RFM Segmentation: Who to Say What To, and When
An RFM segmentation guide: recency, frequency and monetary scores, segment definitions, the right action per segment and measurement.
You have a list of 20,000 people and the same campaign goes to all of them. The outcome is predictable: low conversion, rising opt-outs and, over time, a channel nobody opens. Yet inside that list are groups behaving completely differently — and separating them requires no new data collection.
Three numbers are enough
RFM derives three values from order history:
R — Recency
Time since the last purchase. The strongest single indicator: someone who bought recently is the most likely to buy again.
F — Frequency
Total number of orders. It shows habit and attachment.
M — Monetary
Total spend. It shows value but misleads alone: one large purchase is not the same as regular buying.
Each customer gets a score of 1–5 on each dimension (the most recent 20% score R=5, for example). The combination produces the segment.
Meaningful segments and their actions
- Champions (high R, high F, high M): your most valuable group. Action: early access, loyalty benefits, review and referral invitations. Do not send discounts — they are already buying.
- Loyal customers (high F, medium R): buying regularly. Action: relevant product recommendations, introducing higher-tier products.
- Potential loyalists (high R, low F): new and promising. Action: encourage the second order. The second order is the critical threshold for loyalty.
- At risk (low R, high F): formerly regular, now quiet. This is the most urgent group — they are both valuable and leaving.
- Dormant (very low R, low F): long gone. Action: one win-back attempt; if there is no reply, remove them from the list.
- New customers (high R, F=1): they have bought once. Action: a welcome, product guidance, a satisfaction question — not a campaign.
The most common mistake: looking only at M
Most businesses produce a “highest spending customers” list and stop there. But that list also contains people who made one large purchase two years ago and never returned. Monetary value describes the past; recency describes the future. R is the variable that best predicts whether someone will buy again.
Practical notes on setup
- Set thresholds from your own data. “Recent” is two weeks in groceries and two years in furniture. Industry averages do not transfer.
- Keep the segments few. Defining ten segments and writing actions for five is no better than not segmenting at all.
- Refresh regularly. Segments are not fixed; customers move between them with every order. Monthly updates suit most businesses.
- Define the action with the segment. A segment without an action is only a report.
Measurement
- Segment transitions: how many moved from “at risk” back to “loyal”? That is the real return on intervention.
- Conversion by segment: how did the same campaign perform across different segments?
- Second order rate: how many new customers bought again? The most critical threshold for growth.
- Opt-out rate by segment: if messages to dormant customers produce high opt-outs, leaving that group alone is more profitable.
Frequently asked questions
How many customers before it becomes meaningful?
Trends appear with even a few hundred orders. RFM is not a complex model; it answers “who is leaving and who is new” on small data too.
Does it work in B2B?
Yes, though frequency reads differently. Order intervals are longer in B2B; set your R thresholds against your sector’s natural ordering cycle.
Is segmentation the same as personalisation?
No. Segmentation works at group level and is easy to implement; personalisation works at individual level. For most businesses, good segmentation earns more than poor personalisation.
Next step: to see segments built from order data, request a demo.