Recovering 3.8 EBITDA points: how we dismantled the reach myth of promoting the restaurant on social media with the Restaurant Model Canvas and the Demand Radar

The starting point: 214,000 impressions and 11 reservations

The trattoria did not have a reach problem, it had a destination problem, and the opening figures settled the argument: 214,000 monthly impressions, 34,000 followers, two Reels a week and 11 reservations attributable to the digital channel in the measured month, against a marketing line of US$1,940 per month covering paid media, a photographer and aggregator commissions. The division is elementary arithmetic and it stings anyway: US$176 of customer acquisition cost per diner brought in, on an average ticket of US$23 in the dining room and US$29 in delivery. To size that disproportion the sector benchmark helps, because according to ChowNow (2025) the average paid CAC in quick service sits around US$27 and the organic one around US$9. A 14-table, 52-seat house, with 7.1% EBITDA and seven years of operation, was paying fine-dining acquisition prices to sell lunchtime pasta.

Why did all that reach never move Tuesday's cash?

Because 82% of the posts asked for nothing, and a piece that asks for nothing neither fails nor succeeds: it simply does not take part in the funnel.

Reviewing ninety posts from the last quarter we found plate photos with no visible address, no opening hours, no booking link and not one measurable call to action; nine empty tables at 9 p.m. on a Tuesday was no algorithmic mystery, it was a consequence of design. The contrast with real category demand is brutal, because according to Restroworks (2024) food near me searches grew 99% year over year and, according to Malou (2025), 79% of restaurant searches are non-brand. Meaning: people were out there looking for somewhere Italian to eat that very night while the house posted desserts without saying where it stands. The first thing we installed was not a campaign but an attribution bridge, because without it promoting the restaurant on social media is an OpEx expense justified by feelings.

The missing bridge between the reach report and the P&L

Using the Masterestaurant framework and the acquisition-cost calculator from the tools ecosystem, we set up three things in eleven days: a unique booking link per channel, a table code the host asked for when seating, and a weekly close where the content manager reported reservations and spend, not impressions. The rule I set was a single line, and it is the one that draws the most resistance in any operation: if a post does not lead to a trackable destination, it does not go out. Within four weeks the report went from seven vanity metrics to two numbers the owner could read standing up: attributable reservations and the ticket of those reservations. We cut posting frequency from seven pieces a week to four and the case result was counterintuitive: less volume, more reservations.

What changed in the content, and what we stopped doing

The underlying decision was to move the content manager's hours away from Reels production and toward the local listing and review responses, and that is the lever almost nobody wants to look at, because according to BrightLocal (2026) 76% of mobile near me searches end in a visit within 24 hours and according to Restroworks (2025) more than 60% of restaurant searches originate on mobile. Every post ended up carrying address, time window and booking link; the daily menu started going out at 10:40 a.m., not at 4 p.m. And yes, engagement fell close to 20% in the first month. Tuesday's cash went up regardless. By the fifth month, reservations attributable to the digital channel went from 11 to 63 per month, on marketing spend of US$1,610, that is US$26 of acquisition cost against the US$176 at the start. The important detail is not the CAC drop but where it came from: US$330 less in paid media and an internal recomposition of the spend, not a fresh injection.

The closing numbers at five months

Impressions fell to 158,000, down 26%, and that decline was the best sign that the money was reaching people who could actually walk through the door. The average Tuesday went from 21 to 34 diners between 7 and 10 p.m., and EBITDA closed at 9.4% versus 7.1% at the outset. To calibrate the achievement, look at the sector's ceiling: according to ChowNow (2025) paid CAC in fine dining runs close to US$180. Under US$500,000 a year, forget paid media this week and spend three hours completing your local listing with hours, menu and booking link; it is free, and it is where the non-brand demand that Malou (2025) measures at 79% actually happens. Between US$500,000 and US$1 million, the first step is exactly this trattoria's: calculate CAC by dividing last quarter's marketing spend by trackable reservations, even if the number embarrasses you.

Transferable lessons by annual revenue band

Above US$1 million, split the budget by channel and demand per-location attribution before approving the next month of paid media, knowing the sector's average CPC is US$2.05 according to PPC Chief (2026). Above US$5 million, the priority is a weekly CAC traffic light per site, with a written cut-off threshold. Above US$10 million, with large-format themed venues and a media-famous chef, the personality's reputation already brings the reach: what you must build there is booking capacity and occupancy control, not more content. I would not expect this result in three contexts, and I would rather say so than sell a universal method. First, in a restaurant under two years old with no review base: here there were seven years and accumulated local reputation, so conversion was dammed up, not absent; without that asset the work is construction and takes considerably longer than five months.

Limits of this case

Second, in operations where delivery carries more than 60% of sales, because the trackable destination is no longer your booking but the aggregator, and whoever controls the link controls the margin. Third, in cities above three million inhabitants, where the ad auction makes the click dearer and the US$2.05 CPC reported by PPC Chief (2026) falls short as a reference. And one condition that always applies: if the kitchen cannot hold a Tuesday service, filling the room only speeds up the reputational damage. This operation had no reach problem: 214,000 monthly impressions across a city of 700,000 is reasonable coverage for a 14-table trattoria. It had a destination problem. Fully 82% of posts asked for nothing, and a piece that asks for nothing can neither fail nor succeed, it simply never enters the sales funnel. Customer acquisition cost only starts existing once somebody calculates it.

What separated the myth from the cash in this operation?

The initial US$41 came from dividing marketing spend by attributable bookings, and that division was the uncomfortable moment of the whole audit:

per ChowNow (2025), average organic CAC in quick service sits near US$9 and paid CAC near US$27, so the trattoria was paying fine-dining prices for a casual-dining check. Online reputation was working like a tap open in reverse. With 41 unanswered reviews, the operation gave away the very signal Google uses to decide who shows up in non-brand searches, which per Malou (2025) account for 79% of restaurant searches. Every dead review was local traffic walking into the trattoria across the street. Delivery conversion had been outsourced to the aggregator, which meant guest lifetime value belonged to the aggregator rather than the restaurant. A customer who orders three times a month through an app and never leaves a contact is rented revenue; when commission rises two points, that revenue evaporates and there is no list left to write to.

What separated the myth from the cash in this operation — in practice?

Audiovisual content was produced outside the operation, in quarterly photographer sessions. The output was technically correct and commercially inert: dishes no longer on the menu, plating the kitchen cannot reproduce during peak hours, zero human faces.

Kitchen filming on a phone cost less and converted more. The 'post every day' myth carried one hour daily of an employee at US$5.20 an hour, roughly US$114 of invisible payroll that never made it into the P&L marketing line. Moving to four well-built weekly pieces halved that hour and multiplied the return per piece.