Affinity Diagram · KJ Method
Affinity Diagram / KJ Method
What Is an Affinity Diagram?
- Write ideas on cards and group similar ones
- Sort silently, without setting criteria in advance
- Each header should state a finding
What this technique actually changes is the order of the meeting. In a typical meeting, someone throws out a frame first ("Isn’t this just a communication problem?") and all the other opinions get pulled into it. Affinity diagramming works the other way around. It postpones judgment to the very end, lets the data cluster on its own, and only then names the clusters based on their shape. Just by switching to drawing conclusions last, issues that never once appeared in the meeting minutes come to the surface.
By nature, this is a bottom-up tool. Where a logic tree or a fishbone diagram sets the big framework first and breaks it down, an affinity diagram starts from individual notes and builds upward. The two are not competitors but sequential steps: when the problem definition is fuzzy, use an affinity diagram to surface the issues, then a logic tree to break down the causes. If the definition is already clear, there is no reason to make cards.
The conditions for it to work are fairly demanding. Somewhere between 30 and 100 cards is manageable; with about 10, plain discussion is faster, and beyond 200 you need two stages: group once, then group the headers again. Aim for 4 to 7 participants and 5 to 9 final groups, which is the range the eye can take in. Cards should contain facts and actual quotes, not solutions. The moment you write 'need to improve communication,' that card stops being data and becomes a conclusion already reached.
The most common mistake in practice is believing that putting up sticky notes means you did affinity diagramming. In reality, most teams group cards by the org chart they already know, such as sales, production, and administration, and once headers get two-word nouns like 'compensation' or 'policy', the result circles back to what people meant to say in the first place. Photographing the wall and dispersing is another problem. Unless you move the header sentences and the arrow relationship map into a document that same day and put them on the next meeting agenda, a half-day workshop ends up as a feel-good event.
John KimInstructor John Kim, author of B2B Marketing Design and CEO of Nanum Consulting
Ara LimSpecialist in global leadership, AI, and business communicationDevised by Japanese cultural anthropologist Jiro Kawakita as a method for organizing field data, it is called the KJ method after his initials and was adopted as one of the seven new quality control tools.
Consider a hypothetical case. Company A makes auto parts, with 320 employees and production-worker turnover in the 20% range for the third year running. HR had accumulated 140 exit interview records, but reports always boiled down to a single line, 'pay complaints,' and that report was sent back three times at executive meetings. In the end, six practitioners were gathered for a four-hour workshop.
- HR Manager Kim went through the interview records and transferred only what departed employees actually said onto 142 cards. Summaries like 'pay is low' were left out; entries were written verbatim, such as "The shift schedule changes every week."
- Six people from production, HR, and safety collected cards for 40 minutes without saying a word. When Foreman Park started to explain something midway, the facilitator cut in: "Hands only for now."
- When someone wrote 'compensation' on a header, the facilitator asked, "So what did you actually discover?" and it was rewritten as "There’s no one to teach you once you join." Seven headers like this, written as full sentences, came out of the session.
- They set aside the 11 cards where opinions split and discussed them for 15 minutes. Senior Manager Choi said, "This isn’t about pay, it’s a shift-rotation problem," so 4 cards were moved, and 3 ambiguous cards were placed in both groups.
- When they drew arrows between groups, a chain emerged: 'skilled workers leave → on-the-job training gap → new hires quit within 3 months.' The title of the executive report changed from a pay-raise proposal to redesigning the mentor program.
- Six months later, the rate of employees leaving within three months of joining fell from 34% to 19%. The exit interview form was also redesigned in card format and is now run the same way every quarter.