In brief
Most loyalty programs are built in the wrong order: software first, reward last. Reverse it. The reward and the threshold are the only decisions that cost money, and both can be tested on ten people before anything is bought.
The short answer
Build it in this order: pick the reward, price the reward, set the threshold, decide who validates a visit, test it on ten regulars, then choose software and launch. Choosing software first is the single most common way to end up with a program nobody uses, because the tool cannot tell you what your customers are worth a visit.
The whole sequence fits in about ninety days without rushing, and the first four weeks of it cost nothing.
This post is the order. The content of each decision — what makes a good reward, how to pick a threshold, who should be allowed to validate — is how to start a loyalty program, which goes through the seven decisions one at a time.
Week 0: answer one question honestly
Which repeat visit are you actually trying to buy?
Not “more customers”. A specific behaviour you can name, from a specific group of people who already exist. The Tuesday-morning regulars who stopped coming. The lunch crowd that never comes back in the afternoon. The customers who buy once and vanish.
If the answer is “everyone, more often”, stop here. A program aimed at everyone has to be affordable for your best customer and interesting to your worst, and no reward is both. Most failed programs failed at this question, not at the software.
Week 1: four numbers, on paper
Nothing digital yet. Four numbers, written down:
- The reward, and what it costs you at cost of goods — not at menu price.
- The threshold: how many qualifying visits earn it.
- The current visit rate of the customers you named in week 0. Best estimate is fine; you will check it later.
- The qualifying event: what has to happen for the count to go up, in one sentence a new hire can apply.
Then do the multiplication. Threshold divided by visit rate is how long the card takes to complete. If that is more than about ten weeks, the threshold is too high and nobody will finish. If it is under three, you are giving away margin to people who were coming anyway.
This is also the moment the budget becomes real. Reward cost times expected completions per month is the biggest line in the whole program — see what a loyalty program actually costs for the worked version.
Weeks 2-3: run it on ten people, on paper
Print ten cards. Actual paper, actual pen. Give them to ten regulars and tell them it is a trial.
Two weeks of that will tell you things no amount of planning will:
- Whether staff remember to offer it without being reminded.
- Whether the qualifying event is as obvious at the counter as it was on paper.
- What people say when they ask what it is — their words are your poster copy.
- Whether anyone hits the threshold faster than you expected, which means it is too low.
Ten people is enough. The failure modes are not subtle, and they show up in the first three days.
Fix the four numbers based on what happened. Changing a threshold now is free; changing it after two hundred people are mid-card is not.
Week 4: choose the software, then launch
Now the tool, because now you know what it has to do.
Three questions, in this order:
Can a customer join without installing anything? An app download between a poster and a stamp loses most of the people who were willing. A card that opens in the browser or drops into Apple Wallet does not.
Can the person on shift validate a visit alone? Not the owner, not a manager. Whoever is actually there at 8am. If validation needs the owner’s phone, the program stops on the owner’s day off.
Can you read the ledger? If you cannot see how many cards are open, how many completed and how many people never came back, you will not be able to answer in March whether any of this worked.
Price is the fourth question, not the first. The gap between free and £9 a month is smaller than the gap between a program staff use and one they do not.
Launch is a poster with a QR code on the counter and one sentence staff say out loud. Not an email campaign, not a launch event. The counter is where the enrolments come from.
Day 30, 60, 90: three numbers
Set a reminder for each. The whole point of a readable ledger is that these take two minutes.
| Checkpoint | The number | What it means if it’s bad |
|---|---|---|
| Day 30 | Enrolments per week | The counter ask isn’t happening. Staff problem, not customer problem. |
| Day 60 | Share of cards with a second stamp | The offer isn’t worth returning for. Reward problem. |
| Day 90 | Completions, and repeat rate of completers | Threshold problem, or the reward isn’t worth finishing for. |
Day 60 is the honest one. A customer who takes a card and never comes back has told you the reward is not interesting. That is cheap information and it arrives early enough to act on.
The three orderings that fail
Software first. You buy a tool, then design the program around what the tool happens to do. The program ends up shaped like the software’s demo account, and the reward is whatever fitted in the config screen.
Launch first, decide later. The poster goes up before anyone has agreed what counts as a visit. Two months in, three members of staff are running three different programs and there is no way to reconcile them.
Everything first. Tiers, points, birthday rewards, referral bonuses and an email sequence, all on day one. Every one of those is a thing to explain at the counter, and the counter has about four seconds. Ship the stamp card. Add the rest when the stamp card is boring.
If you are building a rewards card program specifically
Same sequence, with one addition between weeks 3 and 4: decide what the card physically is before you print anything. Paper, browser link and wallet pass have different costs, different failure modes, and only one of them can be changed after the fact. How to make loyalty cards covers the three and what each one breaks.