RFM segmentation
On this page
Definition
RFM is a segmentation method built on three behavioural axes: Recency (how recently a customer bought), Frequency (how often), and Monetary (how much). Each customer gets a score per axis, and the combination of scores defines a segment. The method needs no training and is directly interpretable by the business.
How to compute
Build 1–5 scores per axis from the base’s percentiles, giving a triple such as 5-4-5. Define segments by rules: “champions” are high R, F, M; “at risk” are high F, M with low R. Then assign an action to each segment: retain, reactivate, upsell.
Pitfalls
Bin boundaries are arbitrary and strongly change the result, so choose them deliberately rather than by default. Monetary correlates with Frequency, so two axes partly duplicate each other. On a small base the scores are noisy. A segment with no assigned action is useless: the point of RFM is working the groups, not drawing a chart.