Paid advertising in restaurants: myth vs reality — the unit-economics white paper your board will ask for

Chapter 1 — The campaign dashboard and the P&L measure two different things

A 4.2 ROAS does not mean advertising is paying the rent, and that confusion drains cash every month. The platform reports gross attributed revenue; the bank receives contribution margin after marketplace commission, packaging and cost of goods sold. When 71% of attributed orders come through an aggregator taking its cut, the revenue shown on screen reaches the P&L shrunk by 60% to 70%. That 4.2 turns into something near a real 1.4, and at that level you are buying volume at cost. The fix is not switching campaigns off: it is changing the metric that governs the decision. The prepared-food delivery segment in the United States moved around 96 billion USD in 2024 according to Statista, and that volume sustains an auction where thousands of operators bid for the same seven-o'clock diner. Define first how much margin a visit leaves, then decide how much you can pay to bring it in; the reverse order is the costliest mistake I see on restaurant marketing dashboards.

Chapter 2 — Budget derives from contribution margin per visit, never the other way around

With an average check of 28 USD and a 30% food cost, gross contribution lands near 19.60 USD per visit before touching any variable service cost. If sector retention averages 55% according to Restroworks (2025), that diner returns a bit more than twice, pushing the value of the relationship above 40 USD of accumulated contribution. Only with that figure in hand can you set an acquisition ceiling: a 12 USD CAC is an excellent business, a 30 USD one is a leak dressed as growth. The Masterestaurant framework treats paid advertising as an OpEx line with a measurable return requirement, not as an image expense. The second diner costs nothing to acquire and leaves the same margin as the first, and there sits the only arithmetic that makes a paid media program profitable in foodservice. Where the agency reports CPM and reach, the operator with judgment chases the return rate of the customer already acquired.

Chapter 3 — Repeat frequency: the lever no agency optimizes

With sector retention at 55% according to Restroworks (2025), every point you add above that baseline multiplies the return on advertising spend that already happened and is already paid. An owned messaging channel changes the equation brutally: SMS records roughly 98% open rates with 90% read within one to three minutes according to Constant Contact (2024), and a 45% response rate against email's 6% according to Omnisend (2025). Diego F. Parra has spent twenty years correcting owners who mistake reach for demand, and the conversation always circles back to frequency. Moving orders to your own channel does not reduce acquisition cost, it amortizes it over a larger check, and that nuance decides whether the campaign survives the quarter. Guests order 35% more items per check on first-party platforms than through third parties, according to Paytronix (2024), a finding Lightspeed (2025) confirms in its reading of the digital check.

Chapter 4 — Owned channel versus marketplace: the same CAC spread over structurally different checks

On a 32 USD base order, that gap adds roughly 11 USD of extra sales arriving with no aggregator commission, which means almost entirely into contribution margin. Add that Google Ads conversion in restaurants and food reaches 7.1% according to WordStream (2025) and the full argument is there: paying for traffic that lands in your own ordering system yields twice what paying for traffic landing in someone else's storefront does. A funnel without contact capture is open at the bottom. Workable budget does not scale linearly with sales, and confusing that destroys small operations. Below 500 thousand USD a year, 400 USD monthly represents close to 1% of sales and is justified only when concentrated within a three-kilometre radius with mandatory contact capture; with no owned list, that money evaporates. Between 500 thousand and 1 million, the healthy range runs from 1.5% to 2%, some 1,200 to 1,600 USD, already split into two campaigns by daypart.

Chapter 5 — What each revenue band can afford, with a different figure in each one

Above 1 million the problem opening this document appears: there is budget enough to reach a 4.2 ROAS and no margin discipline to read it. Above 5 million the mix professionalizes and influencer marketing enters as its own line within a global market that passed 33 billion USD in 2025 according to Socially Powerful. Above 10 million, with an in-house media team, the governing metric becomes cost per retained guest. Above 5 million USD a year two archetypes appear whose cost structure breaks sector averages, and applying a casual dining benchmark to them is a diagnostic error. The celebrity-chef restaurant carries a talent cost —schedule, press, audiovisual production— that consumes budget without generating reservations by itself; its paid advertising works as an amplifier of notoriety that already exists, and there the return gets measured in occupancy of off-peak slots, not in clicks. The large-format themed venue carries scenery, maintenance and performance staff, with a payroll that can double that of an equivalent format by seats.

Chapter 6 — High end: the celebrity chef and the large-format themed venue pay other costs

For both, social discovery outweighs search: Instagram records 2.2% engagement against Facebook's 0.22% according to Restroworks (2025), and 51% of TikTok users dine out because of a restaurant's content. Every visit that arrives through paid advertising and leaves without dropping a data point forces you to buy that same person again next time, and that is the silent drain in nearly every media budget I audit. Capture turns a one-round expense into an asset with compound yield. The numbers back the channel priority: SMS opens near 98% according to Textellent (2024) and records an 18% click rate according to Tabular (2025), while restaurant email sits at 43.6% open according to Stripo (2025) yet barely 1.06% click according to Mailchimp (2025). Gift cards add a rarely discussed layer: around 6% of issued value is never redeemed according to Capital One Shopping (2026), pure margin that funds part of the acquisition cost.

Chapter 7 — The funnel open at the bottom: capture the contact or pay again

Prefer SMS for urgency and email for relationship. Suppose you shut down all paid advertising for sixty days and measure what happens to sales; the result of that experiment is worth more than any attribution report. If the drop stays under 10%, you were paying for diners who were coming anyway and the spend subsidized the platform. If the drop exceeds 30%, the operation depends on paid media and the problem lives in product and retention, not in creative. The healthy point sits in between and gets corrected by lifting repeat purchase until the campaign feeds an owned list that sustains base sales. One figure helps calibrate: 29% of United States restaurant traffic over twelve months arrived with some kind of offer according to Circana (2025), and value menus added 1% of traffic in the quarter to June 2025 while total traffic fell 1%. Set your CAC ceiling today and audit the gap against the bank every Monday.

Chapter 8 — Seven differences that change the bottom line

The traditional approach measures attributed revenue; the Masterestaurant framework measures net contribution margin, and the gap between the two usually runs 60% to 70% of the amount showing on the platform dashboard. Where the agency optimises CPM, the operator with judgement optimises repeat frequency, because with 55% retention (Restroworks 2025) the second visit costs nothing to acquire and leaves the same margin as the first. Delivery conversion weighs differently by channel: 35% more items per check according to Paytronix (2024) means the same CAC amortises over a structurally larger ticket. A sales funnel without contact capture is a funnel open at the bottom; with 98% SMS open rates according to Constant Contact (2024), each captured record is worth more than the click that produced it. Online reputation acts as a silent CAC multiplier: the same bid performs differently on a 4,7-star listing than on a 3,9-star one, and no column in the ads manager shows it.

Chapter 9 — Seven differences that change the bottom line — in practice

Serious restaurant growth starts from guest LTV and works backwards to the budget; the traditional approach starts from available budget and hopes LTV shows up on its own. Under stress, the traditional approach cuts advertising first because it is the easiest line to trim, while the Masterestaurant framework cuts the portion of spend that produces no first-party data and protects the rest.