EBITDA down from 9.1% to 1.8% for scaling on views instead of unit economics: how we turned a four-location group around with the MTIE feasibility model and the Restaurant Model Canvas
Verdict: scaling a restaurant is not decided by views, it is decided by unit economics per location. Here, a Peruvian street-food brand with 11.4 million accumulated views across Reels and TikTok opened three locations in fourteen months and watched consolidated EBITDA slide from 9.1% to 1.8%, because the content pulled traffic into a single catchment area while the new sites inherited the CapEx and none of the demand. The fix took seven months: MTIE feasibility before signing any lease, a replicable operations manual to close the theoretical-versus-actual cost gap, and a content structure built on one brand account plus geolocated local accounts. Consolidated EBITDA at month 7: 8.4%. The myth says the audience travels with you. The reality is that an audience stays where it lives.
Interpretation can wait; here is the case file. A Peruvian street-food group runs 4 locations (one original, three opened between 2024 and 2025), 96 seats combined, 71 employees across front of house, kitchen and a two-person in-house content cell, in a mid-sized market of 1.9 million metro residents, with an average check of 11.80 USD in the dining room and 14.20 USD on delivery, six years of brand history, and dine-in as the dominant channel at 62% of sales while the rest splits between owned delivery and aggregators. Consolidated revenue at the close of the prior fiscal year reached 1.9 million USD, which places the group squarely in the above-one-million band, the stage where a family business stops being one and still has no corporate structure.
No sales drop brought this operator to the audit. Revenue across the four sites had never been higher, and yet group cash had turned into a monthly standoff between payroll and suppliers. Peak billing with bottom cash: that symptom shows up whenever someone confuses growth with scaling, and here the origin was very specific, because the decision to open the second, third and fourth location came from reading an Instagram dashboard instead of a per-unit P&L. The owner put it plainly in session one: we had 11.4 million views, how were we not going to fill another room?
Calibrating the ambition takes a look at the sector first. Franchising generated 578 billion USD of U.S. GDP in 2025, with 5% growth against 1.9% for the broader economy according to the International Franchise Association (2025), and multi-unit operators already control 54% of franchised units, roughly 43,212 operators across more than 223,213 locations per FRANdata. Multi-unit, then, is a trade in itself, with its own learning curve. This group wanted into that league carrying a genuine marketing advantage and a genuine structural weakness.
Twenty minutes in, one figure reframed the conversation: 75% of sector traffic now happens off-premise according to Nation's Restaurant News, while this group booked 62% of its sales inside the four walls. The problem was never the digital channel, it was geography. The three openings sat 4.2, 6.1 and 9.8 kilometres from the original room, inside the same city, while the content audience, the same one cheering every Reel, clustered within a 3-kilometre radius of the first address. The views were real. The physical movement of those views was not.
Side-by-side: scaling a restaurant
| BEFORE (baseline, month 0) | AFTER (month 7) | |
|---|---|---|
| Consolidated group EBITDA | ✕1.8% of sales (down from 9.1% as a single location) | ✓8.4% of sales, with all 4 locations running |
| Consolidated Prime Cost (food + labor) | ✕71.3% of net sales | ✓62.9% of net sales |
| Theoretical vs actual food cost gap | ✕6.4 points (theoretical 29.1%, actual 35.5%) | ✓1.3 points (theoretical 29.4%, actual 30.7%) |
| Labor Cost % at the weakest site (location 3) | ✕41.2% of its sales | ✓31.6% of its sales |
| Average dining-room check | ✕11.80 USD | ✓14.60 USD |
| Annualized front-of-house turnover | ✕128% per year | ✓67% per year |
| Sales attributable to geolocated content per site | ✕0% measurable (single account, no per-site attribution) | ✓23.7% of sales at locations 2, 3 and 4 |
| Days of free cash at group level | ✕6 days | ✓38 days |
The case file: 11.4 million views, 4 locations and a cash position at rock bottom
Eleven point four million accumulated views across Reels and TikTok did not pay a single payroll run at this Peruvian street-food group, and the audit starts from that number. Here is the file, plainly: 4 locations running (the original one, six years old, plus three opened between 2024 and 2025), 96 seats combined, 71 employees spread across floor, kitchen and a 2-person in-house content cell, a mid-sized city with a metro area of 1.9 million people, an average check of 11.80 USD in the dining room and 14.20 USD on delivery, and the dining room carrying 62% of sales. The prior fiscal year closed at 1.9 million USD consolidated. Above one million is exactly where a group stops being a family business without yet owning a corporate structure, and that limbo explains half the case. No sales drop brought the owner in. He was billing more than ever; cash had turned into a monthly arm-wrestle between payroll and suppliers.
Why did the views never turn into covers?
The views never travelled because the audience lived within 3 kilometres of the first location, while the three new openings sat 4.2, 6.1 and 9.8 kilometres away from it.
Geography, not digital channel, and it is worth saying that bluntly before going further. Nation's Restaurant News reports that 75% of sector traffic already happens off-premise; here 62% of sales still walked into four dining rooms, on a skeletal owned delivery operation and badly negotiated aggregator terms. We went through the view counts one by one: real, verifiable, bot-free. What did not exist was the physical movement of those people. A Reel with 380,000 plays produced a 40-minute queue on Saturday at the original site and moved nothing at location 3, which closed that same Saturday at 31% occupancy. Replicating a venue is not scaling. Scaling replicates the DEMAND that sustains the venue, and that rarely fits inside a reach dashboard.
What surfaced when we opened the ledger by cost centre?
Two locations were funding the other two.
Nobody in the group knew it, because the P&L consolidated quarterly and by group, never monthly and by unit, so location 3 had been draining 4,100 USD a month for eleven months (roughly 45,100 USD accumulated) hidden inside a result that still looked positive thanks to the original site. One afternoon of splitting the ledger by cost centre was enough to see unit contribution: 18.3% operating margin at the original, 9.7% at the second, −7.4% at the third, 2.1% at the fourth. Group food cost ran at 35.5% against a theoretical 29.1%. That 35.5% breaks the house ceiling: above 32% there is no room to manoeuvre, there is a broken roof. And labor cost at location 3 climbed to 41.2%, with a staffing plan sized for occupancy that never arrived. Before that week, not one of those four figures appeared in any report.
The action: standardised recipes and a monthly P&L per unit with the Masterestaurant method
Closing location 3 was the obvious move, and it was not the first one: the first was writing 34 standardised recipes with gram weights per component, because every kitchen plated from memory and protein portions varied by up to 22 grams between sites on the same dish. On top of the Masterestaurant costing tool we built the dish-by-dish cost sheet with a food cost target per family. According to Diego F. Parra, restaurant consultant at Masterestaurant, nobody opens a new site until the previous one posts three consecutive months of positive contribution measured in its own P&L, and that uncomfortable rule set the whole calendar. A monthly close by cost centre came in alongside it, reviewed on the 8th, with a single traffic-light metric per unit. By month four: group food cost at 30.2%. That 5.3-point correction, on 1.9 million in revenue, amounts to roughly 100,700 USD a year no longer evaporating in the kitchen. Location 3 moved from −7.4% to 1.9% contribution by month seven, without closing it.
The measurable result at twelve months, and what it cost to get there
Twelve months later, without a single extra location, consolidated EBITDA rose from 4.8% to 11.6%, which is precisely the part the owner struggled to accept. Group payroll dropped from 33.4% to 28.9% of sales once we rebuilt shifts against each site's real traffic curve rather than the original's; at location 3 alone that freed 2,840 USD a month. Owned delivery went from 11% to 19% of sales with the 14.20 USD ticket holding, while aggregator dependency fell from 27% to 19%. The content cell kept producing, under a different mandate: it stopped chasing reach and began geotargeting campaigns within a 2-kilometre radius of each site, with a different offer per location. Views fell to 6.9 million for the year. Sales per seat rose 23%. Less reach, more cash: that trade sums up the whole case better than any chart.
The multi-unit operator's paradox: growing and scaling are not the same trade
Growing is opening; scaling is opening without the system losing margin per unit, and confusing the two burns more cash than anything else in the million-dollar band. Sector data backs it up: franchises generated 578 billion USD of US GDP during 2025, with 5% growth against the broader economy's 1.9% according to the International Franchise Association (2025), and multi-unit operators now control 54% of franchised units, some 43,212 operators across more than 223,213 locations per FRANdata. Behind that 54% there is neither coincidence nor luck, there is a learning curve paid for in venues, and whoever tries to skip it with marketing pays it anyway, only in EBITDA. An honest counterfactual belongs here. Suppose the group opens location 5 in month fourteen as budgeted, with food cost still at 35.5% and no monthly close per unit: the combined drain from locations 3 and 5 would have passed 9,000 USD a month and cash would have slipped into arrears with suppliers by month eighteen. That is not an elegant hypothesis. It is arithmetic.
Transferable lessons by annual revenue band
Each band has a different first step, and giving advice by size adjective is what stops it from landing. Under 500 thousand USD: this week, write down with gram weights the 10 recipes that make up 70% of your sales, and measure real food cost against theoretical; without that there is nothing to scale. Between 500 thousand and 1 million the step is accounting, so split the ledger by cost centre even with two locations and fix the 8th as a non-negotiable monthly close. Above 1 million, this case's band, ban the next opening until the last site delivers three consecutive months of positive contribution. Past 5 million you meet the celebrity-chef archetype or the large-format themed concept, where personal brand sells the first venue and the next three hold up only through menu engineering and waste control; audit there what percentage of margin depends on the founder being physically present. Group or chain above 10 million: real-time unit contribution and a committee able to veto openings with data, not enthusiasm.
Limits of this case
I would not expect these numbers in three contexts, and saying so protects more than any promise. The first is a single-location operation: the biggest lever here was isolating unit contribution and reallocating shifts across sites, a lever that with one site does not exist, so almost all the upside sits in food cost. Next comes the group whose dominant channel is delivery at 70% or 80% of sales, where the geography diagnosis inverts: delivery radius and aggregator commission outrank dining-room occupancy, and the 75% off-premise traffic reported by Nation's Restaurant News stops being a contrast and becomes the norm. Then there are markets where the sector's structural weight changes the labour arithmetic (Mexico, with 12.2% of the country's businesses and 96% microenterprises according to CANIRAC 2024, or Colombia, with 3.9% of GDP and 8% of the workforce according to ACODRES 2024), and there the payroll thresholds that worked here may land too high or too low. The method travels. The numbers do not always.
Root-cause diagnosis: which symptom exposed which failure
SYMPTOM: record billing with strangled treasury. ROOT CAUSE: the P&L was consolidated quarterly and by group, not monthly and by unit, so location 3 had been draining 4,100 USD a month for eleven months without anyone isolating it. Opening the ledger by cost centre produced the tell: two locations were financing the other two. SYMPTOM: actual food cost of 35.5% against a 29.1% theoretical. ROOT CAUSE: zero written standard recipes. Every kitchen plated from memory and from the shift lead's judgement, with protein portions varying by as much as 22 grams between sites on the same dish. That 35.5% also breaks the house ceiling, since food cost above 32% is a broken roof rather than room to manoeuvre. SYMPTOM: Labor Cost of 41.2% at location 3. ROOT CAUSE: headcount got sized against PROJECTED sales derived from views, never against real sales in the first eight weeks. When traffic came in 38% under, nobody touched the schedule: the manager feared giving bad service during the opening.
Root-cause diagnosis: which symptom exposed which failure — in practice
SYMPTOM: 128% annual turnover in the dining room. ROOT CAUSE: a structural Skills Gap. Three locations opened on three days of informal training, with no service script and no promotion ladder, and new hires burned out inside the first quarter. Hiring and rehiring cost 1,180 USD per position across recruiting, uniform and the unproductive ramp. SYMPTOM: viral content that converted nowhere near the new sites. ROOT CAUSE: one mother account was publishing to an audience inherited from the original neighbourhood, with no geolocation, no local accounts, and not one piece telling anyone where location 4 actually was. The Reels sold the brand; no Reel sold an address. SYMPTOM: lease decisions closed in under a week. ROOT CAUSE: no location due diligence at all. None of the three openings passed a location intelligence study measuring density, foot traffic by daypart, competition within 800 metres and the catchment's purchasing power; two of those contracts were signed on price opportunity alone.
Myth against reality, criterion by criterion
The myth: your audience follows you to the new room
- Eleven million accumulated views mean guaranteed demand anywhere in the same city.
- If the first location bills well, copying the menu and the façade copies the result.
- Expansion CapEx pays itself back on opening-week media buzz, no financial model required.
- A single Instagram and TikTok account feeds all four points of sale equally.
- The recipe lives in the founding chef's head and that is enough for four kitchens.
- A new location every four months signals success, not treasury risk.
The reality as measured in the group P&L
- The audience clustered within 3 kilometres; locations 3 and 4 opened with 38% less traffic than projected.
- With no replicable operations manual, the theoretical-versus-actual gap hit 6.4 points and ate 78,000 USD a year.
- CapEx across the three openings totalled 612,000 USD, and the 22-month projected payback stretched to 51 real months.
- Under a single account, per-site attribution was zero: nobody knew which Reel filled which room.
- Four kitchens on oral recipes produced four different plates and a food cost that peaked at 35.5%.
- Free cash fell to 6 days, which is one week of cushion against any surprise.
The numbers we moved in seven months
“I was proud of those 11.4 million views and I signed three leases on that pride. Nobody told me my audience lived inside a three-kilometre radius. The month I found out location 3 had been losing 4,100 dollars a month for eleven months while I celebrated a Reel with 600,000 plays, I understood that content had handed me demand, not a business model. Today I do not sign a lease without the MTIE and without seeing the previous location's P&L stable for two straight quarters.”
Composite case for illustration: the names and figures in it do not describe a real business and are not industry data.
Chronological treatment: seven months, phase by phase
The first move was to stop reading the group as one business. The ledger got split into four cost centres, and every location received its own Restaurant Model Canvas, with its own value proposition, its real customer and its cost structure. The finding landed on day nine: locations 1 and 2 contributed 11.9% and 7.4%, while location 3 sat at -3.1% and location 4 barely grazed zero. That exercise alone reframed the whole engagement, because the owner had spent eleven months reading a consolidated figure that averaged his best room with his worst and handed back a lukewarm number describing nothing.
We ran the MTIE feasibility model across the three existing sites as if the decisions were still open. Location 3 would have failed: pedestrian density 41% below threshold, four direct competitors within 800 metres, and rent at 14.2% of projected sales against the 8% ceiling the model tolerates. We froze any new opening for twelve months, and that produced the first real friction: the owner had a fifth lease verbally committed, and losing that space stung for weeks. We held the freeze anyway. That fifth room would have consumed 190,000 USD of CapEx the group did not have.
All 34 menu recipes were written with gram weights, yield, waste and unit cost, then posted physically on every production line. It did not work the first time. Two head chefs kept plating by eye for three weeks and the weekly count exposed it: location 2 still ran 4.8 points over theoretical. The fix came through a scale at every station, a blind weekly count, and a joint Friday review with all four head chefs in the same room. From week eleven onward the gap genuinely started to close.
This is where the group's own specialty finally worked for it. We restructured the audiovisual operation: the mother account keeps brand narrative and viral volume, and each location got a geolocated account carrying neighbourhood content (the regular who has breakfast every Tuesday, the corner market, the local team's match, the Saturday queue), with a visible address on every piece and a distinct table code per site for sales attribution. Three Reels per week per local account, produced by the same two-person cell from a shot bank filmed in half-day sessions. By month 7, 23.7% of sales at locations 2, 3 and 4 traced back to that content.
Using real sales by daypart over the previous ninety days, we cut 148 weekly labour hours across locations 3 and 4 without a single dismissal: shifts moved toward high-traffic windows, and Monday lunch closed at location 3, which had been billing below its cost of opening the doors for months. On that base we installed meseros.ai for training and service scripting, with weekly assessment. Labor Cost at location 3 dropped from 41.2% to 34.8% in six weeks and kept easing to 31.6%.
Everything above consolidated into a 78-page replicable operations manual covering opening, closing, production, counting, service script, content protocol and an incident escalation matrix. In parallel we switched on the Demand Radar to read which categories each catchment wanted, then reordered every menu on that reading: location 4, sitting in an office district, pushed a midday executive menu, and the group's average check climbed from 11.80 to 14.60 USD. The PHYSICAL menu stayed in all four dining rooms, with QR as a complement for delivery, accessibility, pricing and analytics. On that we did not budge: the physical menu is where service pace and suggestive selling get controlled, and removing it would have cost more than it saved on printing.
The close was not a report, it was a rule. We wrote the group's opening criteria on a single page: no new location gets signed without an approved MTIE, without the previous site holding contribution above 10% for two quarters, and without 45 days of free cash at group level. We also built a weekly eight-indicator scorecard (contribution per site, theoretical-versus-actual gap, Labor Cost, check, turnover, free cash, content attribution and reviews) that the owner goes through every Monday in forty minutes. Consolidated EBITDA closed month 7 at 8.4% and free cash at 38 days.
And with AI?
Standardize and replicate processes to scale and franchise with control. Diego F. Parra is an expert in AI applied to restaurants.
Free tools: scaling a restaurant
The tools that carried the case
None of these pieces was custom-built for this group. They are closed, off-the-shelf products from the Masterestaurant suite that deploy in days rather than months, and that is why the group stabilized in seven months instead of two years: consulting time went into deciding and into fixing friction, not into manufacturing tools from scratch.
Diego F. Parra insists on an order this case confirms: model first, cost second, demand third. Invert that sequence, chase demand for a model that loses money per plate, and the leak multiplies; scaling a restaurant with the order reversed is precisely what turned 11.4 million views into six days of cash.
Diego F. Parra — International consultant, expert in creating and scaling restaurants and in AI applied to restaurants, foodtech and HORECA. Methodology applied in 8.400+ restaurants across 43 countries · Expert in Artificial Intelligence applied to restaurants, hospitality and food businesses · 20+ years in restaurants, catering, large events and business growth · Author of 3 ISBN-registered books: «Triunfar o morir en el intento» (2013) and «De esclavo a dueño» (2023) · International keynote speaker for the HORECA sector.
Questions that always come up in this conversation
How many views does a restaurant need before opening a second location?
How many views does a restaurant need before opening a second location?
No view count justifies an opening. What justifies it is contribution above 10% at the current site for two consecutive quarters, 45 days of free cash, and an approved location intelligence study. Here there were 11.4 million views and six days of cash: the content delivered demand concentrated within three kilometres, not expansion capacity.
Does Reels and TikTok virality help with scaling a restaurant?
Does Reels and TikTok virality help with scaling a restaurant?
It helps as an accelerator for a model that already works, never as validation for one that does not exist yet. Virality fills the room already near its audience; a site nine kilometres away needs its own geolocated account, neighbourhood content and measurable attribution. That single change explained 23.7% of sales at the new locations.
What is the MTIE feasibility model and why does it come before the lease?
What is the MTIE feasibility model and why does it come before the lease?
It is the technical investment and study model that validates a site before any CapEx is committed: density, foot traffic by daypart, competition within 800 metres, catchment purchasing power, and rent as a share of projected sales with an 8% ceiling. Run retroactively here, location 3 would have failed, sparing 4,100 USD a month of leakage over eleven months.
Can you scale without a replicable operations manual if the founding chef supervises everything?
Can you scale without a replicable operations manual if the founding chef supervises everything?
Not past the second location, and this case proves it numerically: four kitchens on oral recipes produced a 6.4-point gap between theoretical and actual cost, with food cost at 35.5% when the tolerable maximum is 32%. The 78-page manual closed that gap to 1.3 points. Personal supervision does not scale; the document does.
Scaling a restaurant by the numbers (2026)
Verifiable industry benchmarks from official, non-commercial sources (government, industry associations, market research) - not competitors.
| Metric | Value | Source |
|---|---|---|
| Jersey Mike's size and growth in fiscal 2025: about 3,300 stores, 250+ net openings, systemwide sales over $4 billion | about 3,300 stores, more than 250 net openings, system sales above 4,000 million USD | Restaurant Dive — Jersey Mike's IPO 2025 |
| Popeyes pace and goal in North America: about 200 restaurants a year, target of 800 new locations | about 200 restaurants a year, target of 800 new venues | QSR Magazine — Popeyes 800 New Locations 2025 |
| Net franchise unit growth 2025 | +20.000 unidades (a 851.000 en EE. UU.) | IFA Economic Outlook 2025 |
| New franchise jobs 2025 | +210.000 puestos (+2.4%) | IFA Economic Outlook 2025 |
| Franchise sector output 2025 | USD 936.4 mil millones (+4.4%) | IFA Economic Outlook 2025 |
| Franchise GDP 2025 vs US GDP | USD 578 billion (+5%, vs +1.9% for U.S. GDP) | IFA Economic Outlook 2025 / CBO |
Related content
Scaling a restaurant: bring this case to your restaurant
Applied in +8.400 restaurants across 43 countries.