Pick one reward rule, such as a dessert and coffee on every sixth visit, and track visits in a sheet or your till's customer list, with each member's first name, contact detail and a recorded opt-in. Use an AI assistant to write three offer types from a fact slip, then measure repeat visits over 90 days.
No app is needed, and the reward costs less than it looks: in the worked example below, a free dessert and coffee on the sixth visit costs about 1% of the spend that earned it. Two rules protect you. Membership is not permission to send offers, so the form asks for each separately, and a reward must never be tied to a review, which Google's policy bars.
Points or visits: pick a rule your staff can say in one breath
Every loyalty scheme is one of two rules. A visits scheme says "every sixth visit earns a dessert and coffee". A points scheme says "one point per dollar, and 400 points buys $10 off". Say a rule aloud to a server on a busy Friday. If it takes more than one breath, guests will not understand it either. The cost comparison below uses a table spend of $62 (two guests at $31 each) and a food and drink cost of 30%; use your own figures.
| Rule | Value to the guest | What it costs you per reward | Spend needed to earn it | Give-back |
|---|---|---|---|---|
| Visits: every 6th visit earns a dessert and coffee | $12 on the menu | $3.60 in ingredients (30% of $12) | 6 visits at $62 = $372 | 1.0% at cost, 3.2% at menu price |
| Points: 400 points ($1 each) buys $10 off | $10 off the bill | $10, because a discount is revenue you give up | $400 (about 6.5 visits) | 2.5% |
A free item and a discount feel similar to the guest but cost you very different amounts. The dessert and coffee is worth more to the guest ($12 against $10) and costs you $6.40 less. At 20 rewards a month that is $128 a month, which is why the setup below uses the free-item version. One caveat applies. The $3.60 assumes the guest would not otherwise have bought that dessert. If they would have, you have given away the full $12, so choose a reward people do not usually order, such as a sharing starter or a coffee with the main, and watch what your sheet says.
The visits rule has one weakness: a $20 solo lunch counts the same as a $200 dinner. Fix it with a minimum spend per stamp, such as "one main course or $15 a table", and write it on the card. Points fix this automatically but are harder to explain and harder to track by hand. A points scheme suits a restaurant with a wide spread of spend, such as large groups and bar tabs, particularly when the till tracks spend for you. A visits scheme suits a small team with similar tickets and a sheet.
Set the visit count from your own data. If a typical regular comes twice a month, six visits take three months to earn, which is already a long wait. If your regulars come once a month, six visits is half a year, and four would suit better. This is my judgement, not a measured rule, so check your sheet after the first quarter. If you also sell gift cards, the gift card programme tutorial shows how the two schemes can coexist without confusing guests.
A no-app setup: a sign-up form, a sheet and a handful of controls
Guests join through a QR code on the table card and at the till. It opens a short form asking for a first name, a phone number or an email address and a set of unticked boxes for the scheme and for offers (consent is covered in the next section). The answers land in a sheet, one row per member. A filled-in version of the columns, with dummy contact details, looks like this:
| Member | First name | Contact | Joined | Offers opt-in | Lifetime visits | Last visit | Rewards earned | Rewards used | Ready |
|---|---|---|---|---|---|---|---|---|---|
| M001 | A. | guest-a@example.com | 2026-09-12 | Yes, email, 2026-09-12 | 7 | 2026-10-27 | 1 | 0 | 1 |
| M002 | B. | 555-0101 | 2026-09-19 | Yes, text, 2026-09-19 | 3 | 2026-10-08 | 0 | 0 | 0 |
| M003 | C. | guest-c@example.com | 2026-09-26 | No | 13 | 2026-10-25 | 2 | 1 | 1 |
Three formulas do the counting once the sheet has lifetime visits in column F and rewards used in column I:
H2 =INT(F2/6) rewards earned
J2 =H2-I2 rewards ready to claim
K2 =MOD(F2,6) visits towards the next rewardLogging works like this. At payment, the server asks for the member's name or the last digits of their phone number and writes it on the bill. The closing manager adds one visit per member per table from the day's bills, which takes about ten minutes. A "Ready" cell above zero is the only thing that allows a server to comp a reward, and the manager records the date when it is used. If your till can attach a phone number or email to a sale, it counts visits for you, so look at its customer report before building the sheet. Square Loyalty is one paid example: its page says guests can earn points by visits, spend or items, enrol at the point of sale and try it for 30 days, though I found no published price on the page I read in October 2026.
A digital stamp card is the next step up. Loopy Loyalty's pricing page lists Starter at $25 a month (the page shows a $ sign without naming the currency) for one card design and one location, with unlimited customers, Apple Wallet and Google Wallet support and a 15-day trial; the page does not say whether customers must install anything, so ask before you commit. I would run the sheet for a quarter first. By then you know how many members you have and how long the logging takes, and you can compare the two honestly.
Controls matter more than software. One stamp per table per visit, not per person. A minimum spend written on the card. Staff initials against each logged visit. A weekly check that the number of rewards used never exceeds the number marked ready. Rewards that expire 90 days after they are earned, with the date stated. The AI assistant can build the formulas and conditional formatting for the sheet if you describe your columns, but test them on three rows by hand.
Consent comes in two parts: membership and offers
Joining a rewards scheme means you hold a name, a contact detail and a visit count for each member. Keeping that count is what lets you give them rewards, so the scheme's terms can say so plainly. Sending marketing offers is a separate act. Someone who joined to collect stamps has not necessarily agreed to texts about Tuesday specials. So the form carries separate unticked boxes, and a member can say yes to the scheme and no to offers:
[ ] Join the rewards scheme: every sixth visit earns a dessert and coffee on us. We keep your first name, your phone number or email and a count of your visits so we can give you your rewards.
[ ] Also send me offers and news by email.
[ ] Also send me offers by text. You can stop at any time.
Record when and how each person opted in, as in the sheet above, and keep the wording version too. Collect the minimum: a first name and one contact detail. If you want a birthday offer, ask for the birth month and nothing more, since a full date of birth is more than you need. Texting rules differ from email rules in many markets, so ask for text consent in its own tick-box, put "Reply STOP to opt out" in every text, and ask an adviser if you are unsure. The Mailchimp consent and unsubscribe checklist and the weekly guest email tutorial cover the email side in more depth. None of this is legal advice.
A short terms card closes the loop. Ask the assistant to draft it from your decisions, then read every line, because it will add promises you have not made:
1. One stamp per table per visit, added at payment, with a minimum spend of $15. 2. The reward is a dessert and coffee, with no cash value, one per visit, and cannot be combined with other offers. 3. Rewards expire 90 days after they are earned. 4. We may change or end the scheme with 30 days' notice, and rewards already earned will be honoured for 30 days after any end. 5. We use your details only for the scheme and for any offers you chose to receive, and delete them if you ask. 6. Rewards are never linked to reviews of any kind.
Three offer types, with the cost of each worked out
Most of the value in a scheme sits in three offers, each with a different job. The numbers below use a gross profit of $43.40 per table, which is the $62 spend at a 70% margin.
| Offer | Who gets it | What they get | Cost per redemption | Break-even share of extra tables |
|---|---|---|---|---|
| Earned reward | A member on their sixth visit | Dessert and coffee, $12 on the menu | $3.60 | One extra visit per 12 rewards |
| Quiet-night offer | Members, Tuesday and Wednesday only | A sharing starter ($9) with two mains | $2.70 | 6.2%, about 1 in 16 |
| Win-back offer | Members with no visit in 60 days | A sharing dessert ($10) with two mains, valid 21 days | $3.00 | 6.9%, about 1 in 14 |
The break-even column is the useful one. If an offer costs $2.70 and a table returns $43.40 of gross profit, the offer pays for itself when about one redemption in sixteen is a table that would not otherwise have come ($2.70 divided by $43.40 is 6.2%). Likewise, a $3.60 reward is repaid by one extra visit for every 12 rewards given ($43.40 divided by $3.60). None of the three needs miracles. What they need is the right audience, and that is where they differ. The win-back offer goes only to members who have been away 60 days, so most of its redemptions are real extras. The quiet-night offer goes to members generally, so many redemptions will be people who would have come anyway. That is why it is limited to the nights when you have spare tables, where the downside is a free starter and not a displaced paying table.
Keep each offer simple enough to claim without a manager. A reward that needs a conversation at the table will be forgotten, and forgotten rewards cost nothing but earn nothing either. Print the offer's end date in the message, and put the end date on the sheet as well, so staff can see a stale one at once.
One slip, three drafts: writing the offers with AI
The same fact slip idea from the guest email tutorial works here. Fill it in once, and the assistant drafts all three offers in your voice:
SCHEME: rewards, every sixth visit earns a dessert and coffee on us.
OFFER 1 (earned): sent when a member has 5 visits, to say one visit to go.
OFFER 2 (quiet night): members only, Tuesday and Wednesday, until 2 December. A
sharing starter on us with two mains. Show this message.
OFFER 3 (win-back): members with no visit in 60 days. A sharing dessert on us with
any two mains, valid for 21 days from the day they receive the message.
CHANNEL: for each offer, an email (subject line and 60 words) and a text of 160
characters or fewer ending with "Reply STOP to opt out".
BOOKING: [booking link]
NEVER SAY: "VIP", "exclusive", "double points" (we have no points), "limited spots",
any allergen claim, any price not listed here.
VOICE: warm, plain, a little dry. British spelling. Use [first name].A first reply for Offer 2 (illustrative) looks like this:
Subject: Your exclusive VIP Tuesday treat
Hi [first name], our members get double points on Tuesday and Wednesday until 2 December, plus a sharing starter on us when you order two mains. Spots are limited, so book now at [booking link]. We look forward to seeing you!
Text: [Restaurant]: Exclusive VIP offer! Double points on Tue and Wed until 2 Dec plus a free sharing starter with two mains. Limited spots, book now: [link] Reply STOP to opt out
It breaks your own rules in several places, which is usual in a first pass:
- "Exclusive VIP" and "limited spots". Both were on the never-say list. Delete them. Invented scarcity is the quickest way to teach members that your offers are noise.
- "Double points". You have no points. The assistant imported the standard loyalty vocabulary. Replace it with the reward you actually have.
- The text is 173 characters. Past 160 characters a text is usually sent and billed as two segments, and an emoji or curly quote lowers the limit further, so check your provider's rules. A corrected text of 132 characters reads: "[Restaurant]: Tue and Wed until 2 Dec, a sharing starter is on us with two mains. Show this text. Book: [link] Reply STOP to opt out". Count again once your restaurant's name and the real link replace the placeholders, since a long link can push it back over.
- "Show this message" is missing. The slip said guests should show it, and the draft dropped the instruction, so staff would not know how to verify the offer.
The win-back draft differs in one way worth copying. It should open with what is new since they were last in, not with guilt. For the line servers say at the table, the server scripts tutorial shows how to build short scripts with AI and rehearse them at a pre-shift huddle; the same method suits a one-line invitation to join.
The reward you must never offer: points for a review
AI drafts love to add "leave us a review for a free coffee". Strike it every time. Google's contribution policy lists offering incentives, such as payment, discounts or free goods and services, in exchange for posting a review, and it lists selectively soliciting positive reviews from customers. Both are barred. A reward tied to a visit is fine; a reward tied to a review is not, whatever the review says.
The safe pattern is to keep the two apart completely. Rewards are for visits. If you want more reviews, ask neutrally after the meal, to everyone, without mentioning the scheme, and keep your Google profile current so guests find the right information; the Google Business Profile tutorial has a weekly routine for that. When a draft or a well-meaning server mixes the two, the fix costs nothing, and the downside of leaving it in is removed reviews or a restricted profile.
Measuring repeat visits without fooling yourself
Three numbers are enough. The repeat rate is the share of members with two or more visits in 90 days. Visits per member per month shows frequency. Reward cost per extra visit shows whether the scheme pays. The trap is selection: the guests who join tend to be your best customers already, so their visits would have been high anyway. A fair comparison needs two lines, not one. Compare members' visits in the 60 days before they joined with the 60 days after, and set that change against the same change for similar guests who did not join, such as regulars from your booking history. Illustrative figures for a 50-seat restaurant after 13 weeks:
| Line | Figure |
|---|---|
| Members after 13 weeks | 240 |
| Members' visits a month, 60 days before joining | 1.5 |
| Members' visits a month, 60 days after joining | 1.7 |
| Similar non-members, same weeks (seasonal change) | 1.4 to 1.5 |
| Uplift attributable to the scheme | 0.1 visits a member a month |
| Member-months (joined gradually, so about 1.5 months each) | 240 x 1.5 = 360 |
| Extra visits over 13 weeks | 360 x 0.1 = 36 |
| Gross profit from extra visits | 36 x $43.40 = $1,562 |
| Rewards given: 31 at $3.60 | $112 |
| Quiet-night redemptions: 60 at $2.70 | $162 |
| Win-back redemptions: 22 at $3.00 | $66 |
| Your time: 13 hours at $25 | $325 |
| Total cost | $665 |
| Net | $897 |
Count member-months, not members. Someone who joined last week has had one week to change their habits, so multiplying all 240 members by three months would have doubled the estimate to 72 visits. The scheme pays for itself if it brings roughly 15 extra visits in the quarter ($665 divided by $43.40), and the example estimates 36. But that 36 depends on a 0.1 uplift, and 0.1 is a small number measured on a small group. Treat it as an estimate with wide error bars. A better test of the sums is to ask whether the break-even of 15 visits looks easy or hard for your restaurant: one extra visit a week from 240 members is a modest bar. Keep a count of rewards earned but not yet used as well, because they are a cost you have promised and not yet paid.
If the numbers disappoint, change one thing at a time. Lengthen the visit count if rewards are too cheap to earn. Tighten the quiet-night offer if redemptions are mostly from people who would have come anyway. Send the win-back sooner if the 60-day line is too late. The barbershop tutorial on testing three reward schemes runs the same one-change-at-a-time comparison on another trade's numbers.
Launch week, day by day
A scheme this small needs a week, not a project. Here is the order I would use:
- Monday: decide. Choose the rule, the minimum spend, the reward and the expiry, and work out the cost of the reward at your food cost. Ten minutes with the table above.
- Tuesday: build. Make the form and the sheet, add the three formulas, and test them with three dummy members.
- Wednesday: write. Ask the assistant for the terms card, the table card and the first offer drafts, then correct them against your never-say list.
- Thursday: brief the team. Print the staff script and the controls, and run a two-minute rehearsal at pre-shift.
- Friday to Sunday: soft launch. Invite the first 40 members at payment, check the sheet each night, and fix what the first weekend shows.
The staff script can be one line: "We have a rewards card: every sixth visit, a dessert and coffee on us. Can I take your number? There is a tick-box if you would like offers too." If a server says it to every table for a week, the first 40 members will arrive without anyone being pushy. Log the number of tables asked and the number who joined for the first week, and you will know your sign-up rate before spending anything on a tool.
Questions owners ask before launching a rewards scheme
Is a visits scheme or a points scheme better for a small restaurant?
A visits scheme is simpler for staff and guests, and a free item costs you ingredients rather than the full price a discount gives away. Points suit restaurants with a wide spread of spend, such as groups and bar tabs, especially when the till tracks spend automatically. If you use visits, add a minimum spend per stamp so cheap visits do not earn full rewards.
Do I need a paid loyalty app?
Not to start. A sign-up form, a sheet and a few controls can run a scheme for the first few hundred members. Square Loyalty and Loopy Loyalty are two paid routes, with Loopy's Starter plan listed at $25 a month in October 2026. Compare them once your sheet shows how many members you have and how much time logging takes.
Can I give a reward for leaving a Google review?
No. Google's contribution policy bars offering incentives such as discounts or free goods in exchange for posting a review, and it also bars selectively soliciting positive reviews. Keep rewards tied to visits only, and ask for reviews separately, neutrally and without any reward attached.
Can I text members about offers?
Only if they agreed to texts, and the rules for texting differ from email in many markets, so check them before you send. Ask for text consent in its own tick-box, include opt-out wording such as "Reply STOP to opt out" in every message and keep the consent record. Ask an adviser if you are unsure.
Further reads
- Restaurant Influencer Marketing: Shortlist and Brief Creators With AI — Bring in new guests who can then join the scheme.
- AI Marketing for Restaurants: A Four-Week Plan Owners Can Follow — Where a loyalty scheme sits among your other channels.
- AI Software for Restaurants: POS-Built AI or an Add-On Tool? — Whether your till's own tools can count visits for you.
- What Are Diners Really Saying? Using AI to Analyse Your Reviews — Learn what guests say about the scheme and the service.
- Restaurant Menu Price Increase: Test Each Change by Dish With AI — Check that reward items still earn their place on the menu.
- Customer Loyalty Programmes for Retail Stores: Can AI Design One? — A small shop's loyalty programme drafted with AI: points, stamps and tiers costed line by line, plus the five corrections the draft needed.
- Food Truck Marketing: Build an Email and SMS Regulars List With AI — A window QR sign-up, consent wording and three templates (welcome, weekly location text, monthly email) for a food truck's email and text list.
- Pizza Shop Marketing: Fill Slow Nights With AI-Written Offers — Find the slow nights in your orders, have AI write five offers, then test each one's margin before it reaches a customer.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: Square Loyalty and Loopy Loyalty product and pricing pages, and Google's contribution policy on incentivised reviews (checked October 2026); all restaurant figures are illustrative.