How to improve social impact measurement in food service: from job counts to traceable operating data

Chapter 1 — Why did the closing survey stop working as proof of impact?

Because it measures declared intent rather than recorded transactions, and multilateral banks no longer accept it as portfolio evidence.

When Springer Nature (2025) documents that food service accounts for 18% of the food-related carbon footprint and that a commercial kitchen emits two to five times more than an equivalent space in another use, the order of magnitude demands a different class of data. A form filled in on the last day of a program captures what the beneficiary remembers or wants to please with; the electronic payroll of month 7 has no such motive. The ILO counts roughly 140 million informal workers in Latin America and the Caribbean, about half of regional employment, and no satisfaction survey tells apart a small business that formalized eight positions from one that churned the same vacancy eight times. That distinction is the ENTIRE finding. Impact gets read from the point of sale, payroll, purchasing and waste weighing, on a monthly cut with traceability back to the source document.

Chapter 2 — Moving the data source into the transactional system

This is not a methodological preference: it is the only way a third party can reconstruct the figure three years later. An indicator is worth what its reconstruction is worth, and if waste kilos are not weighed by production line, the reported number is an estimate wearing the suit of a fact. UNFCCC and FAO (2024) place food loss and waste at 8-10% of global greenhouse gas emissions, at a cost near one trillion dollars a year; that size is not audited with a spreadsheet filled in from memory. SATE Institute defines the monitoring and evaluation framework, while Masterestaurant S.A.S. supplies capture at the source. Diego F. Parra reads what the system records in terms of contribution margin and prime cost. A monthly series separates formalization from churn; a single measurement confuses them forever. With twelve cuts a year, the program officer sees the curve: if headcount rose from 6 to 14 people in March and fell back to 7 by July, no jobs were created, high season happened.

Chapter 3 — Frequency: twelve points a year, not a snapshot

The declarative framework would have reported fourteen. And the cost of that confusion is financial rather than reputational, because the external evaluator arrives at month 36 with access to social security filings and compares. The ILO calculates that roughly 6 out of 10 employed young people in the region work informally, and restaurants are their entry-level employer; tracking their permanence month by month is what makes a youth employment program verifiable. My recommendation is blunt: without a monthly series, do not release the second tranche. Counting declared positions inflates results systematically because it never subtracts departures. In operations below 500 thousand USD a year, net formal employment sustained at twelve months usually lands between 55% and 70% of the reported gross, and that gap —not the big number— is the finding that lets you redesign the retention component. I got this wrong for years: I accepted gross figures because gross is what looks good in the donor report.

Chapter 4 — The right unit of measure is NET formal employment

ECLAC (2024) puts labor informality among the region's small businesses at 46.6%, concentrated in micro and small firms, so the real starting point is almost always worse than the declared one. Define the indicator as hires minus exits, with contract and affiliation active at month twelve, and accept that the number will drop. It will drop because it will finally be true. The same indicator changes meaning with the size of the till, and that is where most frameworks break. Below 500 thousand USD a year the viable data point is social security affiliation, captured from payroll; line-level weighing does not exist yet. Between 500 thousand and 1 million, weekly inventory counts appear and waste can be measured against purchases. Above 1 million, the point of sale allows waste attribution per dish, with reference food cost between 28% and 35% according to the National Restaurant Association. Above 5 million there is internal control and the indicator becomes auditable without outside help.

Chapter 5 — Each revenue band measures something different

With sector net margins of 3% to 9% per Statista, demanding the same dashboard from a 400 thousand USD operator that you demand from a 6 million one amounts to funding consulting with impact money. In the celebrity-chef restaurant or the large-format themed venue, measurement no longer costs to implement but to defend. These operations —5 to 10 million USD a year and beyond— run an ERP, hold internal control and almost always carry a public sustainability commitment that exposes them. Their own cost is cross-verification: station-level weighing in kitchens of 40 people, supplier traceability across menus of 90 references, and auditing the figure before the communications team publishes it. The EPA (2023) estimated that 61% of methane from food buried in US landfills escapes uncaptured into the atmosphere, a data point no group with public reporting can ignore. The tension here is real: large formats own the systems yet face the highest reputational risk, while small businesses carry low risk and no systems.

Chapter 6 — High end: above 5 million, the cost of measuring changes nature

It resolves counterintuitively: audit the large one, instrument the small one. It unfolds as a chain, and the ending is predictable. First the mid-term report claims 300 jobs created; then the external evaluator cross-checks social security records at month 36 and finds 170 active; next the funder freezes the following tranche while asking for explanations; and finally the whole program loses its credit line because it cannot prove additionality. IFC and the World Bank (2024) estimate that 70% of small businesses in emerging markets lack adequate financing to grow, so every line closed over poor measurement gets deducted from that 70%. Operators who did formalize pay the bill. That is why transactional data is portfolio protection rather than added bureaucracy, and why it belongs in place before the first disbursement and not once the evaluator has already booked a meeting. Four captures, none heroic, all tied to a document that exists with or without the program.

Chapter 7 — What gets installed in month one

Electronic payroll with hire and exit dates per person, so net figures can be calculated. Purchases with invoice and unit of measure, so the kilo bought has something to be checked against. Waste weighing by line at end of service, even if it starts at a single station. And sales per dish from the point of sale, which is what turns kilos into margin. SDG target 12.3, which the IDB drives in Mexico, Colombia and Argentina through the #SinDesperdicio initiative, calls for halving per capita food waste by 2030; without those four captures nobody will know whether anyone got closer. Start with payroll, which draws the most resistance and reveals fastest whether the social impact the program talks about exists in the accounting. FREQUENCY. Declarative measurement produces a photograph; traceable measurement produces a series. With twelve points a year, a program officer can tell an MSME that formalised its workforce from one that churned the same vacancy eight times, and that distinction changes the next disbursement decision.

Chapter 8 — The four differences an investment committee actually weighs

UNIT OF MEASURE. Counting declared jobs inflates results systematically because exits are never deducted. Net formal employment sustained at twelve months typically lands between 55% and 70% of the gross figure in operations under USD 500,000 a year, and that gap is exactly the finding a program needs to redesign its retention component. TRACEABILITY. An indicator is worth whatever a third party can reconstruct. If waste kilos were never weighed by line, the avoided-FLW figure will not survive verification, and the whole report loses credibility with the multilateral evaluator even when everything else is sound. MARGINAL COST. Field consultancy runs USD 1,800 to 4,500 per round per operation; pulling the same data from systems the restaurant already runs costs almost nothing. That difference is what turns a 40-MSME sample into a full portfolio census.