Loyalty programmes have a credibility problem, and it is deserved. A large share of them transfer margin to customers who would have returned anyway. That does not make the category worthless: it makes measurement essential.
Do loyalty programmes increase visits?
Loyalty programmes increase visits in a specific and narrow way: they pull forward and consolidate visits among customers who already use the category but spread their spending across several businesses.
The mechanism is not gratitude. It is the sunk-progress effect: a customer holding a card that is seven-tenths finished has a concrete reason to choose you over the equally good shop across the road. That effect is strongest close to the goal and weakest at the start, which is why the goal you set matters more than the reward you choose.
Where do loyalty programmes fail?
Loyalty programmes fail in three recognisable ways, and all three are visible in the data within about two months.
- Subsidising the already-loyal. Your daily regulars enrol first and would have come regardless. Early data always looks fantastic for exactly this reason.
- Enrolment collapse. The programme reaches 8% of customers because sign-up needs effort at a busy counter, and 8% of customers cannot move a business.
- Reward mismatch. The goal is so distant that nobody believes in it, so the card never leaves the first stamp.
How do I tell whether mine is working?
You tell whether a loyalty programme is working by measuring visit frequency per enrolled customer before and after enrolment, segmented by how often they visited to begin with.
| Segment at enrolment | What you want to see | What it means if flat |
|---|---|---|
| Visited 4+ times/month | Little or no change | Expected: these are subsidy, not growth |
| Visited 1 to 3 times/month | Clear increase | This is where the programme earns its money |
| First-time visitor | Second visit within 30 days | The card failed to create a reason to return |
If the middle row does not move, the programme is a discount scheme. That is a fixable problem, usually a goal that is too far away or a reward nobody wants, but it will not fix itself.
What does the evidence support?
The evidence supports a modest, real effect that depends almost entirely on reach. Industry reporting consistently finds wallet-based programmes achieve far higher enrolment than app-based ones, and enrolment is the binding constraint: a programme with a strong mechanic and 10% reach loses to a plain one with 60% reach every time.
The mechanic decides how much each enrolled customer is worth. Enrolment decides how many of them there are. Only one of those is usually the bottleneck.
What is the honest conclusion?
How do I run a fair test?
The problem with measuring a loyalty programme is that the customers who join are not a random sample. Your most engaged customers enrol first, so any before-and-after comparison flatters the programme. A fair test needs something to compare against.
The rigorous version is a holdout: enrol customers as normal but withhold the rewards and messaging from a randomly chosen tenth, then compare their visit rates. That is how large retailers measure it, and it is more machinery than most independent businesses want.
The practical version is segmentation by prior frequency, which is what the table above describes. Comparing occasional customers before and after enrolment is not perfect, but it removes the largest source of self-deception: the daily regular whose visits were never going to change.
A third option costs nothing: compare the same month against the previous year. It is crude, it absorbs every other change in the business, and it is still better than looking at enrolment totals and calling them a result.
How long before I see results?
Expect the first meaningful read at about three months, because a customer has to enrol, accumulate progress, and approach the reward before the mechanic can change anything. Anything you measure in week two is measuring enthusiasm, not retention.
The sequence is fairly consistent. In the first month enrolment numbers climb quickly as your existing regulars join, and the data looks excellent for the wrong reason. In the second month enrolment slows and the figures look disappointing. From the third month onward, the occasional customers who joined early start reaching the reward threshold, and that is the first honest signal.
Judging the programme at the end of month one or month two leads to the wrong decision in both directions: celebrating a subsidy in the first case, abandoning a working programme in the second.
Which numbers mislead?
Three metrics are reported by every loyalty dashboard and none of them indicate whether the programme is working.
| Metric | Why it misleads | What to use instead |
|---|---|---|
| Total enrolled | Counts sign-ups, not behaviour change | Share who reached half the threshold |
| Stamps issued | Rises with footfall you already had | Visit frequency of occasional customers |
| Rewards redeemed | Measures the discount you gave away | Repeat rate after first redemption |
| Push open rate | High for all wallet cards, always | Visits within 14 days of a message |
The redemption row is the most commonly misread. A high redemption rate is often presented as success when it can equally mean you are handing rewards to customers who needed no encouragement, which is precisely the failure mode in the first list above.
When should I stop a loyalty programme?
Stop if, after six months, occasional customers show no change in visit frequency and enrolment has stalled below roughly a quarter of your regulars. At that point the programme is a discount with administrative overhead.
Before stopping, though, check which of the two is broken, because they have different fixes. Low enrolment is an ask problem: the friction at the counter, the sentence staff use, or the fact that they have stopped offering it. No change among enrolled occasional customers is a design problem, usually a threshold too far away or a reward nobody particularly wants.
Most programmes that get abandoned were never given a fair chance at the first of those. Enrolment is fixable, and it is worth one deliberate attempt before concluding the category does not work for you.
What is the honest conclusion?
The honest conclusion is that a loyalty programme is a modest lever that is cheap to pull, and worth pulling if, and only if, you can enrol a large fraction of your customers without slowing the counter down. If enrolment requires the customer to do work, the programme will mostly discount people you already had.