A Loyalty Program That Rewards the Wrong Customers
A mid-market grocery chain launches a points loyalty program with a 3% effective discount on all purchases. After a year, program members are 55% of transactions but the finance team finds the program's cost ($4.2M) exceeds the incremental margin it can attribute ($2.6M). Analysis shows most members were already loyal; the program mostly discounts purchases they would have made anyway. A small group of lapsed customers did return, but they were not the bulk of redemptions. The CMO wants to keep the program to avoid a PR backlash and because competitors have one. The CFO wants it cut. Basket size among members is flat versus non-members.
1. Diagnose why the program is underwater and distinguish the strategic question from the accounting question.
Read the worked answer
The program pays a 3% discount to everyone in it, but most members were already buying. That is a subsidy on existing behavior - a transfer, not an incentive. Incremental margin ($2.6M) is less than program cost ($4.2M), a $1.6M gap, because the discount is applied to purchases that would have happened without it. Flat basket size confirms the program isn't changing behavior; it's just taxing the base.
The accounting question is whether the program is net positive - on these numbers, no, by $1.6M. The strategic question is different: does the chain need a loyalty mechanism to compete, and can it be redesigned to pay only for changed behavior? Those are separable. Cutting it addresses the accounting; redesign addresses the strategy. Conflating them (keep it because competitors have one, or cut it because the math is negative) skips the real work.
2. Recommend a redesign that targets incremental behavior, and specify how to measure whether it works.
Read the worked answer
Shift rewards from 'everyone gets 3%' to behavior the chain wants to create:
- Target lapsed and at-risk customers with reactivation offers rather than discounting the already-loyal.
- Reward basket-building (thresholds, category trials) so the incentive changes what people buy, not just what they pay.
- Cap or tier benefits so the discount isn't a blanket transfer on baseline spend.
Measurement: run a controlled test - treated versus matched control - and measure incremental margin net of reward cost, not redemption volume. Track reactivation rate, basket lift among targeted segments, and whether the previously-loyal base still buys without the blanket discount. Kill or scale based on incremental margin per dollar of reward, and set a threshold (e.g., program must return more than $1 in incremental margin per $1 of cost) before full rollout. Keep a minimal program if competitive parity matters, but pay for behavior change, not for loyalty that already existed.