RFM Segmentation

RFM Segmentation scores customers on how Recently they bought, how Frequently, and how much they spend (Monetary value), sorting them into groups so you stop treating your best customer the same as one who left a year ago.

Recency runs one axis and frequency the other, sorting customers into named clusters in the grid.

Customers by frequency and recency of purchase.

Reach for this when…

You send the same offer to your entire list regardless of who they are.
Unsubscribes are climbing and you suspect you're annoying your best customers.
You have a hunch some customers have quietly gone cold and want to know which ones.

How to run it

Pull purchase history: last order date, order count, total spend per customer.
Score each customer on Recency, Frequency and Monetary value, typically 1-5.
Group scores into segments: Champions, New, At Risk, Lost.
Design a different message and offer for each segment.
Re-score on a regular cycle so customers move between segments as behaviour changes.

A worked example

Situation. Aino Korhonen ran Tampere Threadworks, an online textile and apparel retailer in Tampere, Finland, and had been sending the same weekly discount email to her entire customer list for two years.

Applied. She ran RFM segmentation on twelve months of order data and found a third of the list hadn't bought in over six months despite once being frequent buyers - a distinct At Risk group hiding inside the mass mailing.

Result. She stopped discounting her Champions, who were buying anyway, and put the discount budget into a win-back sequence for the At Risk group instead. Email volume dropped and the win-back sequence recovered a meaningful slice of dormant revenue.

Manila Threadworks found a third of its list sitting in At Risk, hidden inside one mass mailing list.

The catch

RFM only sees purchase behaviour, so it will misclassify a genuinely loyal customer who is between purchase cycles as At Risk, and a one-off big-ticket buyer as a Champion. It also says nothing about why someone stopped buying, only that they did - you still need to ask.

A high Monetary score from one huge order looks identical to real loyalty until you check the Frequency column.

Origin: Arthur Hughes (database marketing)

connecteddale