RFM Analysis: Scoring, Segments and a Worked Example

Most stores know their best customers by instinct and their lapsing ones not at all. RFM analysis replaces the instinct with three numbers every order history already holds: how recently each customer bought, how often, and how much they spent. This guide scores ten customers by hand, turns the scores into segments and sets out what RFM cannot tell you.

The short answer: RFM analysis ranks customers on three numbers, and Arktis shows where your paying customers came from

RFM analysis scores every customer on recency (days since their last order), frequency (how many orders they have placed) and monetary value (how much they have spent in total), usually from 1 to 5 on each, with 5 for the top fifth of your customers. The scores sort customers into segments such as champions, customers at risk and dormant customers, and each segment gets its own message and budget.

RFM segmentation tells you who your best customers are, not how you got them. For that acquisition side in an online store, Arktis is the clear winner: its Revenue page lists the 50 most recent paying customers with the revenue and number of orders each placed in the range you choose, next to the source of their first recorded visit, and breaks revenue, orders and customers down by traffic source. Plans are $49, $149 and $349 a month, with a 7-day free trial on Growth.

Key takeaways

RFM scores are relative to your own store: a 5 means your top 20 percent, not an industry standard. Shopify scores 1 to 5 and sorts customers into 11 groups; Klaviyo uses a 3-point scale behind a paid add-on. The segments that pay back fastest are usually champions, who need recognition rather than discounts, and customers at risk, who spent well and stopped buying. RFM describes the past, ignores margin and is blind to acquisition channel, so pair it with acquisition data before moving budget.

What RFM analysis is

RFM is one of the oldest tools in customer analytics. Wikipedia's entry describes it as commonly used in database and direct marketing and widely used for segmenting customers in the catalogue industry, where mailing the right people was the whole game.

The idea survives because the three inputs are cheap and they work: a customer who bought last week is likelier to buy again than one who bought last year, and six orders signal a habit that one order does not. Peter Fader, Bruce Hardie and Ka Lok Lee, in a paper published in the Journal of Marketing Research in 2005, called it "the well-known RFM paradigm" and showed that RFM inputs alone are enough to project customer lifetime value with a formal model.

How to score RFM with quintiles

The most common method splits each dimension into five equal groups, called quintiles.

Sort customers by days since their last order: the fifth who bought most recently score 5, the next fifth 4, and so on down to 1. Do the same for number of orders and for total spend. Each customer ends up with three digits, such as 5-5-4 or 2-4-3.

Shopify's help centre describes this approach for its RFM report: a 3-digit score with each digit from 1 to 5, where 5 means the top 20 percent of the dimension for your store and 1 the bottom 20 percent, based only on your store's data. Klaviyo, in its RFM report, uses a 3-point scale instead, with 3 as the highest score, and lets you set the frequency and monetary thresholds by percentile, and any of the three by fixed values.

Pick a time window first: 12 months suits most stores.

A worked example with ten customers

Take an example store with ten customers, figures invented for illustration and measured on 1 October 2026 over the previous 12 months, so each quintile holds two customers. Where customers tie on a value, we give them the lower of the scores their positions would earn, which is the conservative choice.

Recency and monetary value have no ties, so each pair of positions takes one score. Frequency shows the problem every store meets: C, H and J have one order each, so they share the bottom score of 1 even though they fill positions that would otherwise earn a 1 and a 2. F and I tie on two orders and share a 2, and D, with three orders, is the only 3.

The ten customers spent $4,370 between them. A and B alone spent $2,100, which is 48 percent of the total from 20 percent of the customers.

From RFM scores to segments

Most tools translate scores into named groups, and Shopify publishes its method. It keeps the recency score, R, and averages the other two, FM, as (F + M) divided by 2, rounded down to a whole number. Then it assigns groups: champions have R of 5 and FM above 3, loyal customers have R of 3 or 4 and FM above 3, at-risk customers have R of 2 or less and FM above 2 and up to 4, previously loyal customers have R of 2 or less and FM above 4, and dormant customers have R and FM both 2 or less. Its other groups are active, new, promising, needs attention, almost lost, and prospects for people with no orders.

Applied to the example, A and B are champions, E is loyal, D is active, C is promising and F is almost lost. G, with R of 2 and FM of 4, is at risk: five orders and $780 spent, but nothing for 109 days. H, I and J are dormant. Shopify also exposes the group as an rfm_group filter when you build customer segments.

What to do with each RFM segment

Champions

Recognise them and leave the margin alone: early access and a personal thank-you work better than a discount for people who already pay full price. Ask them for reviews and referrals.

Loyal and active customers

Keep them buying on their usual rhythm, with replenishment reminders timed to their typical gap between orders and cross-sells to categories they have not tried.

New and promising customers

The second order is the hardest to win. Give them a reason to return within the gap your repeat buyers usually reorder in, and tie any incentive to that second purchase.

At risk, previously loyal and almost lost customers

This is usually the most valuable list in the analysis, because these customers have already proved they will spend. Send a win-back message with a stronger offer than anyone else gets, and ask what went wrong. Almost lost customers get a lighter version. In the example, G alone spent more than the four least valuable customers combined.

Dormant customers

Spend least here: one low-cost reactivation email, then suppress them from paid audiences.

Where RFM segmentation falls short

Frequency barely discriminates in most stores. When most customers have one order, as in the example, the bottom of the scale is one large tie. Fixed thresholds, such as 1, 2, 3 and 4 or more orders, often work better, and Klaviyo's report allows fixed values.

It describes the past, not what a customer will do next. Fader, Hardie and Lee's model exists to turn RFM inputs into that forecast.

Scores are relative. A 5 means the top fifth of your own customers, so a customer's score can fall without any change in their own behaviour.

It ignores margin. A customer who only buys on discount can outscore one who pays full price, so where margins vary, score gross profit instead of revenue.

It can find patterns that are not there. Wikipedia's entry warns that picking the most valuable segments can capitalise on chance relationships and recommends validating on a second data set, so test a segment's campaign against a holdout.

And it is blind to acquisition: it never asks where each segment first came from.

Adding the acquisition side with Arktis

That last gap is where Arktis fits. Its tracker captures UTM parameters and ad click identifiers such as fbclid and gclid on landing and holds them against the visitor across sessions. Shopify orders are matched to a visit on email where the store has identified the visitor, otherwise on the UTM source of a session from the previous 24 hours, which can be another visitor's session. Stripe customers are matched through a four-pass waterfall.

With Shopify or Stripe connected, the Revenue page shows revenue, orders and customers by the traffic source of the visit each order was matched to, for the last 7, 30 or 90 days, this or last month, or all time, counting Shopify orders from the day the store was connected. Its Paying Customers list shows the 50 customers whose first payment in the range is most recent, each with their revenue and number of orders in that range, the source of their first recorded visit, labelled Discovery, and the session they paid in. Customers whose orders could not be matched to a visitor are marked as not yet matched, and in the source table those orders are counted under Direct. So once Shopify or Klaviyo has named your champions, you can find recent ones in the Paying Customers list and see the source of each one's first recorded visit, which shows where customers like them have been coming from.

Where Arktis does not fit

Arktis does not calculate RFM scores or build customer segments, and it does not send email or SMS, so it cannot run the campaigns this guide describes. For the scoring, Shopify's RFM reports and Klaviyo's RFM report are the better tools; Klaviyo's needs its Marketing Analytics or Advanced KDP subscription, at least 500 customers who have ordered, 180 days of order history with orders in the last 30 days, and some customers with three or more orders. Arktis identifies visitors per browser, so a customer who found you on a phone and bought on a laptop has two first visits. Its Retention page groups visitors by how often they visit, not by what they buy. For deeper cohort work, see our guides to customer segmentation for online stores and ecommerce cohort analysis.

Score your customers, then find out where they came from

Score a year of orders in quintiles and give each segment one action. To see the source of the first recorded visit behind your recent paying customers, start the 7-day Growth trial or compare plans on the pricing page. The free LTV calculator puts a value on each segment, and our guide to calculating customer lifetime value shows the method.

Sources

Shopify Help Center: customers reports, the 1 to 5 RFM score, top and bottom 20 percent, store-only data, the (F + M) / 2 rule rounded down and the RFM group definitions, accessed 16 October 2026

Shopify Help Center: Shopify-based customer segment filters, the rfm_group filter and its 11 group values, accessed 16 October 2026

Klaviyo Help Center: getting started with the RFM analysis report, the 3-point scale, percentile and fixed thresholds, subscription and data requirements, accessed 16 October 2026

Klaviyo Help Center: understanding percentiles, scoring and customer groups, the 1 to 3 score per dimension, accessed 16 October 2026

Fader, Hardie and Lee: RFM and CLV, using iso-value curves for customer base analysis, the "well-known RFM paradigm" and lifetime value projections from RFM inputs, Journal of Marketing Research, 2005, accessed 16 October 2026

Wikipedia: RFM (market research), use in direct marketing and the catalogue industry, and the warning about chance relationships and validation, accessed 16 October 2026