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.

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

Run it on a typical open orders dataset and the same pattern shows up, one that usually surprises the owner:

• A small Champions segment, often under 15% of customers, drives an outsized share of revenue.
• A large At Risk segment, customers who used to buy often and have gone quiet, holds revenue that is already earned and quietly leaking away.
• The Hibernating tail is big but low-value. Discounting to win it back costs more than it returns.

The actionable read is not “get more customers.” It is protect the Champions and win back the At Risk group before they are gone, two moves that need no new acquisition spend.

That is the shape of it from public data. The interesting part is how little math RFM needs. Three groupby operations and a quintile split reframe a business more usefully than most dashboards. It is a recurring theme in this work: the method that changes the decision is often much simpler than the method that looks impressive.