+2.1 EBITDA points from an optimized Google Business Profile: closing a local traffic leak with the Restaurant Model Canvas

The starting point: 3.7 stars and 640 USD a month covering the hole

The trattoria was billing between 500 thousand and 1 million USD a year with 14 tables, 11 employees and a 28 USD average check, and still, Monday through Thursday the dining room looked like a place under renovation. The Google listing had gone two years without a new photo, it still carried the lockdown hours, it held 3.7 stars across 214 reviews and —this was the serious part— it had 61 unanswered reviews, fourteen of them with specific complaints about waiting times. Meanwhile the owner was paying 640 USD monthly on Instagram ads to bring in people who, when they searched the restaurant's name on Google, ran into that storefront. The dining room accounted for 72% of sales, meaning the channel that listing was choking while nobody kept score. Because almost everyone checks it before deciding where to eat: 92% of diners read reviews before choosing a restaurant, according to Restroworks (2024), and 71% read them specifically on Google, according to BrightLocal (2024).

Why does the listing weigh more than the ad spend?

A neglected profile is not a pending marketing chore, it is the filter most of the demand passes through before crossing the door.

I put the figure in front of the owner because that is the only way the size of it lands: every ad dollar bought a click that ended on a page with 3.7 stars and fourteen silent complaints. And reputation arithmetic is not sentimental either: raising one star lifts revenue between 5% and 9%, according to Harvard Business School (Michael Luca, Reviews, Reputation, and Revenue), which turned those seven tenths into measurable money. The restaurant that invests most in buying new traffic is usually the one that takes worst care of the free asset converting that traffic: you pay for the click and give away the conversion. That tension shows up in nearly every restaurant growth audit we run at Masterestaurant, and here the owner was asking for the opposite of what was needed, he wanted MORE advertising.

The paradox we solved backwards from what the owner asked for

We froze the ad budget for six weeks —640 USD monthly, 960 USD freed over the period— and put all of it into the listing. The decision was not cosmetic: with sector net margins running 3% to 9%, according to Statista, a restaurant in this band cannot fund two fronts at once, and picking the cheaper one to operate is arithmetic, not preference. Six weeks later conversion moved from 11% to 34%. Nobody touched a photo until we knew which dish delivered contribution margin and which one destroyed it, and that sequence is no fussiness about order: promoting the wrong dish on Google multiplies a loss instead of a gain. We ran the Masterestaurant menu engineering matrix over 90 days of sales, dish by dish, crossing popularity against contribution margin in dollars. The dish the owner wanted to push on the listing —his favorite, the one he showed off to visitors— carried a 38% food cost, eleven points above the menu's 27% average, and it left the promotion on day one.

Diagnosis starts at the register, not at the listing

In its place we raised four dishes with food costs between 24% and 29%, and those were the ones that took over the product photos and the listing posts. Answering for the sake of answering is cosmetics; we crossed every complaint with its shift, its server and its ticket. The fourteen waiting-time complaints clustered in two Thursday shifts with a single server on the floor, an operations fact disguised as a reputation problem, and the real answer was not a polite comment but a different staffing grid with a second server on Thursdays from 8:00 to 11:00 PM. Only then did we reply to the 61 pending reviews across eleven days, with answers of 40 to 70 words and no template. Some 43% of diners consider it very important that a brand responds to comments, according to Toast (2024, via Tablein), yet the signal that moves the register arrives when the reply comes attached to the operational change that makes the next complaint unnecessary.

The team shoots the photos, on a phone, every week

A quarterly agency session at 900 USD produces dead images: twenty perfect dishes, lit identically, uploaded the same day and nothing new for ninety more. We put the head chef to work uploading six to eight phone photos a week —a plate on the pass, a table served at 9:00 PM, the bar on a Saturday— and reached 96 new images in three months against the agency's twenty, at zero cost. Visual appetite runs this business: 58% of diners visited a restaurant after seeing it on TikTok, up from 38% in 2022, according to MGH (2024), and the same mechanism operates on the Google listing, where last night's photo sells better than the perfect shot from a quarter ago. The listing climbed from 3.7 to 4.4 stars in six months. Each size has a different first move, and mixing them up costs weeks. Under 500 thousand USD a year: this week fix the hours, post eight photos from your phone and answer the ten most recent reviews; almost all the return sits there.

Transferable lessons by annual revenue band

From 500 thousand to 1 million —this trattoria's band—: audit food cost dish by dish BEFORE deciding what you promote, then cross every complaint with its shift. From 1 to 5 million: name a single owner for the listing, with two weekly hours blocked on the calendar and a conversion dashboard. Above 5 million, the media-chef archetype with three themed venues: separate each location's listing from the personal brand, because one venue's review contaminates the reputation of all of them. Above 10 million, group or chain: govern listings through the API, with monthly audits of hours, categories and NAP. I would not expect these numbers in three contexts, and saying so avoids selling smoke. First, a new restaurant with no history: there were 214 reviews and seven years of operation here, so climbing from 3.7 to 4.4 meant shifting an already loaded average rather than building it from scratch, and with twenty reviews the arithmetic is different.

Limits of this case

Second, a delivery-dominant business: the dining room carried 72% of sales, and where aggregators concentrate the bulk —an online delivery market projected at 473.49 billion USD in the United States for 2026, according to Statista— the listing matters far less than the ranking inside the app. Third, a kitchen with a genuine product problem: ratings rose because the food was fine and service failed on two shifts; if the food is bad, optimizing the listing only speeds up the bad news. Diagnosis starts at the register, not at the listing. Nobody touches a profile before knowing which dish carries contribution margin and which one destroys it; promoting the wrong plate on Google multiplies a loss rather than a gain. In this trattoria the dish the owner wanted to push ran a 38% food cost and came off the promotion on day one. A review is not answered, it is worked.

Three differences behind the 2.1 EBITDA points

Every wait-time complaint was cross-checked against shift, server and ticket, and fourteen of them clustered in two Thursday shifts staffed with a single server. The review stopped being online reputation and became an operations data point that rewrote the schedule. The team shoots the photography, not an agency. A quarterly 900 USD session produces dead images; the head chef's phone, with a framing and light protocol, produces fifty live images a month at zero added OpEx. Consistency beats production value in the local algorithm, and that turned out to be the cheapest lever in the whole project.