Selling on delivery apps in 2026: what actually moves cash, and what is just noise

Why can 41 million pesos in app sales leave a 4% margin?

Because commission, co-funded promotions and in-platform advertising eat 25% to 30% of gross sales before you buy a single gram of protein.

The Chapinero operator who started this conversation billed 41 million pesos in July 2026 and had 11.3 million withheld across those three lines; with a 34% plate food cost and 4% packaging, the channel's operating margin landed at 4%. That picture is not a Colombian anomaly: 65% of limited-service operators already offer delivery according to the National Restaurant Association (2025), and most of them run the same menu at the same dining-room price. That is where the hole opens. A digital menu nobody costed separately is not a sales channel, it is a leak that generates revenue.

REAL TREND — local search became the front door to delivery

Direct ordering driven by local search is the highest-return trend of 2026, because it captures high-intent demand without paying commission. Google reports double-digit growth for the third straight year in «near me» queries that end in a transaction, and the Google Business Profile listing now supports a direct order button that skips the 15% to 30% the app withholds. The measurable signal: the direct channel's share of total delivery orders, tracked month by month in your POS. A neighborhood restaurant with its own kitchen and a short radius should take the listing to 100%, upload twenty real plate photos —not stock images— and ask 30% of dining-room guests for a signed review. A five-location chain also needs separate listings per site, because a single parent listing distributes traffic badly and cannibalizes whichever location sits closest to the customer.

REAL TREND — the algorithm stopped rewarding discounts and started rewarding reliability

Rappi, Uber Eats and DiDi Food now weight acceptance time, preparation time, cancellation rate and rating above whatever discount you co-fund. DiDi Food delivered more than 360 million orders in Mexico over five years (DiDi Food, 2024) and DoorDash closed 2024 with roughly 2,583 million orders, up 19% year over year in the fourth quarter (DoorDash, 2024 results): at that volume, the platform optimizes for completed deliveries, not for coupons. The measurable signal to watch is your average prep time and your cancellation rate caused by missing product. Bring prep time down to nine minutes on the ten best-selling dishes, freeze the rest of the digital menu, and visibility rises without spending a peso on promotion. Small kitchen: cut SKUs. Large kitchen: assign one station exclusively to the channel.

REAL TREND — virtual brands consolidate while ghost kitchens get pruned

Ghost kitchens already account for around 15% of U.S. foodservice delivery sales (Statista), but the game moved from launching brands to pruning them. What works today is one virtual brand per kitchen, six to eight SKUs sharing mise en place with the core menu, and a 32% plate food cost as a ceiling rather than a target. Diego F. Parra keeps pressing one criterion at Masterestaurant that saves money before you spend it: if the virtual brand needs its own purchasing, it is a new restaurant dressed up as a cheap idea, and it must be costed as such. The measurable signal is plain: sales per square meter of kitchen and station utilization at peak. If the new brand lifts neither one within ninety days, close it without ceremony.

What price should the delivery menu carry for the channel to leave profit?

Between 15% and 20% above the dining-room price, and that gap is arithmetic on commission, not greed. At a 25% commission, a 30,000-peso dish leaves 22,500 in the app;

with a 32% food cost (9,600 pesos) and 1,200 in packaging, contribution drops to 11,700 before payroll. Price it at 35,500 —18% higher— and contribution climbs to 15,400, which is what holds the break-even point up. Remember the costing rule: payroll, rent and utilities are not loaded onto the plate, they are covered by accumulated contribution. The most repeated error I find in settlement reports is raising the digital price by 8% and assuming commission is covered; at 25% commission that 8% barely pays for packaging and half the platform fee.

The overrated trend: generative AI on the menu before the data is clean

More than 25% of operators already use artificial intelligence according to the National Restaurant Association (Restaurant Dive, 2026), and a good share of that spending sits in the wrong place. Buying an engine that rewrites dish descriptions or «optimizes» prices while theoretical inventory does not match the physical count means paying to decorate bad data. Operators who profit from AI apply it first to demand forecasting and purchasing, never to menu copy. The contrast matters: 63% of operators planned to invest in digital marketing in 2024 and 48% prioritized point-of-sale technology (National Restaurant Association, 2024) — that order, POS before ornament, is the right one. Skip generative menu assistants for now. Watch instead the forecasting models that cut waste, because there the saving is measured in kilos and shows up on next month's invoice.

2026 horizon: what to adopt this quarter and what to keep under watch

Adopt three things now and hold two under observation. Now: a digital menu costed separately with a 15-20% price premium; a business listing at 100% with direct ordering; and weekly control of prep time and cancellations, which is what actually moves the algorithm. Under watch: in-app advertising —worth it only when your digital contribution margin clears 45%— and the aggregators of aggregators, which promise to consolidate channels and usually add one more commission layer. The market leaves room to experiment: Statista projects 1.45 trillion dollars in global food delivery by 2026, and in Asia Meituan and Ele.me already exceed 60 million orders per day (Mordor Intelligence, 2025). Volume exists. Margin is what runs short, and margin gets decided in the costing, not on the platform.

What would happen if tomorrow the app raised commission from 25% to 30%?

With the Chapinero operator's structure, that five-point jump wipes out the margin and turns it negative:

4% minus five points of gross sales is an operating loss, and the only quick way out would be another 7% on the digital price, dragging conversion down with it. So the decision that matters is not which app to join but how much direct sales you built while the app was bringing you customers. There sits the paradox of the channel: the platform becomes profitable precisely when you stop depending on it, because then you can negotiate, switch off promotions and pick zones. Treat the apps as paid acquisition with a cost per new customer, not as your dining room. Start this week by measuring what share of your delivery orders already arrives through your own channel; below 20%, that is your risk number.

Real trend or fad: how to tell before you spend

REAL TREND — Local search became the front door to delivery. Google reports that «near me» searches ending in an order have grown double digits for a third straight year, and a Google Business Profile listing with a direct order button captures that traffic commission-free. It hits neighborhood restaurants with their own kitchen and a short delivery radius first. Action within 90 days: complete the listing to 100%, upload 20 real plate photos, switch on your own ordering link, and ask 30% of dine-in guests for a signed review. REAL TREND — App algorithms stopped rewarding discounts and started rewarding reliability. Rappi, Uber Eats and DiDi Food weight acceptance time, preparation time, cancellation rate and rating above in-app ad spend. It hits kitchens with long menus and a single hot line first.

Real trend or fad: how to tell before you spend — in practice

Action within 90 days: trim the digital menu to 12-15 items, time each one, and accept orders in under 60 seconds. REAL TREND — The virtual brand stopped being an experiment and became a capacity tool. A second brand running through the same kitchen during dead hours raises equipment utilization without extra payroll, provided it shares 80% of the inventory. It hits operations with idle kitchens between 3 and 6 p.m. first. Action within 90 days: launch ONE virtual brand with four dishes drawn from existing inventory and measure it for thirty days against the cost of that idle hour. EXPENSIVE FAD — The premium in-app subscription promising «priority visibility». When the algorithm weights operations rather than spend, that plan buys impressions a slow kitchen cannot convert.

Real trend or fad: how to tell before you spend — key points

Measure your prep time before paying: above 18 minutes, the money evaporates. EXPENSIVE FAD — Opening three virtual brands at once to «see which one sticks». Each brand fragments your rating, your history and your team's attention; three new listings without reviews compete against each other for the same scroll. One brand, thirty days, a decision backed by data. HALF-TRUTH — Dark kitchen vs physical restaurant is not an either-or. The honest comparison is not site versus dark kitchen, it is cost per delivered order against customer lifetime value. A restaurant with a dining room captures the customer twice —table and delivery— and that second capture is free; a dark kitchen depends on someone finding it in a list where everyone looks the same.