Decide what behaviour you are buying
A programme is a lever, and a lever needs something specific to move. More visits, bigger baskets and fewer lapses look alike from a distance. They take three different mechanics.
- More visits. Reward a rhythm a regular can reach; count visits, not money.
- Bigger baskets. The reward must scale with spend: points, or a spend threshold rather than a flat stamp.
- Fewer lapses. Identification matters more than the mechanic — you cannot win back a customer you cannot name.
Pick one. A programme aimed at all three rewards nothing in particular. If the answer is lapses, the first job is knowing who your regulars are — where most in-store loyalty decisions start.
Cost the reward against cost of goods
A reward is a discount with a delay. What decides affordability is what it costs you to produce, not what it sells for.
| Input | Illustrative value | What it drives |
|---|---|---|
| Average qualifying purchase | $5.50 | Revenue behind each stamp |
| Purchases to a reward | 8 | $44.00 of qualifying spend per card |
| Menu price of the reward | $6.00 | What the customer perceives |
| Cost of goods on the reward | $2.10 | What it costs you |
| Effective discount | 4.8% | $2.10 against $44.00 |
Menu price would call this a 13.6% programme — $6.00 against $44.00, nearly three times the real figure, and an easy way to talk yourself out of something workable. The threshold is the dial: six purchases is 6.4%, eight 4.8%, ten 3.8%. Check a regular can reach it inside a month.
Illustrative example — not a customer result. The figures are placeholders. Replace them with yours; nothing here predicts how many people join or finish a card.
Pick the mechanic
Four mechanics cover most small shops. What separates them is not features but what a customer must understand and what staff must say.
| Mechanic | Use when | What staff say | Main risk |
|---|---|---|---|
| Stamps | One repeatable purchase, one fixed reward | “Ninth one is on us” | A threshold picked as a round number |
| Points | Baskets vary, or several reward costs | “A point per dollar; 200 is a free one” | A rate nobody converts into value |
| Tiers | A minority spend far more | “Three visits to the next tier, which earns double” | Thresholds most customers never reach |
| Visits | Price per job varies | “Fifth appointment is half price” | Equal reward for unequal spend |
Stamps and points are the real choice for most shops; the longer comparison is in the points versus stamp cards guide. Tiers come later — they need a member list to tier.
Choose the format — join friction dominates
Every design above assumes members, and members come from a join that happens while somebody is holding their order and wanting to leave.
- Paper. Lowest friction, least information — you cannot tell a regular from a stranger.
- An app. Most capable, hardest to join: an install request lands at the worst moment in a transaction.
- A scan. A card in the phone's own wallet, or a hosted member page. One scan survives a queue.
Rewardfinity issues a hosted member page and a PWA. Google Wallet is a configured path; Apple Wallet public launch remains dark pending its physical canary. Confirm the exact state before printing signage promising either wallet. Put the code where customers already look while paying.
Write the sentence, and make the ask part of the transaction
One sentence a person can say while handing over an order, plus a follow-up for whoever asks what it is.
Ask and follow-up
Ask: “Want this on the card? Ninth one is free.”
Follow-up: “Scan that code — ten seconds.”
Say it while payment is processing, not after the bag is handed over, and keep it shorter than a customer's patience.
Somebody also has to want to say it in week six. A staff signup competition is the usual advice, and it needs three things to run on: attribution, a scoreboard and a reward. Decide before you launch which of those your own setup can actually produce, because advice without a mechanic is a memo. Rewardfinity attributes counter actions to the staff member who performed them and turns those into a staff-facing spin wheel with an owner-set allowance. Check what your counter devices record.
Launch to one shift first
Pick one shift, ideally a busy one, and run for a week with everything else unchanged. You are not testing the reward — you are testing the join: whether the sentence gets said, whether the code is reachable, whether a refund does the right thing to a balance.
- Run one real transaction end to end, including a refund.
- Watch the first ten signups happen rather than reading about them later.
- Fix the sentence before widening it — twelve people unlearn slower than two.
Pick the review date and the number that changes the rule
Write down the date you will look and the result that would change the rule. Deciding afterwards produces a programme nobody cancels, because nobody defined failure.
- 01When is the review?Four to six weeks shows joins, not retention.
- 02What join rate do you need?Members added divided by transactions. This number fails first.
- 03What lowers the threshold?The completion rate at which the card is unreachable.
- 04What would make you stop?A defensible exit; without one it runs on sunk cost.
Measure joins first, completions second, repeat behaviour last — the order they become observable.
Pre-launch checklist
- The behaviour you are buying, in one line.
- The reward costed at cost of goods, not menu price.
- A threshold from that discount, reachable within a month.
- The reward named precisely, including size or variant.
- The remainder rule when someone earns two rewards at once.
- The sentence written and visible to whoever is on shift.
- Every shift briefed, including whoever works Sundays.
- The review date and trigger number in the calendar.
How these programmes fail
| Failure mode | What you see | Usual cause |
|---|---|---|
| Nobody joins | Footfall fine, few members | The ask sits outside the transaction |
| Joins stall by week three | Good first week, then flat | No staff attribution; the sentence stopped |
| Cards never complete | Members stuck at two stamps | Threshold set as a round number |
| Costs more than expected | Margin drift, no obvious source | Reward costed at menu price |
| Staff argue with customers | Disputes about what qualifies | An edge case nobody decided |
| No idea whether it worked | Opinions instead of an answer | No baseline, no review date |
Questions owners ask before starting
How do I start a loyalty program for a small business?
Decide which behaviour you want more of, cost the reward at cost of goods rather than menu price, pick a mechanic that matches, and reduce joining to one scan. Then write the sentence staff will say and run it on one shift for a week.
How much should a loyalty reward cost?
Treat it as a discount rate. A regular spending $5.50 a visit across eight visits puts $44.00 of qualifying spend behind one completed card. A reward costing you $2.10 in goods makes that 4.8%. Six visits would be 6.4%, ten would be 3.8%.
Do I need an app for a loyalty program?
No, and requiring one is a common reason programmes stay small. Asking someone to install an app while they are paying is a large request at a bad moment. A card added by scanning a code, held in the phone's built-in wallet, keeps the queue moving.
How long does it take to set up a loyalty program?
Configuring the software is usually an afternoon. The decisions take longer and matter more: the behaviour you are buying, the reward's cost of goods, the threshold, and the wording of the qualifying rule. A week of decisions, one shift live for a week, then launch.
What is the best loyalty program for a small shop?
There is no single best one. One repeatable item is usually clearest as a stamp card. Varied baskets or several reward costs usually need points. Services priced per job often fit visit counts. The useful test is which rule staff can say in one sentence.
How do I get customers to actually sign up?
Make the ask part of the transaction rather than an addition to it, said while payment is processing rather than after the bag is handed over. Keep joining to one scan with no install, and credit signups to whoever got them so the ask survives week six.
Settle the four decisions, then open the software.
Configuration takes an afternoon. The sentence takes longer.