+3.1 EBITDA Points While Scaling a Viral Restaurant: Closing the Location 2 Leak with MTIE Pre-Feasibility and the Restaurant Model Canvas

The diagnosis: 92,000 followers, 19.4% EBITDA and a second location returning nothing

Scaling a restaurant is not repeating the location that works, it is repeating the unit economics that works, and those are two different things. The group came in with two locations, 41 employees and 22 tables in the flagship against 18 in location two, open for seven months on 214,000 USD of CapEx covering construction, equipment and working capital. The flagship ran at 19.4% EBITDA with a 24.60 USD average ticket; location two billed 62% of that with the same social push and a 19.10 USD ticket. Instagram brought in 71% of group reservations and WhatsApp from the bio another 14%, with a Reel at 1.4 million plays as proof the channel worked. It worked to fill tables. It did not work to decide where to put the second kitchen. Location two sold less because its audience was the SAME person, not a new person, and the reservation system's phone cross-check exposed it: 34% of location two's guests that quarter had eaten at the flagship within the previous six months.

Why did location two sell 62% of the flagship on identical traffic?

That is not expansion, it is relocating the same cash between two leases, except now there are two administrations, two insurance policies and two payrolls.

Eight blocks between doors reads like area coverage on a map; read from the reservation database, those eight blocks sit inside a single catchment radius. The group had not opened a market, it had split its own in two, and the Instagram account that looked like the growth engine was in fact shuttling diners from one dining room to the other. Second finding, and the one that eats the profit: location two's Prime Cost sat at 71.8% against the flagship's 61.4%, with two linked causes. The recipe costings were inherited without recalculation —same menu, different suppliers, different waste, a ticket 5.50 USD lower— and Labor Cost held at 38.4% through defensive overstaffing: the owner had added extra people «just in case» in a location he did not know.

Prime Cost at 71.8%: inherited recipe costings and defensive overstaffing

That cost stays invisible in the P&L until somebody names it. And it is not a country problem: according to ABRASEL (2025), annual food service payroll in Brazil exceeds 107 billion reais, which gives you the scale of what every percentage point of labor is worth. Ten points of Prime Cost on this revenue were, in this case, the difference between funding the third location and killing it before opening. The engagement opened with the Unit Economics Matrix by Location, the tool Diego F. Parra uses at Masterestaurant to separate what generates cash from what only generates motion. We built it per location and per daypart: contribution margin per cover, real food cost recalculated dish by dish, labor cost per shift and daily break-even expressed in covers, not in money. Location two needed 104 daily covers to break even and was doing 78. With that matrix on the table, the conversation moved from «how do I sell more» to «how many covers am I short and at what margin», which is a question with an operational answer.

The tool: the Masterestaurant Unit Economics Matrix by Location

We recalculated 34 recipe costings, rebuilt the shift grid against measured demand and froze construction on the third location for eleven weeks. Freezing that build was the most profitable decision of the year, and the hardest to argue. Six months in, location two's Prime Cost dropped from 71.8% to 63.9%, with labor at 31.2% and EBITDA at 11.7% where there had been an operating loss. Average ticket climbed from 19.10 to 22.40 USD through menu reengineering, with no list-price increase: the higher contribution margin dishes moved to the reading zones guests scan first. Content continued, under a different brief: four weekly Reels became two, and the target stopped being reach and became acquisition outside the flagship's radius, measured by the share of new phone numbers in reservations. That indicator went from 66% to 81% of guests with no history at the flagship.

The six-month result: Prime Cost from 71.8% to 63.9%

The third location returned to construction in month nine, this time with its own cost model and 3.2 kilometers away, not eight blocks. Here is the tension that gets me into the most arguments with owners of socially strong brands: the content that performs best is the content that most easily hides a unit economics problem. A full dining room covers a negative margin for months, because cash comes in every day and the loss only surfaces when somebody closes the month. What would have happened if the group had opened the third location in month eight, as planned? With 71.8% Prime Cost replicated and another 214,000 USD of CapEx, they would have entered month twelve with three operations, two below break-even and one financing them all: the classic script where the profitable location dies of somebody else's success. The brand would have kept growing in followers while the group decapitalized.

The paradox of content that works too well

And that 1.4 million play Reel would stand as the best evidence that the attribution was wrong. The lesson applies differently depending on what you bill per year, and each band has one concrete first step this week. Under 500,000 USD: calculate break-even in daily COVERS, not in money, and tape it to the kitchen door. From 500,000 to 1 million —this case's band—: cross-check the phone numbers from your last six months of reservations and measure what share of the new location already ate at the old one. Over 1 million: recalculate every location's recipe costings with its real suppliers, never inherited. Over 5 million: audit Labor Cost per shift and per location; defensive overstaffing multiplies with distance from the owner. Over 10 million, group or chain: a frequent archetype is the media chef whose personal brand opens sites the unit economics cannot hold, and there the first step is arming the new-openings committee with veto power over any location that has not closed its cost model BEFORE the lease is signed.

Transferable lessons by annual revenue band

Per FRANdata, roughly 43,212 multi-unit operators control more than 223,213 units in the United States, 54% of the total, and that concentration exists because they replicate a cost model, not a pretty room. I would not expect this result in three contexts, and it is worth saying so before somebody copies the plan. First, operations where the dining room does not rule: when 75% of traffic happens off premise, as Nation's Restaurant News reports for much of the sector, the bottleneck is not the menu or the floor staffing but delivery cost and platform commission, and the cover matrix loses explanatory power. Second, high-rotation drive-thru formats, which per Circana move 43% of fast food orders in the United States, close to 140 billion USD a year: there unit economics gets decided in window seconds, not in contribution margin per dish. Third, highly atomized markets —in Mexico 96% of sector businesses are micro-enterprises, according to CANIRAC (2024)—, where the scaling problem is usually capital access and formalization long before inherited costings.

Limits of this case

The method holds; the order of the levers changes. Symptom: location 2 booked 62% of flagship revenue on identical social traffic. ROOT CAUSE: the audience was the same person, not a new person. The reservation-system phone cross-match gave it away — 34% of location 2 guests that quarter had eaten at the flagship within six months. That is not expansion, it is relocating the same till across two leases. Symptom: location 2 ran a 71.8% Prime Cost while the flagship sat at 61.4%. TWO ROOT CAUSES: inherited recipe costs nobody repriced, and a 38.4% Labor Cost propped up by defensive overstaffing — the owner added bodies «just in case» in a room he did not know. Brazilian food service payroll alone exceeds 107 billion reais a year (ABRASEL, 2025), and that structural payroll weight is exactly what punishes an operator who opens a second unit without sizing shifts from data.

Where the capital was actually leaking?

Symptom: the P&L showed profit, the bank account showed anxiety. ROOT CAUSE: a deferred P&L hiding cash flow. The 214,000 USD CapEx amortized over five years on paper while the loan behind it amortized over 36 months;

the gap between the statement and the checking account ran 3,900 USD a month of real OpEx invisible in the income statement. Symptom: four Reels a week, rising reach, flat reservations. ROOT CAUSE: a vanity metric replaced a commercial one. Nobody tracked acquisition cost per new reservation, sitting at 4.80 USD, or the save-to-reservation conversion rate. With off-premise running near 75% of sector traffic per Nation's Restaurant News, the team produced content for a dining-room channel while ignoring the channel growing fastest. Symptom: 128% annual turnover at location 2 against 61% at the flagship. ROOT CAUSE: Skills Gap. The talent holding the flagship together was six years of tacit knowledge, impossible to transfer through a three-day onboarding.

Where the capital was actually leaking — in practice?

With no manual, every resignation erased a recipe. The diagnosis closes on a tension the owner did not want to hear: the brand was BETTER than the company.

It had a name, an audience, a visual identity — and no replicable unit economics. In my judgment that is the most dangerous setup for scaling a restaurant, because traction funds the mistake just long enough for the mistake to become structural.