What is RFM analysis?

How scoring works

Each customer gets three scores, typically 1 to 5 by quintile: the most recent 20% of customers get R=5, the least recent R=1, and the same for frequency and spend. A customer who bought last week, has ordered eight times and spends in the top fifth is 5-5-5.

The combinations are then grouped into named segments. Typical ones:

• Champions — high on all three. Reward them; ask for reviews.
• Loyal — buy often, recently. Offer early access.
• Potential loyalists and promising — recent, not yet frequent. Encourage the second and third order.
• At risk and cannot lose — used to buy a lot, not lately. Win them back first.
• About to sleep and hibernating — low recency and frequency. Cheap reminders only.

Why RFM works

RFM needs only order data — no surveys, no modelling — and recency in particular is a strong predictor of whether someone will buy again. It turns "email all customers" into "send this message to these 400 people", and it shows movement: the number of champions sliding into "at risk" is an early warning.

RFM in TapCub

The commerce report scores recency, frequency and monetary value from 1 to 5 by quintile, using the orders built from your purchase event, and lists 11 segments with their change compared with 30 days ago. It also shows the top transitions over the last 30 days — for example how many champions became at risk — which is often more useful than the snapshot.

RFM rules are also available when defining a segment, so an RFM group can be reused as an analysis cohort, a monitor target or an export.

Common mistakes

• Applying it to non-transactional products. RFM needs repeated purchases; for a one-time purchase it says little.
• Static snapshots. Scores should be recomputed regularly; last quarter's champions may be at risk now.
• Same message to every segment. The point is to treat them differently.

Related guides: Customer Segmentation Analysis.

See also

• How do I create user segments?
• Run an RFM analysis