RFM Segmentation: A Leaderboard, Not a Reading of the Customer

RFM segmentation scores each customer from 1 to 5 on recency, frequency and monetary value, then groups similar scores into named segments. The scores are quintile ranks: each one says where a customer stands against the rest of the list, not what that customer is likely to do next. RFM is a leaderboard, and a leaderboard is not a decision.

Relvino replaces segments with a decision per shopper, so read this as informed but interested. The two guides that rank first and second for the term are quoted directly, and the point this piece makes is one the guides half-make themselves: the hard part of RFM is not the arithmetic.

What RFM segmentation is, in the guides’ words

The Churned glossary, which ranks first, keeps it short: “RFM segmentation is a strategy for grouping customers based on their specific buying behavior.” Omniconvert’s guide, which ranks second and was updated in July 2026, spells out the three signals: “RFM segmentation groups customers by their purchase behavior using three signals: recency (how recently they bought), frequency (how often they buy), and monetary value (how much they spend).” Its quick answer adds the scoring and the segments: customers are scored “on a 1 to 5 scale, then sorting customers with similar scores into named segments such as Champions, Loyal Customers, At Risk, and Lost.”

Omniconvert also says which of the three signals carries the weight: recency “is the strongest predictor of whether someone will buy again.” That is worth holding onto, because recency is also the signal that changes fastest.

The RFM formula is a ranking

The People-also-ask box for this search asks for the RFM formula. There is no equation. There is a ranking. Omniconvert’s step two: “Rank customers on each dimension and split them into fifths. The top 20 percent for recency score a 5, the next fifth a 4, and so on down to 1. Do the same for frequency and monetary.” Churned describes the same step: “Rank customers on all three factors, typically on a scale of 1 to 5, where 5 means the most recent purchase, the highest frequency, or the highest spend.”

Then comes the sentence that makes the method what it is. Omniconvert: “Ranking into fifths keeps the scores relative to your own base rather than to arbitrary thresholds.” A recency score of 5 does not mean a customer bought this week. It means the customer is in the top fifth of this list by last-order date. On a furniture brand’s list, where most people bought once, a customer who ordered forty days ago can be a 5. On a coffee subscription’s list, the same forty days is a 2. Churned makes the same point from the other direction, for teams that use thresholds instead of fifths: “Set your own thresholds and segment names based on what fits your business, since a frequent buyer for a coffee subscription looks different than a frequent buyer for a furniture company.”

Both guides are describing a leaderboard. Each score is a position among the brand’s own customers, set either by the rest of the list (fifths) or by a threshold a person picked. The position changes when other customers change. A customer who has done nothing new can slide from a recency 5 to a 4 because a few hundred other people ordered this week. Nothing about that customer moved. Their rank did.

Why a leaderboard is useful

A leaderboard is a good instrument for the questions it was built for, and the guides are honest about which those are. Omniconvert frames the distribution as a health check: a base “thick with Champions and Loyal Customers and a thin lost tail” is a retention success story, while a base where most customers sit in At Risk, Hibernating and Lost is “a warning that acquisition is masking a leaky bucket.” That is a reading of the list as a whole, and RFM is good at it.

It also guards against the most expensive habit in lifecycle marketing. On the Champions segment, Omniconvert writes: “These customers already love you, so the mistake is discounting them, which trains your most profitable buyers to wait for a deal.” A brand that sends one newsletter to everyone has no way to know who it is training. A brand with a leaderboard at least knows who sits at the top.

And the inputs are free. Every store already has each customer’s last order date, order count and total spend. No survey, no new tracking. For a team deciding how to spend one afternoon on segmentation, RFM is the right afternoon.

What a leaderboard cannot tell you

Three limits follow from the construction, and the guides name two of them.

It is relative, so it is not a forecast. A 5-5-5 says “best among yours.” It does not say whether this customer is ready to order this week, or whether a message now would be welcome or an annoyance. Two customers with identical scores can be in opposite moments: one just received a delivery and is weeks from needing anything, the other is browsing the site tonight. RFM cannot see the difference, because the inputs are three totals and a date.

It ages, by design. Omniconvert’s FAQ: “RFM segments should be refreshed regularly, because they are a snapshot of behavior that ages quickly.” Its own recommended cadence: “Re-score at least monthly, because customers move between segments as their behavior changes, and stale segments quietly send the wrong message to the wrong people.” And the section that sells its software says the quiet part plainly: “The hard part of RFM segmentation is not the math but keeping it live: a spreadsheet is out of date the day you finish it, and segments only work if they reflect this week’s behavior.” Churned agrees in its closing tip: “A buyer who looks incredibly healthy today might stop engaging next month, and a manual spreadsheet will miss that shift.”

Omniconvert goes further and names the cost: “A spreadsheet you score at month-end has already missed the At Risk customers who slipped mid-month, and by the time you notice, the win-back window has closed. That lag is the real enemy of RFM.” The guide’s answer is to re-score continuously. That fixes the lag. It does not change what the score is.

The segment is the unit of action. Even a continuously re-scored leaderboard ends in the same place: nine named segments, each with “one priority action,” each getting its own campaign. New Customers, listed in the guide’s table at a rough 8 to 12 percent of a base, get an onboarding sequence. At Risk, 10 to 15 percent, get a reactivation offer. The campaign is written once for the segment, and every shopper in the segment gets it. A shopper, meanwhile, is one person in one moment. The segment is the compromise a team makes when it cannot decide one person at a time, and that compromise is the whole product.

Where RFM sits among the four types

The same search box asks about the four types of segmentation: demographic, geographic, psychographic and behavioral. RFM is the purest behavioral method there is. It uses nothing about who a customer is, only three facts about what they did, and all three carry a date. That is why it ages faster than any demographic segment and why the guides spend their best paragraphs on refresh cadence. The behavioral segmentation guide takes that argument further: a segment built from events expires at the next event, whatever the refresh schedule says.

From a score to a decision

The alternative to a better leaderboard is no leaderboard. Relvino does not rank a brand’s customers against each other. For each shopper, it decides whether to act at all, on which channel (email, SMS/MMS or a pop-up), with what offer, and when, as one decision, from that shopper’s own live store events: visits, carts, checkouts, orders and the gaps between them. There is no threshold to set and no fifths to recompute, because there is no group. Each decision is made in under 80ms and its outcome feeds the next, so a shopper whose pattern changes is not waiting for month-end.

The case RFM handles worst is the one every guide files under New Customers: a first order, no frequency, no history to rank. Relvino starts that shopper from a prior learned by a Large Retail Model trained across 10K+ retailers and 7M+ data points, which knows what a first order in that category tends to lead to, and then lets the shopper’s own behavior take over. The bucket is replaced by a starting point.

The leaderboard’s best lesson survives the change. Omniconvert’s warning about training good customers to wait for discounts is exactly the decision a per-shopper system makes explicitly: for a shopper who would have ordered anyway, the right action is often no message and no code, and declining to send is recorded as a decision rather than left to a timer. The abandoned cart recovery software guide makes the same point about discounts on a clock.

RFM scoring and per-shopper decisioning, side by side

In the list below, the RFM column is assembled from the two guides quoted above; quoted phrases are theirs and the rest is a plain reading of the method. The decisioning column describes Relvino.

• Unit of action · RFM segmentation: A named segment with “one priority action” · Per-shopper decisioning: One shopper, one moment
• What a score means · RFM segmentation: A rank within the brand’s own list (“relative to your own base”) or a threshold a person set · Per-shopper decisioning: An estimate for this shopper from this shopper’s events plus a category prior
• Inputs · RFM segmentation: Last order date, order count, total spend · Per-shopper decisioning: Every store event: visits, carts, checkouts, orders and the gaps between them
• Refresh · RFM segmentation: “Re-score at least monthly”; continuous in the vendor’s software · Per-shopper decisioning: Continuous; each decision in under 80ms
• Output · RFM segmentation: A segment membership and the campaign written for that segment · Per-shopper decisioning: A decision: act or not, channel, offer, timing
• New customer · RFM segmentation: The New Customers bucket, “High R, low F (first order)” · Per-shopper decisioning: A prior from a Large Retail Model trained across 10K+ retailers, then the shopper’s own behavior
• When it is wrong · RFM segmentation: “stale segments quietly send the wrong message to the wrong people” · Per-shopper decisioning: The decision is scored by its outcome and the next one adjusts
• Example · RFM segmentation: Omniconvert’s guide, Churned’s glossary method · Per-shopper decisioning: Relvino

For a brand that runs RFM today, the useful exercise is to pick one customer who changed segments this month and check whether anything that customer did caused the move, or whether the list moved around them. For a brand weighing whether segmentation is the right layer at all, the Klaviyo vs Relvino comparison sets up a 14-day pilot beside the current setup, and pricing is public. The lifecycle marketing guide covers the stage model that RFM segments usually feed, retention marketing covers the second order that every RFM segment is ultimately about, and win-back email examples covers the segment, lapsed customers, where the lag the guides describe costs the most.

Frequently asked questions

What does RFM stand for?

Recency, frequency and monetary value. Recency is how long since a customer’s last order, frequency is how many orders they have placed, and monetary value is how much they have spent. Each is scored, usually from 1 to 5, and the three scores together describe where the customer sits in the list.

What is the RFM formula?

There is no single equation. The standard method splits customers into fifths on each of the three dimensions, the mechanics described above, and reads the three resulting digits together: 5-5-5 for the best customers, 1-1-1 for the long lapsed. Some teams use fixed thresholds instead of fifths. Either way the score is relative to the list or to a threshold a person chose, not an estimate of what one customer will do next.

What are RFM models?

An RFM model is the scoring scheme plus the map from scores to named segments: Champions, Loyal Customers, Potential Loyalists, New Customers, Need Attention, At Risk, Can’t Lose Them, Hibernating and Lost in the version most guides use, each with a typical score pattern and one priority action. Vendors publish their own variants with more or fewer segments. All of them share the same unit of action, which is the segment, and the same refresh problem, which is that the scores describe the list at the moment they were computed.

What are the four types of segmentation?

Demographic, geographic, psychographic and behavioral. RFM is a behavioral method: it uses what a customer did rather than who they are.

What is AI customer segmentation?

Software that computes the segments instead of a person, usually by clustering customers on more variables than three and re-scoring more often than monthly. It removes the spreadsheet and the manual thresholds. It keeps the segment as the unit: a group gets a campaign. A per-shopper decision system goes one step further and removes the group, deciding for each shopper whether to act at all, on which channel, with what offer and when, from that shopper’s own events.

How long does it take to migrate from RFM segments and flows to Relvino?

About 30 minutes for the technical cutover: connect the Shopify store, point the sending domain, connect SMS, and order and browse history ingest automatically. No segments are rebuilt and no scores are maintained, because each shopper is handled by a decision made in under 80ms from their own behavior, inside guardrails the brand sets once such as margin floors and quiet hours. Revenue is then proven in a 14-day pilot run beside the current setup.