Journal · 2 February 2026
When RFM stops being enough
RFM is a filing system. It is not a theory of value. Commerce teams in Thailand still treat the three letters as a finished Customer Lifetime Value Analytics stack, then act surprised when the “high monetary” bin is full of people who only shop when the discount code is loud.
Recency still matters. A buyer who has not appeared in fourteen months is usually not about to fund your next campaign. Frequency still matters if the category is genuinely habitual. Monetary still matters if it is contribution, not gross merchandise value. The failure starts when those three ranks are averaged into a single “VIP” and handed to media as if they described a future.
Discount as the hidden fourth letter
Once a brand’s repeat is mostly promo-led, frequency becomes a count of how often you trained the customer to wait. We ask flagship students to rebuild RFM with a discount-depth flag on the first and last order. In several Bangkok fashion extracts, the top monetary quintile collapsed after that flag. The people left were quieter, less decorated on the dashboard, and much harder to replace with lookalikes.
What to do instead of throwing RFM away
Keep recency as a hygiene filter. Replace the composite score with a ranking that includes expected repeat probability and contribution. If you cannot estimate probability yet, at least stop calling the composite “CLV.” Call it a recency-weighted spend rank, and refuse to feed it to a bidder that thinks it is a forecast.
The module titled “RFM as a leftover, not a religion” in Predictive CLV for Commerce Teams is where we practise that sentence until it is boring. Boring is the point.