RFM Analysis: Recency, Frequency, Monetary Segmentation Explained

By Harry Nguyen · Updated Oct 12, 2026

RFM analysis scores every customer on three behaviors taken from order history: recency (how recently they bought), frequency (how often they buy), and monetary value (how much they spend). Each score runs from 1 to 5, and the three digits combine into a code that places the customer in a named segment, such as Champions or At Risk, with a matching action.

Key Facts: RFM Analysis at a Glance

• RFM groups customers using three inputs: last purchase date, number of purchases, and total spend. (Braze)
• The common method ranks customers on each metric and splits them into five equal groups, quintiles, where the top 20% score 5. (Braze)
• Scoring each metric from 1 to 5 gives 125 possible score combinations (5 x 5 x 5), which most teams collapse into a smaller set of named segments. (Bloomreach)
• Bloomreach recommends 12 to 24 months or more of purchase data, and says its segment table is a starting point to adjust for your own customers. (Bloomreach)
• Segment names are a convention, not a standard. Putler lists 11 segments, while Oracle's model uses 6 personas assigned from recency and frequency alone. (Putler; Oracle)

What RFM Analysis Actually Does

Most customer lists are treated as one lump. The person who bought last week and the person who bought once three years ago get the same newsletter, the same discount, the same everything. RFM breaks the lump apart using only data you almost certainly already have in your store or billing system.

The three questions are simple:

• Recency: how many days since the customer's last purchase? Fewer days is better.
• Frequency: how many orders did they place in the chosen window?
• Monetary: how much did they spend in that window?

Those definitions match the way analytics vendors describe the method (Putler; Oracle). The logic behind it is plain. Someone who bought recently, buys often and spends a lot is the most likely to buy again. Someone who has gone quiet is the least likely, unless you give them a reason.

On where it came from: one vendor guide says catalog marketers used it to cut printing and shipping costs by mailing only the customers most likely to respond (Putler). The exact originator is hard to pin down, so treat any "invented by" claim with caution. What matters for you is that it needs very little data.

If you want the broader picture of how RFM fits among segmentation approaches, start with customer segmentation for e-commerce. This article goes deeper on the RFM method itself.

RFM vs Other Ways to Understand Customers

RFM is often confused with its neighbors. They answer different questions, and they work better together than apart.

A useful way to separate them: RFM ranks the customers you have today, cohort analysis shows whether each new batch of customers behaves better or worse than the last, and customer lifetime value estimates what a customer is worth over time. Demographics describe who someone is, while RFM describes what they did, which is usually the stronger signal for the next purchase.

How to Run an RFM Analysis, Step by Step

You can do this in a spreadsheet in an afternoon.

1. Pull one row per customer. You need a customer ID, the date of the last order, the number of orders, and total spend, all over the same window. Bloomreach suggests 12 to 24 months or more (Bloomreach). Pick the window to fit your buying cycle: a grocery brand and a furniture brand shouldn't share one.
2. Turn the last order date into recency. Subtract it from today's date to get days since last purchase.
3. Score each metric from 1 to 5. Rank customers on each metric and split them into five equal groups. For frequency and monetary, the top fifth gets a 5. Recency runs the other way: the most recent buyers get the 5 (Braze).
4. Handle lopsided frequency. Most customers buy once or twice, so equal groups can break down. Bloomreach notes that frequency quantiles can be uneven and suggests manual bands such as 1, 2, 3, 4 and 5 or more orders in that case (Bloomreach).
5. Combine the scores into a code. Write them in R-F-M order, so a customer with recency 5, frequency 4 and monetary 5 is "545" (Braze).
6. Map codes to named segments. 125 codes are too many to run campaigns on. Group them into 6 to 11 segments, each with a rule on the three scores.
7. Assign an action and an owner to every segment. A segment with no action is just a label.
8. Recalculate on a schedule. Customers move between segments. Bloomreach says fixed thresholds may need monthly readjustment (Bloomreach).

Two choices matter more than they look. First, use either percentile cutoffs or fixed thresholds (such as "ordered within 30 days"), not a mix. Percentiles adapt to your data but shift as it changes. Fixed thresholds are stable but go stale. Second, skewed spend can distort monetary scores, because a handful of very large buyers stretch the scale. Quintiles handle this reasonably well because they rank rather than measure distance.

Ready-to-Copy RFM Segment Template

This is a starting set of segments with score patterns and suggested actions. Rules are read top to bottom, and a customer takes the first segment they match. Adjust the cutoffs for your business. Segment names vary by vendor, so rename them to whatever your team will actually remember.

Add a one-line owner and a send limit to each row. For example, Champions might get at most two messages a month, and Lost customers one reactivation email before they move to a quieter list.

A Worked Example (Illustrative Numbers)

Here are ten customers with made-up figures, to show the mechanics. They are not benchmarks. With ten customers, each quintile holds two people. Real lists need thousands for the scores to be stable.

Notice what the raw numbers hide. Customer I spent the most of anyone ($1,500) yet scores a 1 on recency. A report sorted by revenue would call I your best customer. RFM flags I as a Can't lose them customer who hasn't ordered in 240 days, which calls for a phone call, not a newsletter. Meanwhile customer B has spent little but ordered 12 days ago, so B is worth a nudge toward a second purchase.

Ties are a real issue. Customers G and J both ordered once, so their frequency scores are the same even though a strict ranking would split them. Decide a tie-break rule (for instance, rank ties by recency) and apply it consistently.

Examples by Business Function

The same scores serve different teams. These scenarios are illustrative.

For subscription businesses, "frequency" often needs a different definition, because a customer on an annual plan only has one purchase a year. Use logins, usage events, or renewals instead. If the product has a clear churn problem, pair this with churn analysis.

Turning Segments Into Action

A segment earns its keep only if it changes what you do. Three plays cover most of the value:

• Protect the top. Champions and Loyal customers are your most reliable revenue. Don't train them to wait for discounts. Offer access, recognition and referrals instead.
• Rescue the valuable ones who are slipping. At risk and Can't lose them are customers who have already shown they'll pay. A targeted win-back campaign here usually has a better chance than the same effort aimed at Lost customers.
• Climb the middle. New customers, Promising and Potential loyalists are where a second or third order changes their trajectory. Think of it as moving them up a value ladder.

Finally, track movement. If customers never migrate between segments, either your thresholds are off or your campaigns aren't doing anything. A simple monthly table of how many customers moved from each segment to each other segment tells you more than any single snapshot, and it sits naturally beside your growth metrics hierarchy.

Limits of RFM

RFM is quick and useful, but it has edges you should know.

• It describes the past. A high score says someone bought a lot, not that they will. Forward-looking models, such as CLV, add the projection.
• It ignores why. Two customers with the same code may differ in product, margin, channel or reason for buying.
• Monetary value isn't margin. A big spender on heavily discounted goods may be less valuable than the score suggests.
• It needs repeat behavior. It works best where customers buy more than once. For one-off purchases like a mattress, it says little.
• Segment definitions are subjective. Even vendors disagree. Bloomreach's own table assigns some of the same score combinations to two segments and relies on engagement signals to tell them apart (Bloomreach). Oracle's model leaves monetary value out of its persona assignment, so a high spender with low recency can still land in a low-value persona (Oracle).
• Small lists are noisy. With few customers, a single order can move someone two segments.

Braze makes a similar point: RFM is easier to apply and scale for value-based targeting, but it is not better in every case, and it captures behavior rather than attitudes (Braze).

RFM Dos and Don'ts

Do

• Use one consistent window and one scoring method. Mixing them makes scores incomparable over time.
• Invert recency. The most recent buyers must get the highest score.
• Name segments in plain language your team will use. "Can't lose them" beats "Segment 7".
• Attach a specific action and a send limit to every segment.
• Refresh the scores on a schedule and watch segment movement.
• Test. Hold out a control group in a segment to see whether your action actually lifts orders.

Don't

• Don't offer the same discount to every segment. That wastes margin on people who would have bought anyway.
• Don't treat the labels as facts. They are shorthand for a score pattern.
• Don't build 30 segments. If two segments get the same email, merge them.
• Don't run it on tiny lists and expect stable results.
• Don't use RFM alone to set budgets. Combine it with margin and lifetime value.
• Don't forget to exclude refunds, test orders and internal accounts before scoring.