RFM segmentation
RFM segmentation scores every customer on recency (days since last order), frequency (orders in a window) and monetary value (spend or margin in that window), usually 1-5 on each axis. Ecommerce operators use the combined score to rank customers, target retention and win-back campaigns, and see where value concentrates.
| R (recency) | Days since the customer's last order. The most recent fifth of the file scores 5 |
| F (frequency) | Orders placed inside the scoring window, typically the trailing 12 months |
| M (monetary) | Total spend in the window - or contribution margin, the margin-true version |
Conventions differ: some stacks concatenate the three digits (555), some sum them (15), some use fixed thresholds instead of quintiles. Pick one convention and keep it stable.
Worked example
Score a file on trailing-12-month quintiles, 1-5 on each axis, then compare three customers (example numbers, flagged as such).
B and C share the same lapsed label with roughly twenty times the margin at stake. The monetary axis does the work - which is why it should score contribution margin, not revenue.
What is a good RFM score?
Quintile scoring is relative by construction: one fifth of your file always scores 5 on each axis, so RFM ranks your customers against each other rather than judging the brand against an external benchmark. There is no universal good score.
What the ranking is worth depends on three choices. The scoring window has to cover your category's repurchase cycle, or recency punishes normal behaviour. The recency scale has to fit that cycle too - 90 days silent means something different for coffee and for furniture. And the monetary axis should score contribution margin, because a discount-hunting high spender can rank above a full-price customer who leaves far more contribution behind. Our free margin LTV calculator shows what a segment is worth in contribution once discounts and costs are counted.
RFM segmentation vs related metrics
| Metric | What it measures | How it differs |
|---|---|---|
| Cohort analysis | How groups of customers behave as they age | Cohorts are fixed at first purchase and tracked over time. RFM groups by current behaviour, so it describes now rather than change. |
| Customer lifecycle | Which state each customer is in | Named states (new, active, at-risk, lapsed) with explicit boundaries and migration between them. RFM is the score cube those states can be derived from. |
| Value concentration | How much value the top customers hold | One distribution-level summary. RFM assigns a score to every individual customer. |
| Customer lifetime value (LTV) | What a customer will be worth | A forward projection of future value. RFM is a backward-looking score built from what already happened. |
Common mistakes
- Scoring M on revenue, so discount-heavy buyers outrank full-price customers who leave more contribution. Score it on margin.
- Copying another brand's thresholds instead of deriving quintiles from your own file and repurchase cycle.
- One recency scale across categories with different repurchase cycles.
- Building the 125-cell cube and acting on none of it. Collapse it into a short list of named segments, each with one owner and one action.
- Scoring once and never refreshing. Recency decays daily; the cube is a snapshot, not a fixture.
Frequently asked questions
In Blufire, S4 Customer Value & Segmentation ships the margin-true RFM cube out of the box - scored on contribution margin, with lifecycle states, migration and CLV alongside - and S8 Persona Analytics turns the segments into audiences.
Updated July 2026