Prime cost: the traditional method stopped matching the digital channel, the Masterestaurant one does not

Step 1: rebuild the denominator before touching a single cost line

Your prime cost only tells the truth when the denominator is NET revenue by channel, not menu revenue piled into one number. An owner in Bogotá sent me a March P&L showing 57,8 % prime cost, comfortably inside range, while the bank showed 4.100 USD less at month end: he booked Rappi and Uber Eats orders at menu price when the platform settled between 22 and 27 % below that, and the gap never landed on any cost line. What this step produces is a five-column table —dining room, in-store pickup, own delivery, marketplace A, marketplace B— with gross and settled revenue for each over 90 days. Verify it by reconciling total settled revenue against actual bank deposits; a gap above 2 % means the error sits in the denominator, not the kitchen, and tightening portions further just burns the month. Thirty-two percent food cost per dish is the operating ceiling and the working target lives between 28 and 30 %, measured on a standardized recipe with gram weights, trim loss and real yield, never on the monthly average.

Step 2: lock food cost dish by dish with a 28-32 % ceiling

Cost the ten references that move the most volume first, because roughly 70 % of your purchasing hides there. This matters more in 2026 than it did in 2024: USDA ERS projects all food prices up 3,2 % and beef up 7,5 %, with the U.S. cattle herd at a 75-year low, so a protein plate closing at 31 % today crosses the ceiling with the first supplier adjustment. The step is done when every top-ten recipe carries a signed cost sheet with a date. Verify by re-costing two random dishes against the latest invoice. Marketplace commission is a channel cost and belongs on its own P&L line, between food cost and labor, never quietly netted out of revenue. The traditional method asks how much I spent; here we ask what is left per channel after the platform takes its cut.

Step 3: book platform commission as a cost, not an invisible discount

At 25 % commission, a dish running 30 % food cost in the dining room jumps to an effective 40 % in third-party delivery, and that dish stops funding your rent —around 53 USD per square foot per year in Los Angeles, per Pepperlot 2025— and starts funding somebody else's growth. The deliverable is a prime cost per channel, five separate figures. Verify it when the weighted sum of those five reproduces your consolidated prime cost within one point. Somebody packs the order, somebody answers the one-star review and somebody updates the holiday hours on the Google profile so the listing does not drop out of the local pack; those hours exist whether or not anyone logs them. Assign them by observation over two weeks, stopwatch and shift in hand, and value them at loaded cost rather than nominal wage: BLS reports a 14,92 USD/hour median for food service workers in May 2024, and California sets a 16,50 USD/hour minimum in 2025 even for tipped staff, so twenty misplaced weekly hours move close to 1.400 USD a month.

Step 4: measure the real labor cost of the digital channel

Diego F. Parra keeps pushing this line at Masterestaurant because it almost never reaches the P&L. It is done when a shift-lead-signed hours-by-channel matrix exists; verify it against the time clock. A monthly prime cost gives you twelve chances a year to correct course; a weekly one gives you fifty-two, and that gap decides whether you close a leak or inherit it. Kitchen waste, ungrammed portions and petty theft get fixed the same week they surface or they turn chronic, because after thirty days nobody remembers which shift did what. Build a short Friday close: inventory of the fifteen expensive references, net revenue by channel, hours for the period, prime cost calculated before Saturday noon. Running it costs about two administrative hours a week. The payback shows up fast next to ReFED's figure of 7 USD in future benefit per dollar invested in preventing food waste.

Step 5: move the cadence from monthly to weekly

The deliverable is a thirteen-week series with its chart. Four failures account for nearly every false reading, and none of them start in the kitchen. Averaging channels with different margins into one figure comes first: it REASSURES without informing, which is how a restaurant that prints money in the dining room dies from a channel it treated as secondary. Forgetting opening and closing inventory comes second, turning food cost into a purchase log. Third, measuring labor with kitchen payroll alone, leaving out benefits, overtime and servers, whose median tipped wage BLS puts at 16,23 USD/hour in May 2024. Fourth, celebrating a low prime cost achieved by cutting staff: service degrades, reviews slip, sales fall two months later. Fix them in that order. If your figure swings more than three points week over week with no operational cause, the method is what broke. Assume you grow 20 % in volume and every point of that growth arrives through a marketplace charging 25 %.

What happens if you do not separate: the scenario almost nobody runs?

Gross revenue climbs, consolidated prime cost barely moves because food cost per dish is unchanged, and you celebrate. Profit falls anyway, since each point of growth arrives with a negative margin once commission and packing hours are loaded onto it.

With margins of 3 % to 8 % in full service and 4 % to 10 % in fast casual, per WhippleWood CPAs' 2026 benchmarks, there is no cushion to fund a channel that subtracts. The paradox resolves itself right there: the digital channel is not the enemy, the average is. With prime cost split by channel, you accept that growth only when the channel closes under 60 %, or you repair it by lifting digital menu prices 12 to 18 %. Your system is properly built when you can answer six questions without opening a new spreadsheet. One: is the denominator settled revenue, reconciled to bank deposits within 2 %? Two: does a signed sheet with gram weights and unit cost exist for the ten highest-volume references?

Closing checklist: how to know it came out right

Three: does platform commission have its own P&L line? Four: are digital-channel hours assigned and valued at loaded cost? Five: does prime cost appear split by channel, each one under 60 % in full service or under 55 % in fast casual? Six: does the series run at least thirteen consecutive weeks? A single failure means the consolidated figure is not trustworthy yet. Start with question one next Monday, using the last 90 days of settlement reports downloaded from every platform. The traditional method asks how much you spent; the Masterestaurant one asks what you keep per channel once the platform takes its cut. It sounds like an accounting nuance, and it is the difference between growing and funding a marketplace's growth with your own capital. Cadence rules. A monthly prime cost gives you twelve chances a year to correct course; a weekly one gives you fifty-two, and most food cost leaks —kitchen waste, ungrammed portions, small theft— either die in the week they appear or turn chronic.

Where the two methods genuinely part ways?

The digital channel's labor cost exists whether or not anyone measures it. Someone packs, someone answers the one-star review, someone updates holiday hours on Google Business Profile so the listing keeps its spot in the local pack.

That runs six to eleven hours a week in a single-unit restaurant, and once those hours sit inside prime cost, the number climbs 1.8 to 3 points. It climbs because they were always there. Theoretical versus actual cost only means something with standard recipes. Without written grammage, variance is unreadable: you see 34 % food cost and cannot tell whether you bought badly, someone stole, or the cook plates 15 % extra every time. Traditional practice quarantines marketing; we tie it to contribution margin. A geo-targeted campaign delivering orders at 4.20 USD when contribution margin per order is 6.80 USD works; the same campaign against a 3.90 USD margin is buying volume with your EBITDA.