RFM Decay: Why Recency Windows Need a Reset Every Quarter
An RFM score built on a 90-day window goes stale the moment the season changes. The segments it produces describe last quarter's buyers, not this quarter's.
RFM scores decay because recency is a moving window and buying cycles change with the season. Reset the recency thresholds every quarter, re-score the list, and retire segments built on last quarter's windows before they send the wrong offer.
Why RFM goes stale
RFM scores every customer on recency, frequency, and monetary value, usually on a 90-day window. The score is a snapshot, and the snapshot ages fast. A buyer who was a champion in January, three purchases and 400 dollars, is a lapsed buyer by April if they have not returned, but the score still says champion until the window rolls past them. The store keeps sending champion offers to a buyer who has already left.
The season makes it worse. A gift buyer who purchased in December looks like a high-frequency buyer in January and a ghost by March. A swimwear buyer who purchased in June looks lapsed in October. The RFM model does not know about seasons, so it reads seasonal behavior as loyalty or churn, and both readings are wrong.
Related reading on this site: SegmentMuse and RFM Segments That Lie: Recency Without Frequency cover the neighbouring problems in more depth.
The quarterly reset
The fix is a quarterly reset of the recency thresholds. Every quarter, re-derive what recent means from the current buying cycle: the median days between purchases for the last two quarters, not the last two years. A store whose median cycle is 45 days should call 45 days recent, not 90. A store whose median is 120 days should call 120 recent.
Re-score the whole list at the reset, and retire the old segments. A segment built on last quarter's thresholds is a segment built on last quarter's buyers. The reset takes an afternoon, and it is the difference between an RFM model that describes the list and one that describes history. Document the new thresholds in the same place as the priority order, so the next person to run the reset does not have to guess what recent meant last quarter.
What to do with the decayed
The buyers the reset downgrades are the interesting ones. A champion who falls to at-risk is not a lost cause. They are a win-back candidate, and the win-back should reference the actual last purchase, not a generic we miss you. A buyer who falls from frequent to occasional is a candidate for a frequency nudge, a subscription offer, or a bundle that raises the order count.
The mistake is to keep emailing the decayed buyer as if nothing changed. The champion offer sent to a lapsed buyer is the fastest way to train them that the store's mail is irrelevant. The reset is not just a scoring exercise. It is the moment the store admits the relationship changed and adjusts the message to match.
Frequently asked questions
What is RFM segmentation?
It scores customers on recency, frequency, and monetary value: how recently they bought, how often, and how much they spent. The three scores combine into segments like champions and at-risk buyers.
Why does RFM decay?
Because recency is a moving window and buying cycles change with the season. A score built on a 90-day window describes last quarter's buyers, not this quarter's.
How often should I reset RFM thresholds?
Every quarter. Re-derive what recent means from the current median days between purchases, re-score the list, and retire segments built on old windows.
What should I do with buyers RFM downgrades?
Treat the downgrade as a signal. A champion who falls to at-risk is a win-back candidate, and the offer should reference their actual last purchase.