RFM Analysis

RFM analysis is a segmentation method that ranks customers on three measures taken from transaction history: how recently they purchased, how often, and how much they spent. It uses no demographics and no trained model, which is why it can be built from a single table of orders.

Its premise is behavioral and should be treated as an assumption rather than a fact: customers who bought recently, frequently, and in quantity are more likely to buy again. That assumption earned RFM its place in direct marketing and retail, where purchases repeat on a short cycle.

Where cycles are long or contractual, it misreads. A customer who replaces a mattress every eight years is not disengaged in year three, and a subscription that renews annually produces a frequency of one regardless of satisfaction. In those settings the recency dimension measures the product’s replacement interval rather than the customer’s interest.

Two properties matter when reading the output. RFM is relative — scores rank customers against each other at one moment, so a customer can lose a band because others improved. And it is descriptive — it identifies who to look at first and explains none of the behavior it ranks.

RFM can provide a transaction-based starting point for prioritization. Its scores do not explain why customers buy, estimate every component of lifetime value, or show which intervention will change their behavior.