Creating Segments of Observations for Business Reasons (RFM)#

Stage 4 · 🧩 Sampling, Partitioning & Segmentation · Lesson 30 of 56 · intermediate

◀ Previous · RFM Analysis · Next · Least Squares Regression ▶ · ↑ Section

Important

✨ AI-generated content. This page was written with the assistance of an AI language model and is provided as a learning aid. Despite careful review, it may still contain mistakes, omissions, or out-of-date information. Whether you are new to the topic, a team lead, or a senior practitioner, treat it as a starting point rather than an authoritative reference: read it critically and independently verify anything you act on (code, commands, figures, and factual claims) against official documentation and primary sources before relying on it.

From codes to segments#

125 RFM codes are too many to act on. The final step is to group them into a handful of named segments — typically six to ten — each describing a recognisable kind of customer and, crucially, each calling for a different business response. This is where scores become decisions.

A common taxonomy#

A widely used starting taxonomy names segments by their RFM profile:

Champions — high on all three (555, 554): the best customers.
Loyal Customers — buy consistently and often.
Potential Loyalists — recent buyers with growing frequency.
New Customers — recent, but few purchases so far.
At Risk — once frequent and high-spending, but lapsing (low recency).
Can’t Lose Them — high past value, gone quiet.
Hibernating / Lost — low on everything, long inactive.

Each segment, an action#

The point of naming segments is the “so what”. Champions get rewards, early access and referral asks — not blanket discounts that erode margin. At Risk and Can’t Lose Them get win-back campaigns and personal outreach, ideally before they fully churn. Hibernating / Lost get a re-permission push or are suppressed to save budget. One message for a new buyer and a ten-year loyalist would waste both; segmentation lets each be treated for who they are.

Segments as clusters#

Notice this is the clustering idea from earlier in the stage, made concrete: customers are grouped by similarity — here, similarity in RFM space — so that each group can be understood and served differently. Whether the groups come from a rule-based RFM taxonomy or from an algorithm like k-means, the goal is the one this stage began with: turn a mass of observations into meaningful segments a business can act on. With customers understood, the course turns next to predicting outcomes — starting with regression.

Hint