日本語

Affinity Diagram: How to Organize Ideas (Steps + Examples)

Affinity diagram showing scattered note cards grouped into labeled clusters on a board

An affinity diagram takes a pile of raw ideas, complaints, or observations and sorts them into meaningful groups based on natural relationships. It's one of those tools that looks almost too simple on paper but consistently changes how teams think about problems.

What is an affinity diagram?

An affinity diagram is a structured group method for organizing large volumes of qualitative data, ideas, or observations into clusters based on natural relationships, revealing patterns that would stay hidden in an unorganized list.

It belongs to the seven management and planning tools, a set codified by the Japanese Union of Scientists and Engineers (JUSE) in the 1970s alongside tools like the interrelationship digraph and tree diagram. The method originated with Jiro Kawakita, a Japanese anthropologist who developed it in the 1960s as a way to process field data, which is why it's also called the KJ Method.

The core mechanic is simple: each idea or data point lives on its own card or sticky note. The team sorts those cards silently into groups, then names each group based on what the members share. The result is a visual map of themes, not a ranked list and not a fishbone with a predetermined cause-structure. You're discovering the structure from the data, not imposing it.

Key Facts

  • The affinity diagram is one of the seven management and planning tools (7 MP tools) standardized by JUSE and published in the 1979 book "Seven New Quality Tools for Managers and Staff," translated into English in 1983. (Source: JUSE / ASQ Quality Resources)
  • Jiro Kawakita developed the KJ Method in the 1960s to organize ethnographic field data, making it one of the earliest formalized methods for qualitative data synthesis. (Source: ASQ, "What is an Affinity Diagram?")
  • ASQ lists the affinity diagram as the first of the seven management and planning tools, typically applied at the earliest stage of problem-solving when data is still unstructured and categories are unknown.

When to use an affinity diagram

Not every situation calls for this tool. It works best when:

  • You have a large, unstructured set of ideas or data (typically 20 or more items) that need to be organized before further analysis.
  • The team doesn't yet know what categories exist. If you already know the buckets, a simple list is faster.
  • You want to surface consensus without a single person imposing structure. The silent sorting step is deliberate: it reduces anchoring to the first voice in the room.
  • You're in the early stage of a root cause analysis and need to organize complaints or symptoms before building a fishbone diagram.
  • You've run a voice of the customer session and have pages of customer quotes to make sense of.
  • Your team disagrees on priorities and you need a shared picture before moving to solutions.

It's less useful when your data set is small (fewer than 15 items), when the categories are obvious, or when you need a tool that shows cause-and-effect relationships. For those situations, a fishbone diagram or five whys is a better fit.

How to create an affinity diagram

The process takes anywhere from 45 minutes to two hours depending on the volume of data. Five to eight participants is the practical range. Fewer and you don't get enough cross-functional perspective; more and the silent sorting step turns chaotic.

Step 1: Define the question or problem

Write a clear focus question at the top of the workspace. Something like "What are the barriers customers face during onboarding?" or "Why are defects increasing in the final inspection step?" The question keeps the ideation on topic and gives you a filter when someone adds a card that belongs in a different conversation.

Step 2: Generate ideas on individual cards

Each participant writes one idea, observation, or data point per card. Physical sticky notes on a whiteboard or virtual cards in a collaboration tool both work. Aim for concrete, specific language: "checkout confirmation email arrives 20 minutes late" is more useful than "emails are slow."

Don't evaluate at this stage. Write everything, even things that seem obvious or redundant. Volume matters here because the grouping step will filter signal from noise.

Step 3: Sort cards silently into groups

Post all cards on the wall and have participants move them into groups without talking. Silence is not a gimmick. It prevents early anchoring, where whoever speaks first shapes everyone else's thinking. If someone moves a card you just placed, let them. Move it back if you disagree. Repeated tug-of-war on a single card is useful data: it usually means the idea belongs in two groups, which itself is a pattern worth noting.

Keep sorting until movement slows. Most teams reach rough agreement in 10 to 20 minutes.

Step 4: Name each group

Once the groups stabilize, the team names each cluster. The name should capture the essence of what the cards share, not just repeat the most common word in the group. "Communication delays" is more useful than "emails" if the cluster contains feedback about late emails, missed Slack messages, and unanswered calls. A strong group name often becomes the framing for a later problem statement or improvement hypothesis.

If a card genuinely belongs in two groups, you can copy it. If a card fits nowhere, give it a "miscellaneous" group of one and revisit it after the main clusters are settled.

Step 5: Draw the affinity diagram

Arrange the named groups on a clean surface, either physically or digitally. Draw a boundary around each cluster and write the group name at the top. Optionally, draw lines between groups that are related to each other. The finished diagram is your shared picture of the problem landscape. Take a photo or export it, then use it as the input for the next tool in your process.

Affinity diagram examples

Use case How it's applied Output used for
Sprint retrospective Team writes one observation per card (what went well, what didn't, what's unclear). Cards sort into themes like "communication gaps" and "unclear requirements." Retrospective action items, prioritized by theme size
Voice of customer analysis 80 customer interview quotes transferred to cards. Silent sort reveals top themes: "onboarding confusion," "pricing transparency," "missing integrations." CTQ inputs for a kano model analysis
Root cause brainstorm Team generates possible causes of a defect surge. Cards group into "equipment," "process," "training," and "materials." Starting point for a fishbone diagram or five whys session
Product discovery Design team organizes 120 user research observations from usability tests. Groups reveal three distinct user mental models. Design principles and navigation structure

The retrospective use case is worth expanding. Many teams run retros as a free-for-all list, which produces 40 items no one has the energy to prioritize. Running it as an affinity exercise first means the team is voting on themes, not individual cards, which cuts the list to 5-7 actionable areas and keeps the conversation focused.

Affinity diagram vs other tools

Tool What it does Best for Key difference from affinity diagram
Affinity diagram Groups ideas by natural relationships, discovered through sorting Early-stage organization when categories are unknown Inductive: structure emerges from data
Fishbone diagram Maps causes to an effect using predefined branches (people, process, equipment, etc.) Root cause analysis with a known effect Deductive: structure is imposed before you start
Mind map Radiates ideas outward from a central concept, usually built by one person or a facilitator Brainstorming, note-taking, planning Hierarchical and facilitator-driven vs. group-driven flat sorting

The practical implication: use an affinity diagram to discover what your categories are, then use a fishbone to structure the analysis within one of those categories. They're sequential, not competing.

Common mistakes

Starting with categories already written. If you write "people, process, technology" on the board before participants sort their cards, you've made an affinity diagram into a sorting exercise for pre-decided buckets. The value of the method is that the groups emerge from the data.

Letting one person dominate the sorting. The silent-sort step exists to prevent this. If someone keeps explaining their moves aloud, the facilitator should gently redirect: sort first, discuss patterns once the groups stabilize.

Groups that are too broad. A cluster named "process problems" that contains 30 cards isn't useful. Break it into sub-groups. Most teams stop too early. Push for specificity in the group names until each cluster is narrow enough to act on.

Skipping the naming step. Unlabeled clusters leave interpretation to whoever reads the diagram next. Names create shared meaning. If the team can't agree on a name, that's a signal the cluster needs splitting.

Treating it as a one-time exercise. Affinity diagrams work at the start of a problem-solving cycle. But if you re-run the exercise six months later with new data, you'll often find that your old categories no longer fit, which tells you something important about how the problem has evolved.

Best practices

Limit sessions to eight participants. Above that, the silent-sort becomes harder to manage and sub-groups start forming, which undermines the shared picture you're building.

Use physical materials when possible. There's something about moving a sticky note with your hands that creates different cognitive engagement than dragging a card on a screen. For remote teams, virtual whiteboards work well, but build in extra time for technical setup.

Brief participants before the session. Explain the silent-sort rule, the goal of the question, and what you'll do with the output. A five-minute briefing prevents the "wait, why aren't we talking?" confusion mid-session.

Set a time limit for sorting. Open-ended sorting drags on. Tell the group they have 15 minutes to sort. The slight time pressure keeps the session moving and prevents over-analysis of individual card placements.

Pair with the seven quality tools or A3 problem solving. The affinity diagram produces a landscape of themes. Those themes need structured analysis to move from observation to solution. Pairing it with A3, fishbone, or five whys closes that gap.

Frequently asked questions

How many participants does an affinity diagram session need? Five to eight is the practical range. You need enough perspectives to surface ideas a single person would miss, but not so many that silent sorting becomes unmanageable. For a simpler problem with a smaller data set, three to four participants can work.

Can you run an affinity diagram remotely? Yes. Virtual whiteboard tools like Miro or FigJam replicate the card-sorting mechanic well. The main adjustment: mute everyone during the silent-sort step and agree in advance that moving someone else's card is allowed. The same social friction that slows remote teams in discussion helps here, because participants tend to sort more deliberately when they can't explain their moves in real time.

What is the difference between an affinity diagram and affinity mapping? The terms are used interchangeably. "Affinity mapping" is more common in UX research and design contexts; "affinity diagram" is the term used in quality management and the seven management and planning tools framework. The mechanic is the same.

How long does an affinity diagram session take? For a set of 20 to 40 cards, plan for 60 to 90 minutes: 15 minutes for idea generation, 15 to 20 minutes for silent sorting, 20 minutes for group naming, and 15 minutes for discussion and next steps. Larger data sets, such as 80 or more cards from a customer research batch, can take two to three hours.

When should I use a fishbone diagram instead? Use a fishbone when you already know the effect you're analyzing and want to systematically explore its causes using predefined categories (people, process, equipment, environment, measurement, materials). Use an affinity diagram when you don't yet know what your categories are and need to let them emerge from data.


The affinity diagram's real value isn't the pretty cluster map at the end. It's the shared understanding the team builds by sorting together. When five people move cards silently and end up with similar groups, they've found consensus without the politics of open debate. That's a faster and more honest starting point for any improvement project.

Use the diagram as an input, not an output. Once your themes are named, the work of root cause analysis, prioritization, and process redesign can begin with a shared picture of the problem that the whole team built together.