Market Segmentation

RFM Analysis for Customer Segmentation: Scores, Segments and an Example

RFM analysis scores customers on recency, frequency and monetary value to form segments. Learn how to score, group and act on them, with an example.

On this page
  1. What the three measures mean
  2. How scoring works
  3. Turning scores into segments
  4. A worked example
  5. Where RFM falls short
  6. Steps to build an RFM model
  7. Related articles

RFM analysis groups existing customers by how they buy. It uses three measures from transaction records: how recently a customer bought (recency), how often they bought (frequency), and how much they spent (monetary value). Because it needs only purchase data, it is one of the most practical ways to segment customers. It is a form of behavioral segmentation, and it works on existing customers rather than on the whole market. This article explains how to score customers, turn the scores into segments, and where the method falls short.

What the three measures mean

  • Recency (R): how many days have passed since the customer’s last purchase. A recent buyer is more likely to respond to an offer.
  • Frequency (F): how many purchases the customer made in a chosen period, such as the last 12 months. Regular buyers tend to be more loyal.
  • Monetary value (M): how much the customer spent in the same period. High spenders often matter most to revenue.

Choose the time window before you start, and use the same window for every customer. A 12-month window is common for many retail businesses, but the right length depends on how often your customers buy.

How scoring works

Each customer gets a score on each measure. There are two common approaches:

  • Quintiles: rank customers on each measure and split them into five equal groups, scored 1 to 5. A score of 5 for recency means the customer is in the most recent fifth.
  • Simple cutoffs: set thresholds by hand, such as “bought in the last 30 days.” Cutoffs are easier to explain, but they need checking against your own data.

The three scores can be joined into a code. A customer scored 5 on recency, 4 on frequency and 5 on monetary value has the code 545. Codes make it easy to sort customers, but a model with 5 × 5 × 5 scores produces 125 possible codes, which is too many to manage. Most teams group codes into a handful of segments.

Turning scores into segments

Segment names vary between tools, and there is no official list. The groups below are common examples, with the kind of action each might suggest:

Group Typical pattern Possible action
Champions Recent, frequent and high spend Rewards, early access, referral invitations
Loyal customers Frequent buyers with solid spend Upsell and cross-sell offers
New or potential loyalists Recent first or second purchase Onboarding and a second-purchase offer
At risk Once valuable, now lapsed Win-back campaign, personal outreach
Hibernating Low on all three measures Low-cost reactivation, or stop spending on them

These actions are suggestions to test, not guaranteed results. Measure each campaign against a control group where you can.

A worked example

This example is fictional. A small online homeware shop uses simple cutoffs for the last 12 months:

  • Recency: 3 = bought within 30 days, 2 = 31 to 90 days, 1 = over 90 days.
  • Frequency: 3 = four or more purchases, 2 = two or three, 1 = one.
  • Monetary value: 3 = $300 or more, 2 = $100 to $299, 1 = under $100.
Customer Days since last purchase Purchases Total spend R F M Code Group
A 12 6 $540 3 3 3 333 Champion
B 45 3 $210 2 2 2 222 Loyal customer
C 8 1 $60 3 1 1 311 New customer
D 150 5 $420 1 3 3 133 At risk
E 200 1 $45 1 1 1 111 Hibernating
F 60 2 $320 2 2 3 223 Loyal customer

Customer D is a useful case. They spent a lot and bought often, but they have not returned for five months. A general offer might miss them. A personal win-back offer, aimed at their earlier purchases, is more likely to fit. Customer E is unlikely to repay a costly campaign, so a low-cost reminder is a sensible use of budget.

With only six customers, the groups are illustrative. A real model needs many more customers and some checks on how stable the groups are over time.

Where RFM falls short

RFM is useful, but it has limits:

  • Frequency is weak for infrequent purchases. A furniture or car buyer may buy once in five years, so frequency tells you little.
  • Monetary value ignores margin. A high-spending customer may buy low-margin products. Consider profit where you can.
  • Monetary value can be skewed. A few very large orders can push a customer into the top group, even if their usual spending is modest.
  • It ignores why people buy. RFM shows what happened, not the reasons behind it. Survey or interview research can add that.
  • Results depend on the time window and cutoffs. Change either, and the groups change too.
  • It works best with clean data. Duplicate customer records, refunds and missing dates can distort the scores.

For these reasons, RFM works best alongside other bases, such as demographic or psychographic data. For how behavioral and other bases fit together, see the types of market segmentation.

Steps to build an RFM model

  1. Pull transaction records with customer ID, order date and order value.
  2. Choose a time window and apply it to every customer.
  3. Calculate R, F and M for each customer.
  4. Score each measure, using quintiles or cutoffs you can explain.
  5. Combine the scores into a code and group the codes into a small number of segments.
  6. Name and profile each segment, then choose one action to test per segment.
  7. Review the model regularly, since customers move between groups over time.

For the broader process of building segments, see how to do a market segmentation analysis.

The complete guide to market segmentation

Types of Market Segmentation: Bases, Levels and Examples
Read the guide

Keep exploring

Guides from other topics

All topics →