RFM Analysis

RFM · Recency Frequency Monetary · RFM Segmentation

Definition

RFM analysis is a method of segmenting a customer base by the recency, frequency and monetary value of purchases, used to personalize marketing and manage the customer lifecycle.

What RFM is and how it works

RFM analysis is one of the most practical customer-base segmentation tools in e-commerce. The method
groups buyers along three dimensions of their purchase behaviour:

R (Recency) — how long ago the last purchase was. More recent means a higher score.
F (Frequency) — how many times they bought over the analysis period. More orders, higher score.
M (Monetary) — how much they spent in total. More spend, higher score.

Every customer is assigned a score from 1 to 5 on each parameter, producing a three-digit RFM code
such as 5-4-2. It is a simple structure, but a powerful one for reading the state of a customer base
without any ML models.

The key RFM segments and what to do with them

Segment	RFM profile	Description	Action
Champions	5-5-5	Bought recently, often and a lot	Retain, VIP access, do not spam
Loyal	4-4-X or 4-5-X	Buy regularly	Grow AOV, cross-sell
Potential loyalists	5-1-X	Bought recently, but only a few times	Push for the second purchase
At risk	3-3-X with a falling R	Used to buy actively	Reactivate urgently
Cannot lose them	1-5-5	Spent a lot before, have not returned	Win-back with a strong offer
Lost	1-1-X	No activity for a long time, low spend	Minimal effort or ignore

Building RFM step by step

1. Fix the calculation date (today) and the analysis window (12 months)
2. Compute R, F and M for every customer over that window
3. Split customers into quintiles (20% buckets) on each parameter
4. Assign scores of 1–5 on each axis
5. Group similar RFM codes into named segments

Tip: do not try to work with all 125 combinations. Six to ten consolidated segments with clear
names and a defined strategy for each are more useful than a mathematically precise table nobody
can act on.

The limits of RFM

RFM is a retrospective method. A high score today does not guarantee loyalty tomorrow: a customer may
have switched to a competitor yesterday, and RFM will keep calling them a champion for several more
weeks.

For churn forecasting, RFM is usually complemented with:

Churn prediction models — machine learning on a wider feature set
Trend analysis — not just the current R, but its direction (is a champion’s frequency falling?)
Category preferences — the affinity profile shows which category a customer actually lives in, which matters for personalization beyond segmentation

RFM and onsite personalization

RFM segments are the basis for targeting personalized content on the site. The logic: loyal
customers see new arrivals and cross-sell; at-risk customers see an abandoned-cart reminder or a
personal offer; potential loyalists see a frequently-bought-together block tied to their last order.
It lets you address each segment with a relevant message without paying for an outbound delivery
channel.

Frequently asked questions

How is RFM segmentation built technically?

Each customer scores from 1 (worst) to 5 (best) on each of the three parameters: the more recent the last purchase, the higher the Recency score; the more orders, the higher the Frequency score; the more spent, the higher the Monetary score. The three scores combine into an RFM code such as 5-4-3, which gives 125 unique combinations that are then grouped into five to ten named segments.

What should you do with each RFM segment?

Champions (5-5-5) — retain them, do not bury them in promotions, give them early access. At risk (4-4-X with a falling R) — reactivate urgently. Lost (1-X-X) — either a win-back with a strong offer, or no budget at all. New (5-1-1) — push for the second purchase.

What is the main limitation of RFM?

RFM is retrospective: it describes past behaviour and does not predict the future directly. A customer with a high RFM score may have switched to a competitor yesterday. For predictive segmentation, RFM is complemented with churn prediction or purchase propensity models.

Does RFM need to be recalculated in real time?

A monthly recalculation is enough for most purposes. Triggered scenarios need a faster signal — the fact of an add to cart or a first purchase — rather than a recalculation of the full RFM score.

Gravity Field

RFM Analysis in Gravity Field

Gravity Field Audience Segmentation builds behavioural segments, RFM logic included, to personalize onsite content and target campaigns by customer lifecycle stage.

✦Behavioural segmentation on recency, frequency and monetary value of purchases

✦Dynamic segments — customers move between groups automatically as their behaviour changes

✦Affinity profile — RFM extended with category and brand preferences

✦Onsite personalization targeted at RFM segments

Explore the platform →