Territorial prefeasibility for new restaurants (MTIE): how we moved a pilot portfolio of 18 gastronomic MSMEs from 71.4% Prime Cost to 61.8% and stopped the default spiral

What did the 18.6% delinquency rate reveal that financial scoring missed?

The 18.6% ninety-day delinquency in this pilot portfolio was not a credit problem but a street-address problem, and that distinction redesigned the entire program.

Eighteen food-service MSMEs in a mid-sized Latin American city of 620,000 residents, 214 formal jobs sustained, weighted average ticket of 8.40 USD, average age of 14 months: the committee had approved using conventional scoring expecting 6% delinquency and got three times that. Eight locations held margin, ten sank, with identical financial products and comparable teams. Latin American food service carries a reputation for risk —in Colombia 95% of the market consists of independent establishments, per Acodrés via Revista La Barra (2024)—, yet that reputation hides the fine reading. Uniform risk does not exist. What exists is one district that absorbs demand and another that does not, separated by fourteen blocks. None of the 18 files contained a pedestrian count, an estimate of the district's food spending, or a supplier map within 40 kilometers: territorial prefeasibility existed as a signed form, never as a measurement.

The bank was financing addresses, not restaurants

Diego F. Parra put it to the committee in a sentence that stung —the bank was not financing restaurants, it was financing addresses—, and the sting came from being verifiable file by file. Worth stating what prefeasibility is NOT, since operations manuals hollowed the term out: it is not the 60-page market study filed to satisfy a requirement, nor the competitor list drawn inside an arbitrary radius. It measures absorption capacity, four variables and nothing else: real food demand in the district, spending elasticity, installed supply already serving it, logistics cost of supply. Anything the instrument produces beyond that is expensive noise. The MTIE —Territorial Economic Impact Model contributed by Masterestaurant S.A.S. as technology partner of the program run by SATE Institute with a second-tier bank credit line— works on four layers measured on the ground, not declared by the applicant.

How the MTIE was applied across the 18 diagnostics?

Pedestrian flow was counted by time band across seven days in each district; available food spending was estimated by crossing housing density with average income of the census tract;

installed supply was inventoried with its seat capacity; suppliers within 40 kilometers were mapped along with their delivery frequency. Out of that comes a territorial revenue ceiling, the number no file carried. Four locations in the portfolio had invested between 62,000 and 118,000 USD in build-out against a territorial ceiling of 340,000 USD per year. At that ratio, no later operating efficiency rescues the loan. CapEx already decided. Thirteen hours open in districts whose real traffic concentrated into four: that was the 38.2% labor cost the committee read as payroll indiscipline. Crossing the pedestrian count by time band against hourly ticket records settled it beyond argument, because staff presence was being paid against nonexistent traffic for nine hours a day.

The 38.2% labor cost was a schedule problem, not a payroll problem

Fitting schedules to the measured curve and reassigning freed staff to peak bands lowered that indicator without firing anyone, which matters when we speak of formal employment in a portfolio holding 214 positions. Food service sustains employment at scale —357 million people live off tourism worldwide, one in ten workers, per UN Tourism (2024), and in Mexico alone tourism generated 2.9 million jobs in 2024, up 3.5%, per INEGI (2024)—, but sustaining it demands that each open hour pay its own payroll. Nine point seven points separated theoretical food cost from actual cost across the portfolio average, and that gap was not born in the kitchen either: it was born in the supplier map. Locations sourcing from beyond 40 kilometers bought on long frequencies, overstocked and lost product, while those with a nearby supplier turned inventory twice a week with marginal waste.

The 9.7-point gap between theoretical and actual food cost

Food waste moves numbers no small owner sizes properly —in the United States food surplus reached 380 billion USD in 2024, of which 85% ended up wasted, per ReFED (2025)—, and inside an MSME with an 8.40 USD ticket every point of deviation eats the month's margin. Measuring distance to the supplier BEFORE disbursement fixes what afterwards can only be administered. Territorial prefeasibility does not stop a restaurant from opening in the wrong place: it stops credit from financing it, and that is the whole thesis of the pilot. Eleven of the 18 companies billed under 500,000 USD a year, five sat between 500,000 and 1 million, two cleared the million; dining-room service dominated with 27% aggregated delivery. Once the MTIE became a step prior to committee, files falling short of a sufficient territorial ceiling stopped being approved at the requested amount and moved to approval with resized CapEx or redirection to another district.

What changed once measurement came before disbursement?

Portfolio Prime Cost gave way and delinquency settled, per the program indicators. A genuine concession fits here:

for years I argued that the MSME food-service problem was operational, and this pilot forced me to accept that much of it gets decided before the stove is lit. Start by measuring your own district this week, with the instrument matching your size. Under 500,000 USD a year: count pedestrian flow in front of your door across three time bands for five days, by hand, and compare it against your hourly tickets; the schedule fix costs nothing and pays inside fifteen days. Between 500,000 and 1 million: map your suppliers by distance and frequency, then calculate what the kilometer costs you in waste. Above 1 million: demand a territorial revenue ceiling in the file before signing any build-out. Above 5 million: audit CapEx per location against the district ceiling, not against the group average.

Transferable lessons by annual revenue band

And above 10 million, the celebrity-chef archetype opening large format on brand reputation —where traffic is assumed guaranteed— is precisely who needs the count most, because fame fills three months and the district pays for the next twenty years. I would not expect these results in three contexts, and saying so matters more than displaying the pretty number. First, in cities above three million residents with dense mass transit: there the pedestrian flow of a single point does not describe its market, because demand travels and the district turns into an elastic concept that the MTIE, as applied here, does not capture well. Second, in formats with more than 60% delivery sales, where the physical address stops being the determinant and the delivery radius rules; aggregated delivery in this portfolio was 27%, half that threshold. Third, in seasonal tourist-destination operations, whose demand curves depend on season rather than residency; the sector added 27.4 million new jobs in a year, per WTTC (2024), and much of that answers to seasonality, not density.

Limits of this case

Measure before assuming your case resembles this one. The symptom read as 18.6% delinquency; the root cause was CapEx sized for revenue the polygon would never produce. Four locations in the portfolio had sunk between 62 thousand and 118 thousand USD into fit-out against a territorial revenue ceiling of 340 thousand USD a year. At that ratio no downstream operational efficiency rescues the loan. The second symptom, Labor Cost at 38.2%, was never a payroll problem either. It was a schedule problem: locations opened 13 hours a day in polygons whose real flow concentrated into four. Staff presence was being paid against traffic that did not exist, and what exposed it was the cross between pedestrian counts by time band and ticket registration by hour. That 9.7-point gap between theoretical and actual food cost came from logistics, not from the kitchen. Twelve operations bought from more than 180 kilometres away, in small orders, with freight prorated at the supplier's discretion.

Where the leak actually was?

Rebuilding provisioning inside the 40-kilometre radius erased 6.2 of those 9.7 points without touching a single recipe. A third finding matters more to a lender than the other two:

territory predicts better than history. In this portfolio the territorial score sorted delinquency cleanly —the six locations scoring low held 74% of overdue balance— while conventional financial scoring had filed them as medium risk. Restaurant credit risk here has a geographic origin the traditional file never captures. The tension of the trade is real, and it has to be resolved rather than dodged: a founder picks a site out of affection, proximity to home, or a lease that happened to appear, and that same impulse sustains the 95% of the Colombian market held by independents (Acodrés, 2024). Killing it with technical bureaucracy kills formalization too. The bridge is a cheap, fast instrument: priced below 1.5% of CapEx and delivered in ten days, founders use it; priced at 6% and delivered in two months, they fake it.