Customer Segmentation That Actually Drives Revenue

We have seen beautiful customer segmentations die in a slide deck. The clusters were statistically sound, the chart was colourful, and absolutely nothing changed because nobody knew what to do on Monday morning. When we take on a segmentation project, our first question is not “how many segments?” — it is “what decision will this change?”

We start from the action, not the algorithm

If marketing wants to know who to send a retention offer to, that is a different model than if finance wants to forecast which accounts will expand. We pin down the decision first, because it tells us which features matter and how many segments are actually useful. Five segments a team can remember beats fifty a dashboard can plot.

We lean on behaviour — what people actually do — over demographics, which often just describe who they are. Recency, frequency and value get us a long way before anything fancier is needed. And we give each segment a human name, because “high-value, at-risk loyalists” drives action in a way that “Cluster 3” never will.

“A segment is only real once a team has changed what they do because of it.”

Then we close the loop. We wire the segments into the tools teams already use, watch what happens when they act on them, and refine. That feedback is what turns a one-off analysis into something that keeps earning its place in the business.