Affinity Diagram: Turn a Pile of Solution Ideas Into a Change Worth Piloting

The Affinity Diagram groups brainstormed solution ideas into natural themes. It reveals structure in a list that otherwise looks like noise, and it lets you prioritise improvement themes rather than isolated suggestions.

You reach for it at a specific moment in the Improve phase of RDMAICS. You have finished Analyse. You have a diagnosed cause. Your team has run a brainstorm and produced a list of ideas. The list is long enough that nobody can hold it in their head, and short enough that each idea still feels worth keeping. That is the gap the Affinity Diagram fills.

The tool does not score ideas against each other one by one. It asks a simpler question. Which ideas belong together, and which grouping carries the most weight. Once you can see that, the choice of what to pilot first becomes much easier to defend in a room.

Everything in this guide is built on one case. It is fictional. It is the sample data that ships with the tool, and it exists so you can see the mechanics without needing your own data first.

The scenario is an accounts payable process. The Analyse phase has pointed at manual handling and approval delays as the diagnosed cause. The team has run a solution brainstorm and come out with six ideas. Each idea has been written down, given a theme header, given a count of how many separate ideas fed into it, and given a priority.

Six ideas is a small set by design. Real brainstorms produce more. But six is enough to show the shape of the method, and small enough that you can check every step of the maths by eye.

The worksheet has four columns that matter. Theme or header. Solution idea. Count of ideas. Priority. You fill one row per idea.

In the sample data the six rows are as follows. Under Automation, OCR capture with a count of 4 and priority High, and Auto 3-way match rules with a count of 3 and priority High. Under Workflow, Auto-routing plus delegation with a count of 5 and priority High, and Mobile approvals with a count of 2 and priority Medium. Under Data Quality, Vendor master cleanup with a count of 3 and priority Medium. Under Vendor Engagement, Vendor portal plus validation with a count of 4 and priority Medium.

The theme header is the field people get wrong most often. Write it after the brainstorm, not before. If you set the themes up front, your team will file ideas into boxes you already believed in. Let the ideas sit on the table first, then look for what they have in common, then name the group.

The count of ideas column is not a guess at value. It records how many separate contributions collapsed into that single written idea. Auto-routing plus delegation has a count of 5, which means five people or five notes were pointing at the same thing. That is a signal about where attention sits, not a measure of benefit.

The tool returns affinity clusters. Ideas are visibly grouped into themes, with cluster size and priority retained. That last part matters. Grouping often destroys detail. Here it does not.

The sample data produces four clusters. Automation holds 2 ideas, OCR capture at High and Auto 3-way match rules at High. Workflow holds 2 ideas, Auto-routing plus delegation at High and Mobile approvals at Medium. Data Quality holds 1 idea, Vendor master cleanup at Medium. Vendor Engagement holds 1 idea, Vendor portal plus validation at Medium.

Read the clusters on two dimensions at once. Size tells you how much of the brainstorm landed in that theme. Priority tells you what the team thought of each idea on its own. Automation is the only cluster where both ideas carry High priority. Workflow is the same size but mixed, one High and one Medium.

The tool also offers an AI review that explains the result in words. Use it as a second reading, not as the decision. The numbers in front of you are the record.

A cluster is not a business case. It is a candidate. Before you commit to a pilot you need some sense of the gain, and the Affinity Diagram gives you three inputs toward that.

First, cluster size. Automation and Workflow each hold 2 ideas. Data Quality and Vendor Engagement each hold 1. The larger clusters carry more of the team's thinking, which usually means more surface area against the diagnosed cause.

Second, priority concentration. Three of the six ideas are marked High. Two of those three sit in Automation. The third, Auto-routing plus delegation, sits in Workflow and has the highest count of ideas in the whole set at 5.

Third, the count column read across the cluster. Automation combines counts of 4 and 3. Workflow combines 5 and 2. Those are not benefit figures and you should never present them as such. They are weight of attention. Use them to decide which theme to size properly with real process data, and leave the single-idea clusters of Data Quality and Vendor Engagement in the backlog until you have evidence to lift them.

The first decision is scope. You pilot a theme, not six separate ideas. In this case the realistic choice sits between Automation and Workflow, because those are the two clusters with 2 ideas each and the two clusters that contain High priority items.

The second decision is sequence. Clusters with a single idea, Data Quality at 1 and Vendor Engagement at 1, do not disappear. They go into the queue. Vendor master cleanup in particular often turns out to be a dependency for something else, so record it rather than discard it.

The third decision is who owns what. Each cluster usually maps to a different owner. Automation points to a systems owner. Workflow points to a process owner. Vendor Engagement points outward to supplier relationships. Naming an owner per cluster is often the most practical output of the whole exercise.

The fourth decision is what you measure in the pilot. You cannot define that until you know the theme. Once Automation is the chosen theme, the measure follows from the two ideas inside it, not from the full list of six.

The most frequent mistake is treating cluster size as a ranking. A cluster of 2 is not twice as good as a cluster of 1. It means two written ideas were grouped there. A single idea with a strong business case can beat a crowded theme. The diagram shows you structure, and you still have to think.

The second mistake is naming themes too early, which quietly forces the result. The third is letting one voice do all the grouping. Affinity work is better done with the people who produced the ideas, because the grouping argument is where the real understanding appears.

The fourth is dropping the low-count clusters out of the record entirely. Data Quality and Vendor Engagement each hold 1 idea here. Keep them visible. A later pilot may need them.

The fifth is skipping Analyse. If you have not diagnosed the cause, the brainstorm is a wish list and the clusters are just tidy wishes.

The Affinity Diagram will not tell you whether a change works. It has no view of cost, effort, risk, time to implement or benefit in dollars or hours. None of those appear in the sample data, and none appear in the result.

It will not validate your priorities. High and Medium in this worksheet are judgements made by the people in the room. The tool carries them through to the output faithfully, but faithfully carrying a weak judgement does not make it strong.

It will not replace a pilot. The output here is a shortlist of themes, with Automation and Workflow holding 2 ideas each and the other two holding 1. Choosing one of those and running a controlled trial against a baseline is still the next step, and that step belongs to you.

It will not stop you from grouping badly. Different teams will group the same six ideas differently, and both groupings can be defensible. Treat the clusters as a structured record of how your team sees the problem today, not as an objective truth. Then go and test it.