Most loyalty programs are a discount paid to people who were already coming back.
That is the central problem. A scheme that gives 10% to your best customers costs you 10% of your most reliable revenue and changes nobody's behaviour. It feels like retention work; it is margin donation.
A program worth running has to shift something — visit frequency, order size, or the number of new people arriving through existing ones.
Decide which behaviour you are buying
Pick one. Programs that chase all three do none.
- Frequency. Get people back sooner. Rewards should be time-bound and reachable quickly.
- Order size. Get people to spend more per visit. Thresholds and bundles rather than punch cards.
- Referral. Get existing customers to bring new ones. Almost always the highest return of the three, and the least used.
- Retention of the at-risk. Bring back people drifting away. Requires knowing who they are, which most businesses do not track.
Why referral beats loyalty
A referral program pays only when a new customer arrives. A loyalty program pays continuously to people already spending. The economics are not close.
Referred customers also tend to arrive pre-qualified — they were told what to expect by someone they trust — and consequently convert faster and churn less.
What makes a referral program work:
- Reward both sides. One-sided rewards feel like the customer is being used as a salesperson.
- Make the ask at the peak moment — right after a great experience, not in a monthly email.
- Make sharing one tap. A pre-written message and a link. Every extra step halves participation.
- Give something worth mentioning. 5% off is not worth a text to a friend. A meaningful credit, an upgrade, or early access is.
- Pay out fast and visibly. A reward that arrives weeks later trains people not to bother again.
If you do run a loyalty scheme
- Make the first reward reachable quickly. Programs where the first payoff is ten visits away are abandoned at visit two. Front-load it.
- Give status, not only discount. Early access, a better table, a known name at the door, first look at a drop. Status costs almost nothing and is valued more than money by exactly the customers you most want to keep.
- Keep the maths visible. If people cannot tell how close they are, the mechanism does not motivate.
- Do not require an app. For most small businesses the download requirement kills participation outright.
- Expire points, gently and clearly. Unbounded liability is a real balance-sheet problem.
The version that works for service businesses
Punch cards suit coffee, not consultancy. For higher-value, lower-frequency services the equivalents are:
- A retainer or membership that makes the next engagement a default rather than a decision
- Referral fees paid properly — and disclosed, where regulation requires it
- A genuine client alumni relationship: keeping in touch with people who no longer buy, because they refer
- Priority access for returning clients when capacity is tight
What to measure
- Repeat rate before and after launch — the whole point
- Time between purchases for enrolled versus non-enrolled customers
- Referral rate — new customers per existing customer per quarter
- Program cost as a share of the revenue it influenced, not of total revenue
- Incrementality: would these people have bought anyway? Uncomfortable, and the only number that settles whether it works
The bottom line
Pick one behaviour. Prefer referral, because it only pays on new revenue. Make the first reward fast, give status alongside discount, ask at the moment of satisfaction, and check honestly whether it changed anything.
Where this sits in retention generally is covered in the metrics that matter.
Want a program designed around your actual economics? Let's talk.