Loyalty, promo codes and the spin wheel: what actually brings people back

Most loyalty schemes fail for one of two reasons: nobody understands them, or nobody costed them. Both are avoidable in an afternoon.

Points are a liability

When you award a point, you have created a small debt. It sits on your books until someone redeems it or it expires. That is not a reason to avoid loyalty — it is a reason to price it before you print the poster.

Do the arithmetic once:

cost of the scheme = points issued per order × value per point × redemption rate

If a point is worth a small amount and roughly half get redeemed, you know your effective discount per order before you launch. If that number is bigger than the margin on the item you are trying to sell more of, redesign it now.

The one-sentence test

Ask a member of staff to explain the scheme to a stranger in one sentence. If they cannot, redesign it.

“Every 100 you spend is a point, and 10 points is 100 off.” That works. “A point per 10, points worth 0.7, not valid on promotions, expiring after 90 days” does not — not because customers are slow, but because nobody at a busy counter will say all that, so nobody will hear it.

The four things that actually earn a point

A loyalty account can be credited from more than a purchase, and the non-purchase ones are usually the ones that change behaviour:

Referral tracking matters because it is the only acquisition channel where you know exactly what you paid per new customer, and both the referrer and the referred can be rewarded once each — no more, which is what stops the loop being farmed.

Promo codes need three limits

A code with no boundaries is a discount you no longer control. Every code should carry:

1. A start and end date. Nothing runs forever by accident.
2. A usage cap. Total uses, so a code posted publicly cannot empty a month.
3. A scope. A percentage, a fixed amount, or buy-one-get-one — and optionally a single product it applies to, or a single customer it belongs to.

That last option is worth more than it looks. A personal code sent to one customer who has not ordered in eight weeks converts far better than a public one, and it costs nothing if they do not use it. Broadcast discounts mostly subsidise the people who were coming anyway.

The spin wheel, used honestly

A prize wheel with weighted probabilities — some points, a small discount, occasionally a free item, sometimes nothing — is a game, and games work because the outcome is uncertain. Two rules keep it from becoming expensive or annoying:

• Set the probabilities deliberately. The expected value of a spin is a real cost. Calculate it the way you calculated points.
• Rate-limit the spins. One per customer per period. A wheel you can spin repeatedly is not a game, it is a coupon dispenser.

What to do with feedback

Ratings with a comment, tied to an actual order, are worth more than a public review because you know which visit produced them. Reward the act of leaving feedback, never the score — the moment a reward depends on five stars, you have stopped collecting information and started buying it.

Then reply. A reply is visible to that customer and takes a minute, and the complaints you answer are the ones that do not become public.

A starting scheme that works

• One point per round unit of spend, ten points for a fixed reward. One sentence.
• Referral: both sides rewarded, once each.
• One personal offer per month to customers who have gone quiet.
• Public codes only for a real campaign, always with a cap and an end date.

Then look at the redemption rate after six weeks. If almost nobody redeems, the reward is too far away. If almost everybody does, you are discounting the customers you already had.