From $18.40 to $6.10 CAC: how we unclogged the paid advertising bleeding a trattoria's EBITDA, using the Restaurant Model Canvas and the Demand Radar

The opening picture: $9,400 a month without a single line of incremental sales

Fourteen months of ad spend with no incremental-sales measurement had left this 46-seat trattoria with $812,000 in revenue and a dry till. The split ran $9,400 monthly across Meta Ads, Google Ads and paid promotions inside the aggregators, meaning $112,800 a year against a $27.50 average ticket, and nobody could say which diner would have walked in anyway. The 19 employees produced well, the dining room held 66% of sales and owned delivery plus aggregators already moved 34%, yet every ad dollar bought cold traffic for an operation that did not retain. The sector pushes in that direction: according to Restroworks (2025), 99% of restaurants keep at least one social media profile, so simply being there stopped being an advantage a long while ago. Before touching a single bid, we split the accounting line called "marketing" into three buckets, each with an owner and its own KPI: cold acquisition, remarketing and content production.

The diagnosis nobody wanted to hear: 61% of spend chased customers already coming

That purely accounting cut exposed that 61% of the $9,400 monthly chased people whose booking was half-decided already —brand searches, old followers, customers on the list—, meaning $5,734 a month buying what was already his. The second measurement covered contribution margin per dish and 60-day repeat rate, because optimizing creative without those two numbers is expensive superstition. Here I was wrong for years: I believed you fix the ad first and the table afterwards. It runs the other way. Paid media MULTIPLIES, and it multiplies what is broken too, except that you pay to accelerate your own bleeding. The Masterestaurant contribution-margin and repeat-rate board turned an argument of opinions into a cash decision. Diego F. Parra applies it always in the same order: first you cross each dish's unit margin against its weekly rotation, then you cross the customer base against its 60-day repeat rate, and only then do you decide where ad money goes.

The Masterestaurant tool that settled the decision: the margin and repeat-rate board

In this trattoria the exercise revealed that eleven dishes concentrated 74% of the margin while the menu listed thirty-one, and that 60-day repeat sat at 23%, far under the full-service benchmark: according to Paytronix (2024), top full-service restaurants retain 57.8% of their members month over month. With those two figures on the table, cutting spend stopped being frightening. The operational turn was shifting 46% of the budget out of pure acquisition and into remarketing plus in-house production. The kitchen began filming the plating of two dishes a day with a phone and a $40 tripod —a one-time outlay, not recurring—, and those pieces fed organic and paid remarketing at once, which stopped showing generic catalogs and started showing the dish leaving the pass. Competitive context explains why producers win: according to Restroworks (2025), 78% of restaurants use Instagram, so the edge no longer lies in having a profile but in holding a publishing rhythm with your own material.

Manufacturing demand instead of buying it: two dishes a day on a $40 tripod

We also moved the campaign objective from "profile traffic" to "confirmed reservation", and cost per reservation fell from $11.80 to $6.40 in nine weeks. With cold spend trimmed, the freed money went to the channels that speak to whoever already ate in the house. We built an email sequence over the base of 4,100 registered diners and an SMS flow for confirmation and reactivation, the two vehicles with the best economics in the sector: according to Stripo (2025), restaurant email marketing returns $36 for every dollar invested, and according to Constant Contact (2024), SMS reaches roughly 98% open rate, with 90% read within the first three minutes. We added a plain rewards program, no proprietary app and no plastic card, because according to Paytronix (2025) more than 90% of restaurants already run one and staying out costs more than arriving late. The 60-day repeat rate went from 23% to 34.6% in five months.

The measurable outcome: less investment, more sales, more margin

Monthly investment dropped from $9,400 to $5,800 and sales climbed 21.3%, precisely the opposite of what any account manager promises. In case figures: $3,600 less in ad spend per month, meaning $43,200 a year that stayed inside the operation, with revenue moving from $812,000 to a projected $984,000 over twelve months. Cost per reservation fell 45.8%, the 60-day repeat rate gained 11.6 points and the average ticket rose from $27.50 to $29.10, because the eleven highest-margin dishes started appearing in the content rather than in a thirty-one-line menu. What would have happened without the cut? Holding $9,400 with the same 23% repeat rate, the trattoria would have bought the same table volume one more year, burning $43,200 on single-visit customers. What transfers is not the cut, it is the order: retention first, paid media after.

Transferable lessons by annual revenue band

Under $500,000 a year: this week measure your 60-day repeat rate on the last 300 tickets and pause every cold acquisition campaign until you hold that number. Between $500,000 and $1 million, this case's band: split the "marketing" line into three buckets with an owner and a KPI, and move 40% into remarketing. Above $1 million: assign a content owner with a weekly calendar and measure cost per confirmed reservation, never per click. Above $5 million: negotiate aggregator rates against volume and pass the saving into your own base. Above $10 million, multi-site groups or celebrity-chef formats built on a personal brand: consolidate the diner base into one repository before spending another dollar, because each location buying its own audience cannibalizes the rest. I would not expect these numbers in three contexts, and it is worth saying so before someone copies the cut blindly.

Limits of this case

First, a recent opening with no customer base: if you hold neither 4,100 registered diners nor a repeat rate to measure, cutting cold acquisition leaves you without flow, and there paid media truly is the tap that sustains cash through the first twelve to eighteen months. Second, operations with structurally low contribution margin by menu design, where no retention gain offsets a food cost already sitting at the 32% ceiling. Third, markets where the dominant channel is the pure aggregator, with 70% or more of sales outside the dining room: there your own remarketing collides with a customer relationship you do not own. The case is an anonymized composite of patterns observed across more than 8,400 restaurants in 43 countries. The first change was accounting, not advertising: we split spend into three buckets with an owner and a KPI each —cold acquisition, remarketing, content production— because until then everything landed on one line called «marketing» that nobody could read.

What genuinely changed between month 0 and month 5?

That move alone exposed 61% of the budget chasing people who were coming anyway. We stopped buying traffic and started manufacturing demand.

The kitchen team began filming two plated dishes a day with a phone and a $40 tripod, and those pieces fed both organic reach and paid remarketing; per Restroworks (2025), 78% of restaurants use Instagram, so the edge no longer lies in showing up, it lies in producing at your own cadence. We switched the campaign objective from «profile traffic» to «confirmed reservation» and accepted losing reach. Reach dropped 44%, attributable reservations rose 88%, and that made it obvious we had been paying for vanity measured in impressions. We built an owned base. From 410 loose records to 3,940 opted-in contacts, segmented by frequency and channel; email stopped being a monthly newsletter and became a repeat sequence firing at 14 and 45 days. Delivery left the aggregator without abandoning it.

What genuinely changed between month 0 and month 5 — in practice

We kept presence for discovery but moved repeat orders to the owned channel with a QR menu sitting alongside the physical menu —the printed menu governs service pace and suggestive selling, the QR handles delivery, accessibility and price updates— and effective commission on total sales fell from 8.9% to 4.3%.