RFM Analysis: How to Score Customers and Segment Them for Action

Not all customers are equally valuable, and the difference rarely shows up in a plain list of names. RFM analysis is a simple, long-established way to see it. It scores every customer on three questions: how recently did they buy, how often do they buy and how much do they spend. The result is a handful of clear groups, from loyal champions to customers on the verge of leaving, each of which deserves a different message. This guide explains RFM, shows how to calculate it with a worked example, lists the segments and what to send each, and covers the limits you should know about.

What is RFM analysis?

RFM stands for recency, frequency and monetary value. It comes from direct marketing, where catalogue sellers needed to decide who deserved the next mailing. The idea still holds: customers who bought recently, buy often and spend more are more likely to respond than customers who did none of these. Because RFM needs only order data, it is one of the easiest ways to start with behavioural customer segmentation, and it sits comfortably beside measures such as customer lifetime value.

How to calculate RFM scores

1. Choose the period. Use a window that fits your buying cycle, such as the last 12 months for fashion or 24 months for durable goods.
2. Pull one row per customer. You need the date of the last order, the number of orders and the total spend in the window.
3. Score each measure from 1 to 5. Rank customers and split them into five equal groups (quintiles). For recency, the most recent group gets 5. For frequency and monetary, the highest group gets 5.
4. Combine the scores. Write them as a three-digit code, such as 555 for the best customers or 211 for those who bought once long ago.
5. Group the codes into segments. There are 125 possible codes, which is too many to act on, so cluster them into eight to ten named segments.
6. Attach a play to each segment. Decide the message, channel and offer for each one.
7. Refresh regularly. Recalculate monthly, because customers move between groups.

A worked example

Take four customers from an online store, measured over the last 12 months.

Asha is a champion. Ravi is a solid regular. Meera used to buy often and spend well, but nothing for over six months, which makes her a high-value customer at risk. Karan bought once and disappeared. Each of the four needs a different message, which is the whole point of the method. The scores here are an illustration of the arithmetic, since the cut-offs for each score depend on your own distribution.

RFM segments and what to send

Using RFM in Flowfox

You do not need a data science team to run this. Orders arrive as events through the data model or FlowConnect. Computed attributes on the contact profile, such as days since last order, order count and total spend, give you the three measures. In the segment builder you can then create rules like “order count is at least 4, days since last order is under 60 and total spend is above Rs 10,000” for champions, and the mirror image for at-risk groups. Each segment can feed a journey with the right messages across email, WhatsApp and SMS. The retention use case shows a ready pattern.

{
  "segment": "At risk",
  "match": "all",
  "rules": [
    { "attribute": "orders_12m", "op": ">=", "value": 4 },
    { "attribute": "days_since_last_order", "op": ">", "value": 120 },
    { "attribute": "days_since_last_order", "op": "<=", "value": 365 },
    { "attribute": "consent.email", "op": "=", "value": true }
  ]
}

What RFM is good for

• Targeting. Send richer offers to valuable segments and fewer messages to the rest.
• Win-back. Find lapsed valuable customers before they are gone for good. See our guide to customer retention strategies and churn rate.
• Budget. Spend acquisition money on audiences that look like your champions.
• Personalisation. Tailor tone and content to the relationship. See marketing personalization.
• Measurement. Track how many customers move up or down between groups each month.

Limits and pitfalls

RFM looks backwards. It describes what customers have done, not why, and it says nothing about those who have not bought yet. It treats a customer who bought a rarely purchased item, such as a sofa, as if they were a lapsed regular, so adjust the window to the product. Monetary value can mislead when a few big orders dominate, and it ignores margin, so a high-spending customer who returns half of what they buy may be less valuable than the score suggests. Finally, it is a simple form of market segmentation, which is excellent as a starting point and less so as the whole strategy. Combine it with product interest, channel preference and lifecycle stage as your data grows.

For a more rigorous look at how groups behave over time, pair RFM with cohort analysis. For larger datasets, some teams replace quintile scoring with clustering, such as the k-means method described in the k-means article, but the simple version is often enough and is easy to explain to colleagues.

Common mistakes

• Too many segments. If you cannot write a different message for each, merge them.
• One-size period. Use a window that suits how often people really buy.
• Discounting champions. They would have bought anyway. Reward, do not subsidise.
• Never refreshing. Customers move. Recompute at least monthly.
• Ignoring consent. A segment is not permission to message on every channel.

RFM will not tell you everything about a customer. It will tell you who to call first.

Next, read about customer lifetime value and abandoned cart recovery, and browse the blog.