Dessert delivery: the numbers that decide whether you sell or just move boxes

What is actually left from a dessert delivery order?

A 6.90 USD order carrying a 30% food cost leaves 2.68 USD clean once the 25% commission is paid, and that figure rules the business, not the menu price.

Break it down: 2.07 USD goes to ingredients, 1.73 USD to the aggregator fee, and from what remains you still subtract packaging — between 0.35 and 0.55 USD for a dessert that travels properly protected — plus the night shift's share of labor. Push it down to a single 4 USD unit and the arithmetic collapses: 1.20 USD of product, 1.00 USD of commission and 0.45 of packaging leave 1.35 USD that will not cover the kitchen minute or the oven's energy peak. So the two-unit minimum ticket is no marketing trick; it is the condition under which the unit economics exist at all. Negotiating commission is nearly impossible when one operator owns the market, and the numbers confirm it bluntly.

The channel is concentrated, and that concentration sets your fee

DoorDash closed 2024 with 60.7% of US delivery against Grubhub's 6.3% (Earnest Analytics 2024); in India, Zomato and Swiggy together exceed 95% of online orders (Business of Apps, Food Delivery App Report 2025); in China, Meituan and Ele.me pass 90% combined (Mordor Intelligence, 2025); and Grab took 53.9% of Southeast Asia (Momentum Works 2024). A duopoly never lowers the fee; it raises it when it can. The operating read is direct: if your gross margin per dessert cannot survive a 30% commission — the reasonable ceiling to assume when you project — the problem is not the aggregator, it is the price on your digital menu. Fix that before you sign. Diego F. Parra argues that dessert delivery is a business with its own unit economics, and at Masterestaurant it gets built as a virtual brand from day one even when the kitchen is shared.

A virtual dessert brand, not an extension of the restaurant

The reason is measurable: while dessert rides on the restaurant's listing, 100% of orders pay a fee; once it exists as a brand with its own Google Maps profile, close to 33% of orders arrive commission-free. Across 900 monthly orders of 6.90 USD, that third returns roughly 690 USD a month to the till, a little over 8,200 USD a year that used to fund the app instead. One context figure holds up the thesis: Asia-Pacific already concentrates 48.0% of global cloud kitchen revenue (Grand View Research, Cloud Kitchen Market 2025). The model works; what is missing is your accounting split. A tiramisu that arrives tilted costs you more than a refund: it costs you positions. Three-star ratings are read inside the aggregator's algorithm as a low-quality signal, and a listing sliding from 4.7 to 4.2 loses visibility exactly in the band where volume is won.

Packaging rules the review, and the review rules the ranking

The investment is trivial next to the damage: a rigid base plus a tamper seal adds 0.35 to 0.55 USD per unit, around 6% of a 6.90 USD ticket. Weigh that against losing 20% of orders after dropping one rung in the dessert category. The most repeated mistake here runs opposite to intuition: owners who shave 0.18 USD off the container and then buy 1.50 USD promotions per order to win back the visibility they gave away themselves. Dessert peaks at an hour that does not match the restaurant's, and the hidden margin lives right there. That 11:40 pm Friday order lands when the hot line has already closed, labor is paid and the oven still holds residual heat; the marginal cost of that unit is basically product alone. Concentrate 60% of your dessert orders between 9 pm and 1 am and you are not competing against the neighborhood burger joint or the sushi place, you are competing against a closed supermarket.

The time window: the asset almost nobody works

Now the counterfactual: suppose you open the dessert brand only at lunch. You land in the aggregator's most saturated band, you buy promotion to surface, your ticket competes with 12 USD full menus, and dessert ends up as a 3 USD add-on. Same recipe, half the margin. Last-mile automation does not move your P&L yet, and that deserves saying without decoration. Serve, Starship and Nuro together held 18% of global delivery robot fleets in 2024 (Mordor Intelligence, Autonomous Delivery Robots Market 2024): a fragmented, geographically narrow market, useful on campuses and dense urban corridors, irrelevant to a shop delivering nine blocks away. Inside the kitchen things are riper, with North America concentrating 40.8% of food robotics (Grand View Research, Food Robotics Market), though the equipment that pays back is portioning and repetitive baking, not assembly arms. My recommendation is dry: do not capitalize a robot before packaging, time window and minimum ticket are stable.

Robots and automation: what is real and what is still a pilot

Those three fixes cost under 600 USD and move far more margin. Benchmarks do not apply the same way across the three sizes, so land them before you decide. Small operation, up to 300 orders a month: skip the virtual brand for now, raise the minimum ticket to two units, armor the packaging and live with the 25% commission while you stack reviews; your target is 4.6 stars, not volume. Mid-size, 300 to 1,200 orders: this is where you open your own Google Maps listing and chase that commission-free 33%, worth some 690 USD monthly at 900 orders; dessert food cost must sit under 32% with packaging already loaded in. Multi-unit group: centralize production in a single night kitchen, dispatch from the nearest point and negotiate fees on aggregate volume, because the only real lever against a market where one player holds 60.7% is consolidated volume.

Where these benchmarks come from and how far they reach?

The market figures in this piece come from four houses that publish methodology: Earnest Analytics, measuring US delivery share from card transactions; Momentum Works and Business of Apps for Southeast Asia and India;

Mordor Intelligence and Grand View Research for cloud kitchens, robotics and autonomous delivery. The limits must be stated. None of those sources breaks out the dessert category on its own, so platform shares tell you about the channel's bargaining power, not about your demand. The unit-economics figures — 6.90 USD ticket, 2.68 USD net, 0.35 to 0.55 USD of packaging — are arithmetic built on price and published commission, reproducible against your own menu in five minutes. Swap them for yours: open last month's settlement statement and recompute net per order before you touch anything else. The difference sits in WHO owns the order, not in the recipe. Under the traditional method the customer belongs to the aggregator: you rent demand and pay 25% to 30% every single time, forever.

Where the two paths genuinely split?

Under the virtual-brand approach, 33% of orders arrive through Google Maps commission-free, which on 900 monthly orders keeps roughly 690 USD in your till instead of sending it out as fees.

Packaging decides the review, and the review decides the ranking. A tiramisu that arrives tilted earns a 3-star rating that the Rappi algorithm reads as a low-quality signal and punishes with less visibility; the same tiramisu on a rigid base with a seal holds 4.7 and climbs inside the dessert category. That demand window is the asset almost nobody works. Home dessert orders cluster exactly when the restaurant kitchen is shutting down, between 22:00 and 00:30, and a two-person operation running six references produces in those two and a half hours what the dining room bills in dessert across three days. Costing the packaging changes the pricing conversation.

Where the two paths genuinely split — in practice?

A 6.90 USD dessert with 2.10 USD of raw material looks like 30% food cost, yet adding box, gel and spoon pushes it to 34% and the operation crosses the 32% ceiling Masterestaurant sets as MAXIMUM;

the right fix is redesigning the box or moving to a two-unit format, never shrinking the portion. Direct orders carry a different lifetime value. A customer captured in your own database returns 2.4 times per quarter at near-zero reactivation cost, while the aggregator customer costs full commission on every purchase, even the ninth time they order the same brownie.