Affinity diagram: steps, example and a copyable template

What is an affinity diagram?

An affinity diagram is a working arrangement of related items, often represented by sticky notes or movable cards. ASQ’s method guide describes it as organizing ideas into their natural relationships and traces the method to Japanese anthropologist Jiro Kawakita in the 1960s. The name KJ method refers to Kawakita Jiro; affinity mapping is another common label for this family of practices.

The key move is to let the material influence the categories. You first compare the notes, then write a header that captures what a group has in common. Starting with a set of required categories and allocating every note to one of them is a different operation. It may be appropriate for your project, but it removes the open-ended discovery that an affinity sort is meant to allow.

The finished diagram usually has three layers: individual notes, named groups and, when useful, broader themes joining several groups. It can also include outliers, disputed placements and questions. It does not need a perfectly symmetrical layout, equal-sized groups or a category for every noun appearing in the data.

When to use affinity mapping—and when not to

Interview observations

Compare what participants said or did, preserving a participant ID and page, timestamp or session reference. Separate direct statements from the researcher’s interpretation.

Field or research notes

Look for related conditions, needs, tensions or unanswered questions across observations. Keep the date and setting when they affect the meaning.

Brainstorm output

Group possible approaches after generating them. These are proposals, not findings about what people want; give them a visibly different status from research evidence.

Open survey comments

Compare themes in written responses while retaining response IDs and relevant context. Do not turn the size of a sticky-note cluster into a population percentage.

Use the method when a set of items is hard to inspect in its current order and you need to understand relationships. For generating the ideas themselves, use an ideation technique first. For chronology or a claim with supporting and conflicting sources, an evidence board may be more suitable than a thematic sort.

Do not use an affinity diagram to stand in for a statistical analysis, establish causation or claim that your sample represents a wider population. If the research question concerns how often something happens, return to the original records and an appropriate counting or analysis method. The diagram can point toward a useful question without answering every kind of question.

Prepare one observation per note, with its source

Write a focus question at the top before touching the notes. “What makes the first visit to this workshop difficult?” is more useful than “What themes do we have?” State which sessions, dates or responses are included and what the output should support. Do not silently combine separate studies whose contexts differ.

Nielsen Norman Group’s account of affinity-diagramming pitfalls emphasizes that the focus question changes how notes should be grouped. The same set can reasonably be sorted by location, by task or by underlying difficulty. Choose the comparison that serves the question, and keep that choice visible.

• Give the note a stable ID, such as O01, and write one observation or idea on it.
• Attach a source reference: participant or response ID, source title or file, and relevant page or timestamp.
• Label whether it is a quotation, an observed action, a paraphrase, your interpretation or a proposed solution.
• Split compound notes when the parts could belong to different groups. Preserve the same source reference on both.
• Remove personal information that is unnecessary for this analysis, keeping any required identifying records separately with suitable access controls.

For a real quotation, copy exactly and retain enough context to check its meaning. For a paraphrase, do not use quotation marks. Keep solutions separate from observations: “add a welcome sign” cannot substitute for “the visitor asked where to enter.” Source discipline matters more than matching sticky-note colors.

How to make an affinity diagram step by step

1. Spread the notes out. Give everyone a chance to read the full set rather than hiding parts behind stacks or overlapping cards.
2. Sort silently at first. Move notes that seem meaningfully related near each other without announcing a preferred category system. Leave notes alone when they do not yet fit.
3. Compare and rearrange. Read the actual observations, not just repeated keywords. If a note belongs in two groups, use a marked duplicate with the same ID or a reference to the original.
4. Discuss the emerging groups. Ask what each group has in common, why a disputed note is there, and whether the focus question still guides the arrangement.
5. Name the groups after sorting. Write a specific header that summarizes the shared pattern. A short sentence such as “Visitors cannot predict what to bring” is more informative than “Information.”
6. Add hierarchy only when useful. Join related groups under a broader theme without erasing the distinctions underneath. Skip this layer if the smaller groups already explain the material clearly.
7. Write insights and next questions. Connect each interpretation to the observation IDs supporting it, include conflicting items, and identify what needs checking before a decision.

ASQ describes silent grouping, delayed headers and optional supergroups; these steps follow that broad procedure. Silence is for the initial sort, not a ban on examining disagreements. Allow discussion once the first relationships are visible, particularly when a note’s wording or source context is unclear. A facilitator can keep the question in view without deciding all the groups in advance.

There is no required number of clusters. Stop rearranging when the group can explain the current structure and see what it leaves unresolved. Photograph or otherwise preserve a useful intermediate layout before a major change if you need to document how the interpretation developed.

A fictional affinity diagram example

Suppose you are reviewing a fictional community pottery workshop. The question is “What makes a first visit difficult?” All observations, IDs and circumstances below are invented for this example; they are not customer research or actual participant quotations. Begin with six paraphrased observation cards rather than prewritten categories.

• O01 / fictional session P01: the visitor asks whether tools are provided.
• O02 / fictional session P02: the visitor brings materials that the workshop already supplies.
• O03 / fictional session P03: the visitor waits outside the wrong entrance.
• O04 / fictional session P04: the visitor asks which door to use after arriving.
• O05 / fictional session P05: the visitor cannot tell whether the listed time is arrival or teaching time.
• O06 / fictional session P06: the visitor asks if the clay used in class can be taken home immediately.

One plausible initial sort puts O01 and O02 together, O03 and O04 together, and leaves O05 and O06 separate. After reading the groups, name the first “Visitors cannot predict what to bring” and the second “Visitors cannot identify the entrance.” O05 concerns timing; O06 concerns the expected result. Do not force either into “arrival” merely to achieve two tidy groups.

A broader theme might be “The invitation leaves practical expectations open,” with subgroups for preparation, arrival, timing and outcomes. That theme is an interpretation of these invented cards, not a demonstrated explanation for all first-visit difficulties. Another analyst might separate arrival logistics from expectations about the activity. The diagram should make the reasoning discussable rather than disguise it as inevitable.

Write a limited insight: “In this fictional set, preparation questions concern whether supplies are provided (O01, O02).” Then write a separate proposal: “Try an invitation that explicitly lists supplied and required materials.” Finally record the check: “Compare how first-time visitors interpret the revised invitation.” Observation, interpretation, proposal and next check remain distinct.

Turn group labels into traceable insights

A label names the shared thread; an insight explains why it matters to the question. “Entrance confusion” is a label. “Two observations in this set concern choosing the correct entrance” is a bounded summary. “The workshop needs an app” is a proposed solution and introduces a leap the cards do not establish.

• Pattern: what do the source notes actually have in common?
• Support: which observation IDs demonstrate the pattern in this set?
• Boundary: in which context did it appear, and what conflicting or missing evidence remains?
• Implication: what question or decision does the pattern raise?
• Next check: what could confirm, complicate or overturn the interpretation?

If you want to count, return to the records. Several cards may describe one participant or one incident, and duplicated notes must not be counted twice. An unusual but relevant observation deserves a visible place even if it has no neighbors. A large cluster is not automatically more urgent than a small one: relevance, severity and your decision criteria need separate consideration.

Doing affinity mapping on a digital canvas

Keep a source area apart from the provisional grouping area. Arrange small selections around the focus question, then retain the source IDs when notes move. The guide to organizing research visually explains the same distinction between source excerpts, your interpretation and open questions. The visual research use case shows why keeping them nearby is useful.

Canvas Perch supports notes, images, links and files side by side on an infinite canvas, connections between cards, and grouping selected cards inside a frame. You can use frames for provisional clusters and keep a note beside a connection explaining its meaning. This is manual arrangement, not automated research synthesis.

Do not turn frames into permanent categories before the sort. Leave an area for ungrouped notes and another for disputed interpretations. Use readable written headings; color may help navigation, but the group should still make sense without it. Preserve the source records independently of the board so the diagram does not become your only account of the research.

Affinity mapping pitfalls to catch before sharing

Forcing categories

If a note does not fit, leave it separate or revise the grouping. A tidy diagram is not worth losing an important difference.

Labeling too early

Headers can steer later sorting toward the facilitator’s expectations. Compare the observations before naming their shared thread.

Matching words instead of meaning

Notes mentioning a “door” may concern accessibility, signage or security. Group by the relationship relevant to your question, not a shared noun.

Losing the source

Keep IDs on notes, duplicates and insight summaries. A persuasive header is not a substitute for the original observation.

Letting one interpretation dominate

Begin with silent sorting, invite explanations of disputed placements, and preserve alternative readings when the evidence does not resolve them.

Treating a sort as a final decision

Add an explicit evaluation or research step. The map can organize possibilities, but priorities and conclusions still need reasons.

Better capture makes later sorting easier. If your inputs are sprawling meeting summaries, use note-taking methods to produce clearer records first. Do not repair uncertain notes by confidently rewriting what a participant might have meant.

A copyable affinity diagram template

Copy these fields into a document or put each one in a labeled area on a board. Use the brackets as fields to fill in, not as invented research. For a solo review, the same structure works without a facilitator: record the alternative groupings you considered before choosing one.

• Focus question: [What relationship or problem are we examining?]
• Scope: [Included sources, dates, sessions and exclusions]
• Observation card: [ID] / [One observation] / [Type: quote, action, paraphrase or interpretation] / [Source and location]
• Ungrouped or disputed: [Card IDs] / [Reason for leaving the placement open]
• Cluster header: [Specific shared pattern, named after sorting] / [Supporting card IDs]
• Broader theme, if needed: [Related cluster names] / [What joins them without erasing differences]
• Insight: [Bounded interpretation] / [Supporting and conflicting IDs] / [Uncertainty]
• Next check or action: [What to examine] / [Who will do it, if relevant] / [What would change the interpretation]

Before sharing, read each insight back against its source notes. Check that invented examples are labeled, quotations are exact, duplicates remain identifiable and personal information is necessary for the audience. The useful deliverable is not just a wall of groups: it is an explanation someone else can follow and question.