AI assistant · N2 Ready-to-use piece
Win-Back Campaign Builder for Restaurants
Builds the campaign that brings back guests who stopped visiting: defines who counts as dormant, segments by value, writes the message sequence and sets how to measure returns without giving away margin.
N2Ready-to-use piece
Produces a piece you can use as-is: a menu description, a service script, a supplier email. Copy and go.
◆ This assistant 7 of the 164 published assistants are at this levelIs this for you?
What it does and what you get
- Every restaurant loses guests silently: nobody complains, they just stop coming.
- This assistant turns your contact list into an organized win-back campaign, with messages in your brand's voice, offers sized to each guest's value and an honest way of knowing how many actually came back.
What you end up with
A complete win-back campaign: cause diagnosis, segment table with offers and costs, three ready-to-send messages, a send calendar and three tracking indicators.
EXAMPLE This is what you getAn example built with Sazón de Origen, the house's fictional restaurant.Not your data: your result is built with YOUR restaurant's.See the full example →Sample result
The owner asked this: "I have around fifteen hundred own-delivery phone numbers in a spreadsheet and I cannot tell which ones stopped ordering. Do I send everyone a discount, or how is this done without giving the money away?"
Your house has no customer database. It has the own-delivery order book, and that is enough to start. Of the three channels, only one records who ordered and when: every own-delivery slip carries a name, a phone number, an address and a date. The delivery platform holds the rest and does not hand them over: nobody there gets a message, you use the inactive-customer coupon in its partner panel and the return is read inside the app. The dining room moves 62% of the tickets (assumed) and records nothing: nothing to recover there, because nothing was ever written down.
What to start recording on Monday, in the same delivery spreadsheet: name, WhatsApp with permission asked out loud, date and what they ordered. Captured when the check is closed, not at the door.
Twelve weeks of order slips give a median cycle of 21 days between orders from a repeat guest. Twice that is 42, and that is the cut-off: of 1,460 unique numbers across twelve months, 612 are still active, 468 are dormant and 380 have been away more than 90 days.
| Tier | How many | Last visit | Reason being answered | What is offered | Margin given up |
|---|---|---|---|---|---|
| A · The regulars, 6 orders a year or more | 74 | 43 to 90 days | The craving for home cooking on a weeknight | Sancocho de tres carnes on the house to share, Tuesday to Thursday | 11,800 COP per visit |
| B · The occasional ones, 3 to 5 | 163 | 43 to 90 days | The same one, without the closeness to phone them | Limonada de la casa on the house on the first visit back | 4,140 COP per visit |
| C · The one-timers, 1 or 2 | 231 | 43 to 90 days | None yet: they never made this place their own | Nothing. One touch, no offer | 0 COP |
Why the sancocho and not the bandeja. Bandeja de la casa is the best seller and costs 14,300 COP against 42,000: 34% food cost, the menu average and two points above the 32% ceiling any dish on promotion has to meet. The problem there is not the offer, it is the menu price: at 45,000 it would sit at 31.8%, inside the range you already charge. The sancocho costs 11,800 against 38,000, lands at 31% and does fit.
Illustrative figures, built from the data you gave.
Two axes you can actually count in the spreadsheet you already have: what the guest spent over twelve months and how many days they have been away. The cross gives six cells, and each one decides differently.
| Dormant · 43 to 90 days | Lost · more than 90 days | |
|---|---|---|
| High · 6 orders or more | 74 · The full campaign. Three touches, the first signed by you with your own name, the third carrying the sancocho courtesy. It is the only tier where touch 3 is backed by a phone call. | 61 · Not a campaign, a phone call. One by one, from the owner, with no offer. Here you do ask what happened: a call to 61 people is not a mass survey, it is a conversation the house earned. |
| Medium · 3 to 5 orders | 163 · Three touches, small courtesy. The limonada de la casa on the first visit back. No phone call: someone who ordered four times in a year finds the phone intrusive. | 142 · Out of this campaign. They come back in when the house has something new to tell, a different menu or a new schedule, not a promotion. |
| Low · 1 or 2 orders | 231 · One touch only, no offer. Whoever replies moves into tier B and gets the other two. The rest receive nothing further: every base churns on its own and that is not a leak. | 177 · Nothing. One hundred and seventy-seven people who will not be written to, stated with the number so the decision is made once and not reopened every month. |
The judgement, declared and arguable: effort is handed out for what the guest already left behind, not for what you hope they will leave. The concession: if own delivery were just starting and the list held two hundred names, tier C would get all three touches, because there would be nowhere else to draw from.
They go out on WhatsApp from the own-delivery number, which is where the guest left the detail and where they are already served. Instagram stays out: there is an audience there, not that person's permission.
| Touch | Day | Opening line | Call to action | What it does not carry |
|---|---|---|---|---|
| 1 · Noticing the absence | 1 | "It has been a while since you ordered and I wanted to tell you something" | None. It closes by saying no reply is needed | No offer, no link, no questions |
| 2 · The reason and the dish | 8 | The dish that guest used to order, taken from their last slip | "Reply with a time and we send it over" | No discount |
| 3 · Something on the table | 15 | The tier's courtesy, with a closing date | "Until Sunday the 20th, tell me the day" | No second round, no renewal |
Touch 1 for tier A, sent as written with the name and date swapped:
Don Jorge, how is everything? I checked the order slips and your last one was 14 June; it had been a while since I saw you around. Something changed here: since July two bikes go out on weekdays and the food arrives in 35 minutes, not in an hour, which is what was happening to us at night. I am not selling you anything today. I just wanted you to know.
Touch 2, with the dish from their slip:
Don Jorge, there is sancocho de tres carnes today, the one from the three-hour pot you used to order on Thursdays. If you want it, reply with a time between 6 and 8 and we send it over hot.
One condition before copying touch 1. That message only goes out if the house genuinely fixed something. If evening delivery is still the same, touch 1 says nothing about improvements: it says you remembered him and stops there. Claiming the experience changed when it did not is what turns a dormant guest into a lost one.
Illustrative example generated with a fictional restaurant. Not a promise of results, not a client case.
This is 3 of 5 parts. Behind the wall sit the two parts that get executed. The twenty-one-day calendar row by row, with day, what goes out, to which tier, how many messages, owner and success signal: tier A's 74 signed by you on day 1, tier B's 163 split across two days so the whole list does not reply on the same Tuesday, the 3:30 to 5:00 p.m. window that dodges the full lunch service, and the day-10 cut-off at 9:00 a.m. that decides whether the third touch carries a courtesy or becomes a phone call. And the six-week board with its four columns —came back, cumulative, margin given up and cost per recovered guest—, the 31,680 COP ceiling per recovered guest that switches the courtesy off halfway, and the three boxes ticked every Monday at nine.
See the full example — free accountWorks with these AIs
How to use it
Which of your data it uses
This assistant works with 8 of the 23 fields in «My Restaurant», among them Restaurant name, Your main guest and Why they visit (consumption reasons). It does not ask out of curiosity: these are what make the answer speak about YOUR scale instead of an industry average — the same calculation on a business of another size returns a number you cannot decide with. You fill them once and they apply across the library, so the second assistant you open already starts with them in place. Whatever you leave blank, the assistant asks for one thing at a time instead of inventing it, and says so before answering, so you know what it is working from.
Always up to date
Every assistant carries a visible version and date. When an AI changes how it works, the assistant's version goes up and you see what changed.
v1.3.02026-08-25Card completed (doc 11): who it is for, a sample result with the fictional restaurant Sazon de Origen, and recommended AIs. The assistant body did not change.v1.2.02026-08-20Rooting: the TASK stops talking about concepts and works with the house data. Step one starts from the reasons for visiting -why they came- and step four sets the absence against the promise, because whoever left needs a reason to forgive, not a promotion; the three touches go out in the brand voice and through the sales channel the guest already allowed, the invitation lands on a named signature dish and the average check sets the spending ceiling. The format reads the return against the main customer and the next step is signed with the restaurant name: eight fields across the working sections, six inside the TASK. Body at 793 words, duel still at 100.0.v1.1.12026-08-14Spelling: Spanish accents restored in the prose. No field marker, link or meaning changed.v1.1.02026-08-13Body rewritten under the margin mandate: 6 win-back campaign steps, 5 named output blocks with columns, 8 business fields, named trade traps, verification checklist and visit-cycle and measurement figures with units.v1.0.02026-08-11The assistant is born.Same method, another variant
Inside the app this is a single control: you press it and the assistant reframes itself, including the variants that don't appear here because they don't have their own page yet.
More assistants in this category
Designs a complete campaign end to end: money objective, audience, message, offer, channels and a 4-week calen…
N3 · Analysis with a verdictBuilds delivery campaigns on two lanes: selling more inside the platforms without melting margin into commissi…
N3 · Analysis with a verdictDesigns your loyalty program as a recognition ladder: steps with clear signals, house gestures that cost littl…
Included in the always-growing library
Access to every published assistant, adapted to your AI and personalised with your restaurant's data, updates included.
