In brief
Almost every failed loyalty program failed in one of eight ways, and six of them were decided before any software was bought. Each has a symptom you can see from the till and a fix that is either an adjustment or a restart.
The short answer
Loyalty programs rarely fail because of the software. They fail because the threshold was set too far away, the qualifying event was ambiguous, redemption needed a manager, or nobody at the counter ever asked. Six of the eight failures below are decided on paper before any tool is chosen, which is also why changing vendor almost never fixes one.
The eight, in order
| # | Symptom at the counter | Cause | Fix |
|---|---|---|---|
| 1 | Cards issued, almost none completed | Threshold too far away | Lower it, or add an earlier reward |
| 2 | Customers argue about whether something counted | Ambiguous qualifying event | One sentence a new hire can apply |
| 3 | “Let me get the manager” | Redemption needs authority nobody on shift has | Give the shift the authority |
| 4 | No answer to “is it working?” | Ledger nobody can read | Restart on a readable one |
| 5 | Enrolments stopped after week two | Nobody asks | A scripted sentence, and checking |
| 6 | People ask what it is and leave confused | Rules need explaining | Cut rules until it fits on a poster |
| 7 | Program rewards big spenders you did not mean to reward | Counting spend when you meant visits | Change the qualifying event |
| 8 | Everything half-works | Launched tiers, points, referrals and birthdays at once | Turn off all but one |
1. The threshold is too far away
The most common, and the most expensive because it is invisible for months.
A ten-stamp card given to customers who come in once a month is a ten-month commitment. Nobody finishes. The cards are issued, the enrolment number looks healthy, and a year later there have been four redemptions.
What it looks like: high enrolment, near-zero completion, no complaints — people do not complain about a card they quietly stopped carrying.
The fix: set the threshold from the customer’s existing visit rate so the first reward arrives in six to ten weeks. For a weekly customer that is six to ten stamps. For a six-weekly salon client it is two or three, or a card with a reward at stamp two as well as at the end.
Restart threshold: if the current card is more than double the right length, do not adjust — relaunch, honour the old cards for a stated window, and say the date out loud.
2. The qualifying event is ambiguous
“A stamp per visit” sounds unambiguous until two people buy four coffees together, or one customer comes back at 4pm having been in at 9am, or someone buys a bag of beans and no drink.
What it looks like: two members of staff apply it differently, and a customer notices. That conversation happens in front of the queue and it costs more than the stamp.
The fix: write one sentence a new hire can apply without judgement. “One stamp per customer per day, any purchase.” Not a policy document — a sentence.
3. Redemption needs a manager
The card is full, the customer asks, and the person on the till does not have the authority or the confidence to hand it over.
This is the worst one, because it fails at the exact moment the program was supposed to pay off. Nine visits of goodwill spent in one interaction.
What it looks like: an apologetic pause, then “can you come back tomorrow when the manager’s in?”
The fix: whoever is on shift must be able to honour it alone. That constraint should have shaped the reward in the first place — if the reward is something a junior member of staff cannot give away, it is the wrong reward. There is more on splitting that authority in who on your team can do what.
4. The ledger nobody can read
A rubber stamp on a paper card records nothing. A year in, there is no way to answer how many cards are outstanding, how many customers stopped coming, or whether frequency moved.
What it looks like: the program is neither obviously working nor obviously failing, and the decision to continue is made on feel.
The fix: this one is a restart, not an adjustment. There is no way to recover data that was never written down. Moving a paper stamp card to QR in a weekend is the migration.
5. Nobody asks
Enrolment is a staff habit, not a marketing campaign, and habits decay in about two weeks without reinforcement.
What it looks like: a launch spike, then a flat line. The program is running and nobody is joining it.
The fix: one scripted sentence, said the same way by everyone, and someone actually checking the weekly enrolment number. If the sentence takes longer than four seconds, shorten it. If joining takes longer than the sentence, the enrolment flow is the problem, not the staff.
6. The rules need explaining
Points that convert at a rate, tiers with names, bonus categories, an expiry policy. Each addition sounded reasonable in isolation.
What it looks like: the customer asks how it works, the staff member gives a thirty-second answer, and the customer says “sure” in the tone that means no.
The fix: cut until the whole program fits on the poster in one sentence. If a customer cannot repeat the offer to a friend, it will not spread.
7. Rewarding spend when you meant visits
A points-per-pound program rewards the customer who spends most. A stamp-per-visit program rewards the customer who comes most. These are different people, and most local businesses want the second one.
What it looks like: the program’s biggest beneficiaries are occasional large orders rather than the Tuesday-morning regulars you were trying to keep.
The fix: change the qualifying event to the visit. Which type of loyalty program to choose covers when spend genuinely is the right unit.
8. Everything at once
Tiers, points, a referral bonus, birthday rewards and an email sequence, all live on day one.
What it looks like: nothing is obviously broken and nothing is obviously working. Every part is under-explained, so no part gets used, and there is no way to tell which one was the good idea.
The fix: turn off everything except the simplest mechanic. Add the second thing when the first is boring. A program you find boring is a program that is working.
The two that are not on this list
A name nobody likes. Gets argued about for weeks and matters least. Naming a program is worth twenty minutes, not a meeting.
Choosing the wrong vendor. Real, but far rarer than it is blamed for. If a program is failing for any of the eight reasons above, switching software carries the same failure into a new account. Fix the design first, then decide whether the tool was ever the problem.