Loyalty-program research: choose a reward worth earning and funding

A popular reward can be an expensive way to thank customers for purchases they would have made anyway. A cheaper reward can be so difficult to earn that nobody changes their routine. Loyalty-program research needs to examine both sides: what makes the program useful to customers and what additional behavior would pay for it.

Start with a specific program decision. Are you trying to bring occasional customers back, retain regular buyers, increase the number of categories they buy or make an existing benefit easier to use? Each objective changes the people to study, the rules to test and the evidence needed after launch.

Enrollment, reward redemption and incremental purchasing are different outcomes. A good research plan keeps them separate from the first concept through the commercial pilot.

Put the earning and redemption rules in the offer

Consider Fable, a fictional neighborhood bookshop deciding whether to introduce a free rewards program. All program terms, costs and quantities below are teaching assumptions, rather than findings from a customer study.

The proposed offer is:

• Join for free and earn an $8 store credit for every $120 of qualifying book purchases after enrollment.
• Qualifying spend is the amount paid for eligible books after discounts, excluding tax and delivery charges.
• The credit can be used on a later book purchase, rather than the purchase that earned it. It expires 60 days after issue.
• The earning balance stays available throughout the six-month pilot. Returns adjust the qualifying balance under the stated program terms.

The decision is whether that offer deserves a limited pilot, or whether the reward, threshold or redemption rule needs to change first.

“Earn rewards on every visit” would leave too much unanswered. Customers need to know which purchases count, how far they are from the reward and what they can do with it. Fable also needs a workable answer when someone returns a book after earning a credit. Research should expose unclear rules before the team promises them.

Test whether the reward fits the shopping routine

Show the complete offer and ask people to explain it using a concrete purchase.

Someone with $90 of qualifying spend buys another $30 of books. What happens now? Under Fable's proposed terms, they earn an $8 credit for a later purchase. They do not receive $8 off the current $30 receipt. That distinction matters to both perceived value and the economics.

Then ask what would make the program useful in the person's actual routine. A weekly visitor, an occasional gift buyer and someone who mainly borrows books may see the same $120 threshold differently. Find out how often they buy, their usual basket and which other bookshops or channels they use. Age alone will not explain whether the reward is reachable.

Explore the reason behind rejection. Is the reward too small, the threshold too distant, the expiry too short or the process too much trouble? Those answers suggest different revisions. Someone who likes the reward but expects to forget the credit raises a redemption problem, rather than evidence that a larger discount is necessary.

Reachability deserves its own check. Wang and colleagues' hotel-loyalty field experiment tracked stays before, during and after a reward promotion and examined goal achievement and failure. Its setting does not determine Fable's threshold, but it gives the team a reason to investigate what happens when customers approach, reach or miss a goal.

Keep current loyalty separate from the program's effect

Fable's most frequent customers may be the first to join. They understand the shop, buy books regularly and expect to earn the credit without changing much. If members later spend more than nonmembers, that difference may partly reflect who enrolled.

Leenheer and colleagues' Dutch supermarket research addresses this self-selection problem explicitly. It examines program membership as a choice, rather than treating the spending gap between members and nonmembers as the program's effect.

Use purchase history to define relevant groups before the test: recent frequent buyers, occasional buyers and eligible customers who have become less active. Keep the definitions and observation periods consistent. If you have only stated purchase habits, retain that evidence label instead of presenting the groups as transaction-based cohorts.

Interviewing current members can explain their experience, but it misses eligible people who did not enroll. Include them when the decision concerns program adoption. The sign-up process, requested information or perceived effort may discourage precisely the customers Fable hoped to bring back.

For the commercial comparison, evaluate the effect of offering the program across the assigned eligible group, including people who never join or redeem. Restricting the analysis to enthusiastic redeemers would answer a selected question. If individual customers can be assigned to an offer and a comparison, preserve that assignment; if the program must be introduced by store or period, design the comparison around that constraint and account for the resulting differences.

Calculate what additional buying would have to cover

Suppose Fable expects 30 credits to be redeemed during a pilot. Assume each $8 redemption reduces contribution by $8 relative to the same basket without the credit, and administering the pilot costs another $120.

The specified cost is 30 × $8 + $120 = $360. If an additional paid order contributes $12 before loyalty credits and those administration costs, the program needs 30 additional orders just to cover that $360. More than 30 would be needed to improve contribution under these assumptions.

Those are additional orders compared with what the eligible customers would otherwise have bought, not 30 orders carrying a membership number. The credits used by existing regular buyers still cost money even when their underlying purchases are unchanged.

The conclusion also depends on redemption. At 50 redemptions, the specified cost becomes $520. At the same $12 contribution per additional order, at least 44 additional orders would cover it. If rewarded baskets have different margins, or returns reduce net receipts, recompute the calculation from those conditions.

This is a break-even exercise, not a forecast that Fable will generate the required orders. It tells the team what the pilot must resolve. Include earned credits that remain available after the measurement window when planning later costs; a short observation period should not make future redemptions disappear.

Watch what happens after the first reward

Set a measurement period that fits the buying routine and the earning cycle. An occasional book buyer may need longer to reveal a change than someone who visits every week. Track purchases before the offer, during earning, at redemption and afterward.

Keep several mechanisms visible. A customer might make the same purchases sooner to reach $120, spend more per visit, add a visit or continue buying for longer. The program can affect these differently. Research on a non-tiered hair-salon loyalty program separates spending per visit, visit frequency and attrition when analyzing customer value. Its reported results belong to that program and data, rather than provide a lift estimate for the bookshop.

For Fable, the useful readout joins purchases and costs with customer explanations. Keep refunds, credit issuance and redemption attached to the relevant account and period. Check whether a purchase moved from another Fable channel or replaced a full-price purchase the customer already intended to make. Report the comparison's uncertainty alongside the estimated contribution difference.

Choose what should move into the pilot

Sapien can help explore how a proposed program is understood, compare reward concepts and investigate reactions across data-grounded simulated buyer groups. That can identify an unclear earning rule, a reward that feels irrelevant or a threshold worth revising. Actual enrollment, repeat purchasing and redemption economics need the corresponding customer and transaction evidence.

Fable should advance an offer customers understand and can plausibly use, with a cost model that shows what extra behavior would make it worthwhile. If buyers mistake the credit for an immediate discount, repair the explanation. If occasional buyers cannot see a realistic route to the reward, reconsider the mechanism. Then use the pilot to learn whether the program changes purchasing enough to justify its cost.