In brief
You cannot invalidate the paper cards already sitting in customers' wallets, so the migration is a dual run with an end date you say out loud. The rest is one weekend of decisions, one printed poster, and a sentence staff can say without thinking.
The short answer
Replacing a paper stamp card with a QR one takes a weekend, and the work is mostly decisions rather than setup. Friday you decide what happens to the paper already in circulation. Saturday you set the rules and teach one sentence to whoever is on shift. Sunday you print the poster, test the whole journey on your own phone, and move your named regulars across by hand. Monday you go live and convert paper at the counter as people come in.
The software part of this is an hour. The part that goes wrong is the paper you cannot see, which is why it gets Friday evening to itself.
Friday evening: decide what happens to the paper already out there
Every card you have handed out is a promise, and you cannot cancel it because the customer is holding the evidence. Someone will walk in on stamp nine with a card that says “tenth coffee free”, and “we’ve moved to an app” is not an answer they will accept, or should have to.
Start by counting what is outstanding. You cannot do this exactly, which is the whole problem with paper as a ledger you can read, but you can estimate. Take how many blank cards you have handed out since your last print run, subtract the completed ones you collected, and you have a rough number of open cards. It will be higher than you think.
Then pick one of two policies. Invalidate and you break a promise in public, at the counter, in front of a queue. Honour paper forever and you are running two programs indefinitely, one of which you cannot audit. Neither is acceptable, so the answer is the third thing: a dual run with a stated end date.
Set the window by asking how long the worst-case customer needs. A ten-stamp card held by someone who comes in weekly and is only on stamp one needs nine more weeks. Most holders are somewhere in the middle, so eight weeks is a defensible window and it is short enough to say out loud without flinching. Pick the date, write it down, and put it on the poster, on the counter card, and in the sentence staff say.
The window is only the backstop. The real mechanism is conversion: when a paper holder comes in, you copy their count onto the new digital card, keep the paper, and hand it back to nobody. That converts a liability you cannot see into a balance you can, one customer at a time, and it means the deadline only ever applies to people who did not come in at all.
The weekend, hour by hour
| When | What you do | Done when |
|---|---|---|
| Friday, 30 min | Estimate open paper cards, set the dual-run end date | The date is a real calendar date, written down |
| Saturday, 9am | Set three rules: threshold, qualifying event, who validates | Each one fits in a sentence |
| Saturday, 10am | Build the card and write the reward line | The reward reads as one sentence a stranger understands |
| Saturday, 4pm | Staff walkthrough | Everyone on shift has said the sentence out loud once |
| Sunday, morning | Print and place the poster, then test the whole journey | A card exists on your own phone with a stamp on it |
| Sunday, afternoon | Migrate the regulars you know by name | Every name on the list has a count beside it |
| Monday, open | Go live, convert paper at the counter | A customer joins without you helping them |
Nothing on that list needs a quiet week or a consultant. The two rows that people skip are the Friday one and the Sunday test, and they are the two that produce every complaint in week one.
Saturday morning: three rules, not thirty
Keep the threshold you already had. If the paper card was ten stamps, make the digital card ten stamps. Changing the offer and the format on the same weekend means that when the numbers move you will not know which change moved them. Change the format now, change the offer in three months if the format worked.
Name the qualifying event in one sentence. One stamp per customer per day, regardless of what they spent, is the version a person on a busy till can apply in a second without a rule sheet. Anything involving a minimum spend needs the till and the loyalty screen to agree, which is a second job in the middle of service.
Decide who validates before you decide anything else. If a stamp or a redemption needs a manager, it will fail on exactly the shifts where nobody senior is in. Whoever is on the counter needs to be able to do both, which is what staff validation means in practice: the customer requests, the person serving approves.
Write those three lines on a piece of paper and tape it next to the till. That is your program documentation, and it should not need a second page.
Saturday afternoon: the sentence staff will actually say
Staff need one sentence, not a briefing. Give them this shape, with your own reward in it:
“Want a stamp? It’s the QR code here, takes about ten seconds, and your tenth coffee is free.”
And a second one for anyone holding paper:
“If you’ve got the paper card, give it here and I’ll put your stamps straight onto the new one.”
Two things not to say. Do not say “download our app” if there is no app, because the customer will look for one, fail, and blame you. Do not say “sign up”, which sounds like admin; “join” and “get a stamp” both sound like something that happens to them rather than something they have to do.
Then run the walkthrough properly, which takes fifteen minutes. Each person scans the poster with their own phone, joins as a customer, and gets a card. Then they swap to the staff side and validate a stamp for the person next to them. Someone who has been the customer once will explain it better than someone who has only been told about it.
Sunday: print it, then try to break it
Print the poster and place it where the customer has already stopped moving, which for most counters is beside the card machine rather than by the door. The counter placement guide covers the physical constraints, and the one that catches people is glare: a code behind glass or under a spotlight scans fine at your angle and not at theirs.
Then test the entire journey on your own phone, not just the scan. A successful scan proves the code is readable and nothing else. Go through it end to end: join, find the card again after closing the browser, present the code, validate a stamp from the staff side, and take a card all the way to a redemption so you have seen what your team will see when a reward is claimed. Try the typed-code fallback too, because on the morning a camera refuses to focus that is your only path.
Last job of the weekend, and the one that pays off fastest: migrate the regulars you know by name. Write the list, ask them next time or text the ones you have numbers for, create a card each and set the count to whatever their paper card says. Twenty or thirty names is a normal list for a single site, and it is worth doing by hand because those are the people whose habit you are least willing to interrupt. If your list is under thirty, most tools will carry it on a free tier; NeoLoyal’s stops at 30 customers, which is roughly one weekend’s worth of hand migration before anything is billed.
The two things that break in week one
The customer who lost their paper card
Decide the policy now, in one sentence, so that whoever is on shift can apply it without calling you: credit up to half a card, once per customer, no questions asked.
The arithmetic is why. Say your reward is a coffee that costs you 80p in beans, milk and cup. Half a ten-stamp card is 40p of goods, given away slightly earlier than you planned. That is the entire downside of believing someone. The upside of not believing them is a regular customer who has just been called a liar in front of the queue. There is no version of this where the 40p wins.
Cap it and record it, so the same person cannot do it monthly, and move on.
Staff who forget to ask
This is a prompt problem rather than a discipline problem. The poster faces the customer, which means nothing in the room faces the person who has to speak first. Put something where their hands already are: a sticker on the card machine, a line on the till screen, the three-rule sheet taped where they look when the order is done.
Then measure it instead of nagging about it. Count joins against transactions for the week rather than against footfall, and you have a number that tells you whether the sentence is being said at all. If it is near zero on one person’s shifts and healthy on another’s, that is a coaching conversation with evidence in it rather than a suspicion. The rest of that arithmetic is in the metrics worth tracking.
What to do on the end date
Put two dates in the calendar before you close on Friday. Day 14 is your first honest look at joins per transaction, and it is early enough to fix a poster in the wrong place. The second is the paper end date you announced.
On that date, take the rubber stamp out of the drawer and put it somewhere you cannot casually reach. A program with two ledgers only ever ends when one of them physically stops, and a stamp that stays next to the till will still be inking cards at Christmas. If you decide later that paper had something worth keeping, the three card formats are still there to be compared on purpose rather than by default.