A loyalty program works when it rewards future value, not past spend.

Forrester reports that 90% of US online adults belong to at least one loyalty program. EY reports that more than half of all points go unredeemed and that members actively engage with only 18% of the programs they join. BCG found that 35% or more of members plan to cancel, rising above 50% among 18 to 34 year olds. Enrollment is not the problem. Design is.

The rewards do not match where the customer is going

A customer who bought a premium winter coat does not want 5% off socks. They want the spring collection in their style. A program that only sees transactions cannot know that. Meanwhile the strongest signals sit outside the program: Nielsen finds that 88% of consumers trust recommendations from people they know more than any advertising.

Everyone gets the same treatment

The same points per unit spent, the same emails, the same offers. But a customer likely to spend ten times more over the next year should get different treatment from one who will not return: different timing, different channel, different products, a different incentive or none.

The program is an island

Points live in one tool. Email in another. Recommendations in a third. Ads in a fourth. The email does not know the points balance. The site does not know who is close to a voucher. Nobody coordinates, because nothing shares a memory.

Nothing flows back to acquisition

The program discovers your most valuable customers, and the ad platforms never find out. Meta and Google keep optimizing for fast converters because nobody told them what a loyal customer looks like.

1. Pick the goal before the mechanics

Decide what the program optimizes: long-term customer value, not enrollment or redemption rate. The agent lets you set that goal for the whole store, and the program follows it (how it decides). Without a prediction model, recency, frequency and value segments are the honest starting point.

2. Reward predicted value, not last year's spend

A spend threshold rewards the past. The customer who spent a lot last year and is about to leave gets the best status. The new customer whose behavior matches your best buyers gets nothing. The most useful moment to shape a relationship is the first 30 to 60 days, before the pattern sets, so let predicted value decide the treatment from the first order.

3. Make every loyalty event a trigger

Points earned, points about to expire, a voucher earned, a voucher unused: each one should start a message on the channel that customer answers, with the products they are likely to want. In the agent, points and vouchers live in the same journeys as the rest of retention, next to email, push, SMS, Viber and WhatsApp (retention).

4. Show the balance everywhere

Every email and the customer's own loyalty page should show the points and the vouchers within reach, next to products at that price. A monthly statement is not a program.

5. Send the signal to your ads

When a valuable member buys, send the purchase and the predicted value of that customer to Meta and Google, with purchases counted once and existing customers excluded from acquisition. Retention data then improves who the ads find (ads audiences).

Points plus vouchers, the familiar mechanic: points per unit spent and points for actions like subscribing, redeemed as vouchers. It fits stores where people buy often: grocery, coffee, pet supplies, beauty. It works best when the allocation underneath follows predicted value rather than treating every member the same.

Behavior rewards without points: a benefit for the second order within 30 days, another for a review. It fits brands with a real community, where a points balance feels forced.

Invisible treatment by predicted value: no points at all, just progressively better service, earlier access and better picks as a customer's predicted value rises. It fits categories with 30 to 90 day buying cycles, such as fashion, home and electronics, where nobody needs a points nudge to buy on their natural schedule.

Whatever the model, measure it as incremental value: what members are worth against what they would have been worth without the program, with a holdout group, minus the cost of the rewards. Total enrollment, points issued and member revenue say nothing about that. PwC reports that 63% of US executives increased loyalty spend; the question is what it returns.