In brief
A bank reward program is paid for by the merchant, not by the customer collecting it. That one fact decides which parts a shop can copy and which will quietly take the margin off every sale.
The short answer
A bank reward program is funded by someone other than the person collecting the reward. The merchant pays a fee on every card transaction, part of that fee reaches the card issuer, and the issuer returns a slice of it to the cardholder as points or cashback. The customer is being paid with money a shop already spent.
That is why the headline rates look generous, and it is why you cannot copy them. What you can copy is the structure: tiers, category multipliers, and a balance treated as an obligation rather than a marketing number.
This post describes the mechanics generically. It names no bank and quotes no terms, because those change quarterly and a post that repeated them would be wrong by the time you read it.
Where the money comes from
Three funding sources sit behind a card reward program, and the cardholder is not the main one.
Interchange. Every card payment carries a fee, and a portion of it goes to the bank that issued the card. That flow is the engine. The reward is a rebate on a fee the shop paid, routed through the customer so it looks like a gift.
Annual fees and interest. A premium card charges to exist, and a revolving balance earns the issuer more than the reward costs. The heaviest rewards attach to the products with the heaviest fees.
Partners paying for placement. When points transfer to an airline or a hotel, money moves between the two companies. The issuer is selling access to a customer base, and the reward is the thing the customer sees.
Now count the parties in your own program. There is one. You are the merchant, the issuer and the funder, and every reward you hand over comes out of your own gross margin. There is no fee flowing back to you from anywhere, and no partner buying access to your regulars.
The mistake, stated plainly
The common failure is copying the generosity without the funding.
An owner reads “2% back on everything”, decides to be more generous than a bank, and writes “5% back in store credit” on a poster. That is not a loyalty program. It is a 5% price cut with a delay and some paperwork, given to everyone including the people who were coming anyway.
The fix is to stop measuring rewards as a share of the price and start measuring them as a share of the margin. A coffee that sells for £3.20 and costs £0.80 in beans, milk and cup earns £2.40. Five per cent of £3.20 is 16p, which is 6.7% of what you actually made. Survivable.
Run the same 5% on a £40 item in a shop with a 30% gross margin. The margin is £12, the reward is £2, and you have given away one sixth of the money. The rate on the poster is identical. The cost to the business is six times higher.
Do that division before you pick a number. The reward is affordable or not depending on your margin, and a rate borrowed from a company with a completely different income statement tells you nothing.
Mechanic by mechanic
| Mechanic | Why it works at bank scale | The small-business version |
|---|---|---|
| Cashback rate | Funded by interchange the merchant already paid | No version. You are the merchant |
| Sign-up bonus | Acquisition cost recovered over years of card use | One stamp on joining, costing you one item eventually |
| Tiers | Millions of accounts make every tier a real population | Two tiers, split on visits, never on spend |
| Category multipliers | Steers spend toward categories that pay the issuer more | Double stamps on the high-margin item or the dead day |
| Points as currency | Valued, provisioned and audited as a liability | Count outstanding stamps monthly. That is what you owe |
| Partner network | Points move to airlines and hotels that pay for the traffic | Rarely works. A two-shop swap needs a shared ledger and trust |
| Expiry rules | A compliance function with notice periods and disclosure | One sentence on the poster, honoured without argument |
Three of those seven transfer cleanly. The rest are listed so you can recognise them when a vendor demonstrates them as features.
Tier by behaviour, not by spend
Bank tiers usually key off annual spend or off which product you hold. Both are invisible to the customer until a statement arrives, and both need a system that totals money over a year.
At two hundred customers, spend tiering is the wrong axis twice over. It is hostile, because it tells the customer that the way to matter is to hand over more money. And it is hard to run, because it means capturing basket totals accurately at a counter where someone is also making the coffee.
Frequency is the better axis. The person who came in twelve times this quarter is a different customer from the person who came twice, and counting visits is something a scan already does. A visit-based split needs no till integration and no arithmetic at the counter.
Keep it to two tiers. Two is a distinction people can hold in their heads. Five is an accounting department, and every extra tier is another rule the person on shift has to remember at the moment a customer is waiting. The visit-based program guide works through the counting side of this.
Multipliers are the part worth stealing
Category multipliers are the one bank mechanic that transfers almost unchanged, because they cost nothing to declare and they move behaviour.
A bank pays extra on restaurants and fuel for reasons of its own. You would pay extra on the thing you want sold: the item with the fat margin, the day the shop is empty, the size you would rather people ordered. Double stamps on a Tuesday costs you an extra stamp on the visits that happen on Tuesdays and nothing at all on the rest.
The constraint is the counter, not the concept. A multiplier that requires the person on shift to check a matrix will be applied differently by two people on the same day, and customers notice inconsistency faster than they notice generosity. One multiplier, one condition, stated in a sentence. If you cannot fit the rule on the poster, it is too clever.
Points are a liability, and yours has no ledger line
A bank knows what its outstanding points are worth. The balance is estimated, provisioned and audited, because a promise to hand over something later is a debt whether or not anyone calls it one.
Your outstanding stamps are the same debt with nothing recording it. The number is easy to produce and worth producing.
Two hundred enrolled customers holding an average of four stamps on a ten-stamp card is 800 stamps in circulation, or 80 rewards’ worth. At £0.80 of cost of goods each, you owe about £64 of coffee. That is not alarming, and saying so is the point: most small programs are carrying a debt smaller than a week’s milk order.
It matters at two moments. When you change the offer, that number is what honouring both versions costs. And when you consider stopping, it is what walking away would take from people who kept their side. Knowing it requires a ledger you can read, which is the argument for a dashboard over a drawer of paper cards, and for reading the balance side of the program before the marketing side.
What does not transfer, and why
Scale. A bank’s numbers are stable because they are averages over millions of accounts. Breakage, redemption rates and tier movement are predictable at that size. With two hundred customers, one family moving away is a visible dent in your monthly figures. Averages are not available to you, so run the program on rules you can defend rather than on rates you tuned to last month’s noise.
Partner networks. Transferring points into an airline requires two companies with contracts, settlement and a shared valuation of a point. The local version, a card that works at your shop and the barber next door, needs a shared ledger, a split of the cost, and agreement about who honours what when one of you is closed. It is not impossible. It fails on the second month, when one shop has issued more than the other and nobody agreed in advance who pays.
The funding model. Covered above, and the one that matters. Nothing sends money back to a shop for accepting a card.
Regulatory machinery. An issuer has a department for changing terms: notice periods, disclosures, records of what was promised to whom. Your version of that department is a poster with the rule on it and a decision to honour what it says even when a customer reads it more favourably than you meant. That costs you an occasional free coffee and buys you the ability to change the program later without an argument.
What to take
Three things, in order of how much they are worth.
Price the reward against margin, not against turnover. Steer with a multiplier on the thing you want sold. Count what you owe once a month.
Everything else on the list is a feature of having millions of customers, an income statement funded by other people’s fees, and a compliance department. Take the structure and leave the rate. Then read how a program actually makes its money at your size, because it is not the same five answers.