Delivery profitability checklist: each box with its 'done' criterion (unaudited vs audited) 2026

Box 1: I have the effective commission per platform as a number

The number outranks the contract: this box closes only with a calculated figure, never with the commission printed on the signed agreement. The proof means pulling the last 90 days per platform, adding commission paid, packaging per order, and refunds on incomplete orders, then dividing by gross channel sales. Platforms charge 15% to 30% (on a $25 order at 25% that's $6.25), but add $1.10 in packaging and a 4% refund rate and the channel actually takes $8.35: a real 33%. If you think you pay 25% and can't document it per app, the box stays empty. No group moves forward at Masterestaurant without this figure, because costing off the 25% contract rate leaves 11 points uncovered. What isn't a number isn't measured. It's assumed, and that assumption costs $1.80 to $2.50 of cash on every order.

Box 2: my app price is adjusted per channel, not copied from dine-in

A 15-22% reprice on dishes that can absorb it closes this box, never the claim of 'having it under control.' Margin is the proof: a dish running 30% food cost keeps 70% in the dining room, but on the app, at 33% commission, it barely keeps 37% before labor, so the app price has to climb to hold a similar margin. That takes per-dish profitability analysis behind every reprice, so price-sensitive dishes never move blindly. Per-channel AI pricing is what decides which dish can absorb a +24% bump. Checked correctly, this box shifts channel margin from -$1.80 to $4.10 on a $25 order. The more common mistake runs the other way: leaving the box unchecked out of fear of raising prices, and handing the platform $5.90 of swing on every order. Few boxes get checked as poorly as this one, and it might be the most revealing of the eight.

Box 3: I have a delivery P&L separate from the dining room

It closes only with a delivery-channel income statement kept apart from dine-in: gross sales, total effective commission, packaging, specific food cost, contribution margin per order, and break-even in number of orders. Without that document, the box stays empty no matter how much delivery 'feels fine.' When the channel lives buried inside the dining-room P&L, a delivery losing 7 margin points vanishes from the radar while total cash keeps climbing on app volume. In 70% of the groups Masterestaurant audits, delivery was destroying margin while never measured on its own. At 3,000 orders and $4.10 contribution margin, the channel adds $12,300 monthly to fixed costs. That figure closes the box. The feeling doesn't. Respecting the hard costing rule, with the layers kept apart, is the only thing that checks this box. Commission gets subtracted in the delivery-channel costing because it's a direct variable cost.

Box 4: I separate commission from payroll in my costing

Food cost maximum sits at 32 points, never recommended, and measures ingredients only. Payroll, rent, and utilities aren't charged to the plate or the order: they go to break-even, calculated separately. If commission and payroll land on the same costing sheet, the box stays empty: that costing is wrong, and it ends either in mispriced app menus or in a dining room quietly subsidizing the channel. Confusing these layers is the root of half the delivery mistakes I correct in consulting work. At Masterestaurant we always keep them apart: food cost to ingredients, commission to the channel, payroll and rent to break-even. The proof is a costing sheet with nothing blended. Neither a website nor the wish for one checks this box: it takes a 4-6% gateway actually running, with the direct channel migrating real volume off the app. Against the platforms' 25%, a direct channel at 4-6% recovers nearly 20 commission points on every migrated order.

Box 5: I have a direct channel operating at 4-6%

This isn't about dropping the apps, which bring reach a direct channel can't reach alone, but about turning a repeat customer into a direct one with one simple incentive: 10% off their second order. Even giving away that 10%, you still land ahead of the app's 25%. The box closes once direct exceeds 15-22% of total delivery, tracked month over month. Masterestaurant groups reach 22% within 6 months and recover $9,000 to $13,000 monthly. The proof is migrated volume, not the channel's mere existence. This box rewards habit, not a one-time fix: it gets checked when channel mix is reviewed monthly with the formality of a board KPI. What percentage runs through the expensive 25% app and what runs through the 5% direct channel, with the margin of each order and how it's trending, is what the box demands, not an occasional glance.

Box 6: I review channel mix monthly as a board KPI

Building a direct channel and letting it stall at 8% because nobody reviews it is the mistake that resurfaces in every fresh audit. It takes an owner, a monthly cadence, and a target number. The Masterestaurant goal is pushing direct from 0% to 22% across two quarters and holding it there. Without this box checked, everything built earlier erodes on its own. The proof is the monthly meeting minutes with the mix figure. From profitable to optimized: that's what this box delivers, and it only gets checked once dynamic per-channel pricing runs on real artificial intelligence, not a promise to get to it eventually. AI cross-references hourly demand, platform commission, and per-dish elasticity to decide which app prices can rise without scaring off orders. It isn't raising everything 18% blindly: it's identifying the dish that can absorb 24% and leaving the price-sensitive one alone.

Box 7: I use AI for dynamic per-channel pricing

With per-channel pricing, Masterestaurant groups raise the average app ticket 11% while losing under 3% of orders, a trade that nearly always nets positive cash. This box isn't mandatory to start (measurement and costing come first), but it squeezes the channel once those two are closed. Its proof is an app ticket measured before and after, with order variation documented. Count, don't feel: that's the order behind this eighth and final box, which tallies how many of the previous seven closed with real proof. Fewer than 6 of 8 with evidence means delivery almost certainly loses $1.80 to $2.50 per order, hidden inside the dining-room P&L. Commission, reprice, separate P&L, and cost layers, boxes 1 through 4, are the minimum to stop bleeding cash. Boxes 5 through 8 take the channel from profitable to optimized. Masterestaurant runs this checklist in the first week with every group and repeats it each quarter, because channel mix erodes on its own if nobody pushes it.

Box 8: fewer than 6 boxes checked with proof is a red alarm

Without a 'done' standard, nothing is done, only assumed. This week's action is simple: print the 8 boxes and check only what you can prove with a number.