RFM segmentation scores every customer on three axes — how recently they bought, how frequently they buy, and how much monetary value they represent — and groups them by the combination. It comes from direct mail, predates email entirely, and remains the most reliable way to decide who deserves a different message.
The reason it survives is that those three variables predict future purchasing better than anything else you can cheaply collect, including demographics and stated preference.
Building the scores
Take your customer table with three columns: date of most recent order, number of orders, total spend. Then split each into five equal groups and score 1 to 5.
- Recency: most recent fifth scores 5, oldest fifth scores 1.
- Frequency: most orders scores 5.
- Monetary: highest total spend scores 5.
Using quintiles of your own data rather than fixed thresholds is what makes this portable. "Bought in the last 30 days" means something different for a coffee subscription than for a mattress company; "in the most recent fifth of my customers" means the same thing everywhere.
Every customer now has a three-digit code from 111 to 555. That is 125 combinations, which is far too many to write email for.
The seven groups worth naming
Collapse the 125 into groups you can actually act on.
| Segment | Pattern | What to send |
|---|---|---|
| Champions | R 4–5, F 4–5, M 4–5 | Early access, referral asks, genuine thanks |
| Loyal | R 3–5, F 3–5 | Replenishment, complementary products |
| Promising | R 4–5, F 1–2 | Second-purchase nudge, category education |
| Need attention | R 3, F 3–4 | A reason to come back now |
| At risk | R 2, F 3–5 | Win-back with something specific |
| Hibernating | R 1–2, F 1–2 | One re-engagement attempt, then suppress |
| Lost high-value | R 1, M 4–5 | A personal email from a person |
That last row is the one people miss. A customer who spent a great deal and has not returned in two years is worth an individually written email, not a template. There are usually few enough of them that a founder can write them personally, and the response rate justifies it.
What to actually do with it
Promising is where the money is. The single-purchase customer who bought recently is the most convertible group most businesses have, and the gap between one purchase and two is the largest behavioural step in the whole lifecycle. Most companies send this group the same newsletter as everyone else. A dedicated second-purchase sequence — what to buy next and why, based on what they already bought — reliably outperforms almost every other campaign.
At risk is where the urgency is. These are people who used to buy often and have gone quiet. They still recognise your name, which is a wasting asset. A win-back campaign aimed here converts far better than one aimed at hibernating customers, who have largely forgotten you.
Hibernating is where the discipline is. One attempt, then suppress. Continuing to mail people who have not engaged in eighteen months is the main way senders damage their inbox placement, and it costs you delivery for the customers who do still care.
Champions should not get discounts. They already buy at full price. Discounting to your best customers converts revenue you were going to earn anyway into margin you gave away. Give them access, information and acknowledgement instead.
Keeping it current
RFM is a snapshot, and a customer's segment changes with every order and every quiet month. Recalculate monthly at minimum; weekly if your purchase cycle is short.
The implementation that works is to store the three scores as custom fields on the subscriber record and rebuild them on a schedule, so your segments are queries against current fields rather than static lists that decay. Any platform with custom fields and dynamic segments supports this — in IGSendMail you can push updated scores through the API or the MCP server as part of a nightly job.
When RFM does not apply
If you sell one expensive thing once — a mattress, a course, a car — frequency carries little information and the model degrades to recency plus value. If you are B2B with a long cycle and few customers, you do not need statistics to know who your best accounts are.
For subscription businesses, usage frequency is a better middle variable than purchase frequency, since purchases are automatic and tell you nothing. Substitute it and the rest of the method holds.
For everyone else, RFM pairs naturally with the broader approach in email segmentation strategies and lifecycle email marketing.
Frequently asked questions
How often should I recalculate RFM scores?
Monthly for most businesses, weekly if the typical repeat purchase interval is under a month. The scores are only useful while they are current — a segment built three months ago is describing customers who have since moved.
Should I use quintiles or fixed thresholds?
Quintiles, because they adapt to your own purchase cycle and to seasonality. Fixed thresholds mean your segment sizes swing wildly between your busy and quiet periods, which makes campaign results incomparable across months.
Does RFM work for subscription businesses?
Partly. Recency and monetary value transfer directly. Replace purchase frequency with a usage frequency measure, since subscription renewals are automatic and carry no signal about intent.
What is the most valuable RFM segment to email?
Recent single-purchase customers. They have the highest conversion potential per email because the step from one purchase to two is the biggest behavioural change in the lifecycle, and most businesses do nothing specific to encourage it.
