Data vs intuition decision-making: the case study that raised net margin from 6% to 14% in 2026

The diagnosis that eight years of intuition could not see

El Fogón de Andrés reached January 2026 with a food cost of 38% and a net margin of 6% — two numbers that no instinct had flagged as a problem because no real recipe-level costing existed. The restaurant has operated in Medellín since 2016 with 42 tables and 19 people between kitchen and front of house. Eight years of decisions built on the phrase 'that dish looks like it sells' kept a business alive that was, on paper, surviving — but leaving 32 food-cost points uncontrolled. The mistake I see over and over: owners confuse revenue with profitability, and intuition reinforces that confusion because a full dining room feels like money. The first step in the Masterestaurant method was, precisely, opening a real costing sheet before making any other decision. When Diego F.

Re-costing 64 recipes in 60 days: what the first real number revealed

Parra and the Masterestaurant team costed all 64 active recipes at El Fogón de Andrés, 11 dishes with negative margins surfaced — dishes the business had been selling under the belief they 'drew in customers.' Those 11 dishes accounted for 17% of sales volume and consumed ingredients above their sale price once real waste and portion sizes were applied. Recipe-level food cost averaged 29%, nine points below the 38% estimated by feel, because the gap lived in specific dishes distorting the overall average. The lesson for any owner: averages hide the recipes bleeding cash. Without breaking it down dish by dish, you cannot know which items fund the business and which drain it — regardless of how many years you have been in the trade. Removing 11 dishes from the menu and renegotiating 3 supplier contracts in 60 days is not a decision you make on instinct; you make it when the numbers leave no room for doubt.

Cutting dishes and renegotiating suppliers: a data decision, not a gut feeling

At El Fogón de Andrés, the 3 renegotiated contracts involved variable-price proteins purchased in bulk for discounts — a practice that drove monthly inventory waste to 9%. With a real consumption forecast built from costed recipes and weekly order frequency, waste dropped to 3%, a 6-point swing that in a restaurant of that size represents between 800,000 and 1,200,000 Colombian pesos of monthly difference. Buying on discount without a consumption forecast is one of the most expensive and least visible operating errors; intuition dresses it up as savings. One of the starkest gaps between data-driven and intuition-driven decisions is how fast you catch a losing dish. With the Masterestaurant method and weekly indicator reviews, detection averaged 7 days from the moment a dish started eroding food cost. Without that system, El Fogón de Andrés averaged 14 months before pulling a problematic dish — and only acted when customer complaints or a sales drop made the problem obvious.

Reaction speed: 7 days with data versus 14 months without it

Those 14 months on a dish with a negative 5% margin and 40 covers per week add up to roughly 560 portions sold at a loss before anyone reacts. Intuition has no alarm; data does, and the difference is measured in cash leaving the register every week without anyone noticing. El Fogón de Andrés lifted its net margin from 6% to 14% in the first six months of weekly indicator reviews — without opening a second location or raising the average ticket. The 8-percentage-point gain came from three sources: eliminating negative-margin dishes, reducing inventory waste, and aligning staff scheduling with real demand peaks rather than the owner's memory of when 'people tend to come.' Break-even was calculated by keeping payroll and rent separate from dish-level costs, a rule the Masterestaurant method establishes from day one. When an owner knows exactly how many covers are needed before generating profit, operational decisions change in nature: they stop being bets and become adjustable variables.

Table turns: from 1.8 to 2.6 per service with data-driven scheduling

Peak-hour table turns rose from 1.8 to 2.6 per service when El Fogón de Andrés began scheduling staff based on historical customer flow records instead of the owner's perception of when 'high season' is. The 0.8 additional turn translates, across 42 tables with a 45,000-peso average ticket, to roughly 1,512,000 pesos in additional revenue per service on the busiest nights. The change required no extra servers or kitchen capacity — it required knowing precisely which days and hours concentrate 60% of weekly traffic. Intuition overestimates the 'usual' peaks and underestimates the real ones; 12 weeks of recorded data reveal patterns that an owner's memory quietly distorts. The mistake I see over and over when an owner arrives at Masterestaurant with eight years of experience is always the same: confusing accumulated experience with current information. Knowing the trade does not replace knowing this week's food cost, this dish's margin, or this month's break-even.

The definitive argument for the owner still hesitant about the shift

Intuition is useful for reading a crowded room, calming a nervous server, or negotiating face to face with a supplier. It does not work for deciding which dishes sustain profitability, how much inventory to buy, or when to adjust prices. El Fogón de Andrés proves it with numbers: 8 years of gut decisions produced a 38% food cost and a 6% margin; 6 months of data and weekly reviews produced a 29% food cost and a 14% margin. The single concrete action: cost every recipe this week, all of them, using real portions and real waste. Reaction speed to a losing dish: 7 days with data vs 14 months on average when deciding by intuition. Food cost accuracy: 29% measured per recipe with Masterestaurant vs 38% estimated by eye over eight years. Monthly net margin: 14% with weekly indicator reviews vs 6% with monthly or nonexistent reviews. Inventory waste: 3% with real consumption projection vs 9% with bulk, discount-driven purchasing. Table turnover at peak hour: 2.6 turns per service with data-scheduled staff vs 1.8 turns with intuition-based shifts.