Every business has customers — but not all customers are equal. Some buy frequently, spend generously, and came back just last week. Others placed one order two years ago and have never returned. Treating both the same way is one of the most common (and costly) mistakes in marketing.

RFM Analysis is the framework that fixes this. It scores every customer on three behavioural dimensions — Recency, Frequency, and Monetary value — and uses those scores to group customers into actionable segments. No complex algorithms. No black boxes. Just a clear, systematic view of who your best customers are, who is slipping away, and who you’ve already lost.

Video Lecture

What this video covers

RFM stands for three behavioural dimensions:

  • Recency — How recently did the customer last buy? Recent buyers are far more likely to purchase again.
  • Frequency — How often do they buy? Frequent buyers have already proven their loyalty.
  • Monetary — How much do they spend? High spenders drive a disproportionate share of revenue.

Scoring: 1 to 5 Each dimension is scored independently from 1 (worst) to 5 (best) using quintiles. A customer who scores 555 — the maximum — is your champion: recent, frequent, and high-spending. A customer scoring 111 is lapsed, infrequent, and low-value.

The 4-Step Workflow From raw transaction data to labelled segments in four steps: collect data → build the RFM table → assign scores → create segments.

The 5×5 Segment Map A 5-by-5 grid of Recency vs. Frequency reveals 25 distinct customer groups. Champions sit in the top-right. At-Risk customers once bought frequently but haven’t recently. Lost customers in the bottom-left are the hardest to reach.

Hurdle Rate Analysis Hurdle rates track the percentage of your customer base clearing a defined threshold for each RFM dimension. Track these over time: rising rates signal a healthy business; falling rates are an early warning that your best customers are drifting.

Why RFM works

RFM is powerful because it is behavioural, not demographic. It doesn’t care where a customer lives, how old they are, or which channel they came from. It measures what they actually do — and that turns out to be a far better predictor of what they’ll do next.

It is also interpretable. Unlike clustering algorithms that produce groups you have to reverse-engineer, RFM segments have names you can act on immediately: Champion, At Risk, Promising, Lost. Your marketing, retention, and service teams can speak the same language.

This lecture is part of the NUS Customer Analytics course series. The next topic is Cluster Analysis — where machine learning automatically discovers hidden customer segments without predefined labels.