Last-touch attribution vs. recency-frequency-monetary scoring for retention spend

Last-touch attribution vs. recency-frequency-monetary scoring for retention spend

Attribution models were built for acquisition. Applied to retention and reactivation budgets, they quietly mislead — and an older method often serves better.

The mismatch
Multi-touch and last-click models answer "which channel caused this conversion." For an existing customer's repeat purchase, the dominant driver usually isn't a channel at all — it's the customer's own state: how recently they bought, how often, how much. Attributing a loyal customer's reorder to the email that happened to land that week is the retention-side version of last-click bias.

What RFM does instead
Recency-frequency-monetary scoring segments customers by behavioral state rather than crediting touches. It tells you who is at risk, who is high-value, who is dormant — the inputs to deciding where retention spend should go and to whom.

How they should interact
The error is using attribution to evaluate retention channels in isolation. A reactivation campaign sent to already-active customers will show glorious attributed revenue (they were going to buy anyway). The only honest evaluation holds RFM state constant: did the campaign lift purchase rate within the dormant segment versus a holdout? That's incrementality conditioned on customer state.

Bottom line for practitioners: Don't evaluate retention spend with acquisition attribution models — they'll credit your channels for revenue your loyal customers were already going to deliver. Use RFM to define segments, then measure each retention channel with a within-segment holdout. The question isn't "which touch got credit" but "did contacting this segment change its behavior versus leaving it alone."