than 5 and you’ve probably over-collapsed distinct themes. More than 10 and the map loses its value as a synthesis tool. If you’re working with a second layer of higher-order themes, 3–5 is typical there.
Yes, and it’s one of the most common applications. The process works well when you pull direct quotes and specific observations from transcripts onto individual notes. One effective approach: go through each transcript and highlight anything that surprised you, contradicted your assumptions, or pointed to a clear behavior pattern – then transfer those highlights to your affinity board.
Document your themes with the supporting observations, then write a synthesis document that frames each theme as an insight with implications for your product, design, or strategy. Share it in a format stakeholders can skim – executive summary up front, full map available as reference. Then use the themes to directly inform your research questions for the next round of discovery, your design decisions, or your product roadmap prioritization.
The KJ method is the original name, coined by Japanese anthropologist Jiro Kawakita in the 1960s. Affinity mapping and affinity diagramming are largely the same process, adapted for product and UX research contexts. The underlying logic is identical: gather individual data points, group by natural affinity, name the clusters.
For a typical round of 5–8 user interviews (roughly 100–150 notes), plan for 2–3 hours including the silent sort, discussion phase, and labeling. Larger datasets with 200+ notes often need a full day or two separate sessions. Remote sessions tend to take 20–30% longer than in-person ones, mostly due to coordination overhead.
The core research team plus anyone who will act on the findings: product managers, designers, engineers if they’re involved in discovery. Keep it to 4–6 people if possible. Larger groups slow the session significantly and make the silent sorting phase harder to manage. Stakeholders who weren’t involved in the research can join for the theme-naming phase if you want broader input on prioritization.
