What RFM segmentation reveals on a public ecommerce dataset

Worked example on a public, open dataset — no client or private data. This is the shareable version of the kind of analysis I run privately on real store data.

One of the fastest ways to understand an ecommerce business is to stop looking at totals and start looking at who is buying. RFM — Recency, Frequency, Monetary — scores every customer on three axes and buckets them into segments like Champions, At Risk, and Hibernating.

Run on a typical open orders dataset, the pattern is almost always the same and almost always surprising to the owner:

• A small Champions segment (often <15% of customers) drives a wildly outsized share of revenue.
• A large At Risk segment — customers who used to buy often but have gone quiet — represents revenue that is already earned and quietly leaking away.
• The Hibernating tail is big but low-value; discounting to win it back usually costs more than it returns.

The actionable read isn’t “get more customers.” It’s: protect the Champions, and win back the At Risk group before they’re gone — two moves that need no new acquisition spend.

That’s the shape of it from public data. The interesting part is how little math RFM actually needs — it’s three groupby operations and a quintile split — yet it reframes a business more usefully than most dashboards. A recurring theme in this kind of work: the method that changes the decision is often much simpler than the method that looks impressive.