Referrals are the one channel a small business already owns. The people who like you enough to say so are standing in front of you every week, and the whole problem is that asking them feels like work, both for you and for them.
A wallet-native referral removes most of that work, because the card is the link. What it does not decide for you is the rules, and the rules are what separate a referral program that brings people in from one that quietly pays regulars for friends who were coming anyway. This page is about the rules. What the mechanism does is on the referrals page, and this one does not repeat it.
The rule that matters more than the others
Pay for the visit, not the signup.
Every other decision here is a preference. This one is structural, and it is the difference between a referral program and a leak. A signup can be produced from a sofa in ten seconds by somebody with two email addresses. A person standing at your counter being scanned cannot.
In Fideliya this is not advice, it is the code path. A referral is created when the friend joins and sits at pending. It completes when that friend is scanned on an earning action, and only then are both bonuses written. The claim is a single atomic update against the pending row, so a double scan cannot pay twice. And a deduction does not count: spending from a gift card is presence, not an earn, and the referral block excludes it explicitly rather than letting a subtraction credit two people in the middle of itself.
The practical consequence for your rules is large. Because the money only moves when somebody walks in, you can afford to be generous with the offer and relaxed about policing it. There is nothing to police.
The four choices you actually make
| Choice | The safe default | Change it when |
|---|---|---|
| What the referrer gets | One, in the card's unit | Your goal is long and one barely registers |
| What the new customer gets | One, in the card's unit | You want the head start to be the hook |
| Which cards it runs on | Your main card only | Never, if a second card is staff-only or seasonal |
| When you mention it | At the moment a reward is claimed | Never at the first visit, when there is nothing to endorse yet |
Fideliya defaults both bonuses to one, in whatever unit your card counts, and referrals are off until you switch them on. Those defaults are conservative on purpose and they are usually right. Two reasons to leave them alone: a small bonus keeps the ask honest, because your customer is doing a friend a favour rather than selling on commission; and a bonus that is smaller than the reward keeps the reward the thing people want.
The one case for raising the referrer's side is a long goal. On a card that needs ten visits, a single stamp for bringing somebody in is a rounding error, and two or three is a better match for the size of the favour. The one case for raising the new customer's side is when you would rather the head start did the persuading than the person did.
The ceiling nobody tells you about
Bonuses are clamped to the card's goal. If your card needs ten stamps and a customer sits at nine, a two-stamp referral bonus does not write eleven. It writes ten, and on the visits path it grants the reward at the crossing rather than leaving somebody parked at the goal with nothing available.
This matters when you set the numbers. A bonus larger than a couple of steps on a short card is mostly wasted, because the ceiling absorbs it and the customer experiences a smaller jump than you paid for. Set the bonus against the length of the ladder rather than against how grateful you feel.
What to say at the counter
The mechanism is easy and the sentence is hard. Three rules make the sentence work.
- Say it when the card is already open. The moment your customer is looking at their own progress is the only moment the referral is one tap away rather than a task for later.
- Say it once a reward has been claimed, not at enrolment. Nobody recommends a loyalty program they have not yet benefited from, and asking at the first visit spends your one clean ask on somebody with nothing to say.
- Put the offer in the sentence. Show this to a friend, they start with a stamp and so do you when they come in. It is short, it is true, and it names both sides.
What does not work, and is worth naming because it is the obvious idea: a poster. A referral needs a person to hand a phone to another person. A poster reaches people who are already in the shop, which is the audience a referral program is not for.
What the product will not do for you
Three honest limits, all of them in the code rather than in the marketing.
Nothing invites on your behalf. There is no automatic message asking your regulars to refer somebody. The card carries the link; a person still has to show it. If your plan depends on the platform doing the asking, it depends on something that does not exist.
A referral made while the toggle is off, or while the account is below Pro, is silently dropped. Fideliya checks both at the moment of joining, logs a reason, and lets the friend join normally with no bonus for anyone. Nobody is told. That is the correct behaviour for a stranger standing at a till, and it means the toggle is worth checking before you promise anything out loud.
There are no tiers of referral. A friend of a friend credits nobody twice, and there is no chain, no second degree and no ladder of referral ranks. One customer, one friend, one bonus each, paid once at that friend's first visit.
One more thing worth knowing if your customers are on Android. When several notifications compete for a pass on the same day, a referral bonus sits second in the order Fideliya uses, behind a reward being earned and ahead of a broadcast. So a referral that pays out on a busy day still gets through; a marketing message that day may not. The channel behaviour behind that is on push notifications from a loyalty card.
Reading it afterwards
Watch the funnel, not the total. The referral hub separates enrolled from first visit from pending, and the gap between the first two is the only number that tells you anything: referrals that enrolled and never came in are the ones your rules are failing to convert. A large pending count is not a fraud problem, it is a follow-through problem, and the fix is usually the sentence at the counter rather than the size of the bonus.
We are publishing no conversion figure of our own, because we have not measured one. Referral conversions are on the explicitly unmeasured list in our last production snapshot, so anything we printed here would be a number we invented. What to measure instead is on the seven numbers a loyalty program should report, and the other campaigns that fit around this one are on twenty loyalty campaign ideas.
Sources
Every product behaviour on this page was read from this repository's source on 6 September 2026 and is cited in the source note on this article to its file and to the symbol inside it: the pass token route that makes the card a referral link, the two bonus defaults and the enabled flag, the three checks made at joining and the reason each failure logs, the completion gate that excludes deductions, the atomic claim that prevents a double award, the clamp to the card's goal with its crossing grant, the plan gate in the feature matrix and its server-side enforcement, and the notification priority order. This page quotes no statistic and no figure from our own production, because referral conversions are on the not-measured list in our last snapshot.
Fideliya is wallet-native loyalty program software for small businesses. Customers add a stamp or points card to Apple Wallet or Google Wallet in one tap, with no app to download and no account to create. Fideliya runs stamps, points, gift cards and referrals on one platform in English, French, Spanish and Arabic with full right-to-left support, and the free plan issues a real wallet pass for up to 20 customers. What the referral mechanism does is described on the referrals page, and the plans are on the pricing page.