Rappi delivery strategy: the traditional method versus the Masterestaurant method

1. Why does the Rappi channel eat your margin even as revenue climbs?

Because the delivery order gets costed inside the dining-room P&L, and that is where the bleeding stays invisible.

The quarterly report I mentioned —4,100 orders, an average ticket of 38,000 pesos— showed respectable gross sales, yet channel profit turned negative the moment we loaded packaging, hot-line waste and the hours of the assistant who assembles bags. Commission was not the culprit. The missing piece was a separate income statement, in a market that moved USD 1.18 billion in Colombia during 2024 and grows at a 7.32% CAGR through 2029 (Statista Market Insights, 2024). Volume rises on its own and hides the inefficiency: you bill more every month while contribution margin per order falls, and no dashboard tells you. Close that blind spot before you argue rates with anyone.

2. A P&L per channel: the accounting decision that orders every other one

Open a separate income statement for delivery and charge it every peso the channel consumes. Platform commission, packaging, the extra waste of dishes that travel badly, and —almost nobody does this— the labor hours dedicated to assembly and dispatch. With delivery apps in sustained year-over-year growth, a channel without its own costing stopped being a tolerable experiment: it is a business line you run blind. At Masterestaurant that separation opens every diagnosis, because it changes the whole conversation. Once delivery carries its own prime cost, the question stops being «how much did we sell through Rappi?» and becomes «how much did we KEEP on each order dispatched?». Two different questions, and only the second one decides anything.

3. The dining-room menu is not the delivery menu

Publishing the full menu on the app looks generous and is the fastest way to give away profit. Long-execution, high-food-cost dishes leave with a flat 20% surcharge that never covers commission plus packaging plus waste, while they occupy the griddle your kitchen needs for the dining room. The rule I apply: publish what survives the trip, sits below 30% food cost and executes in under twelve minutes. That usually leaves 35% to 45% of the original menu. Against the urge to show everything, the short catalog sells more per order, because a guest looking at nine options decides in twenty seconds and adds a side. Mexico, at USD 9.22 billion with a 14.66% CAGR (Statista, 2024), rewards whoever tidies the catalog before scaling.

4. Under USD 500K a year: don't scale, prune

In this band there is one decision: no virtual brand, no buying more volume, prune the current catalog until every published dish yields positive contribution margin after commission. The operating threshold is 25% contribution margin per order, calculated on the app price, never the dining-room price. A dish that misses it leaves the catalog that same week. The small band remains fertile ground —the global cloud-kitchen market was estimated at USD 45.65 billion in 2024 (MarkNtel Advisors, 2024)— but the recurring error here is chasing that growth before the arithmetic is settled. First the pruned catalog and a properly built surcharge; afterward, and only afterward, the conversation about volume. A venue in this band with twelve profitable dishes earns more than one with forty undifferentiated ones.

5. USD 500K to 1M: cost the order and split the station

Here the lever is operational rather than commercial: build a separate dispatch station and measure channel prime cost every week. The operator in the case lived exactly this problem, with two channels fighting over the same hot line and an assistant whose cost nobody charged to delivery. The threshold I set him: channel prime cost —food plus direct dispatch labor— stays under 62% of net delivery sales, already net of commission. Two consecutive weeks above it and you cut catalog or raise the surcharge, no debate. That flat 20% surcharge also needs a dish-by-dish review, because a 22% food-cost starter and a 34% main course do not deserve the same treatment. Brazil, at roughly USD 18.8 billion in 2024 (Statista, 2024), proved that volume forgives nothing when there is no dedicated station.

6. Above USD 1M: a virtual brand only if the kitchen has slack

A virtual brand earns its place when your kitchen runs below 70% capacity in the valley shifts and you can launch it on existing inventory. That is the threshold. Without real slack, a second brand generates no incremental sales: it cannibalizes the first and adds dispatch complexity that degrades both their times. With slack, the arithmetic turns good, because fixed cost is already paid and each extra order arrives almost free of structure. Mexico's cloud-kitchen market reached USD 1.1 billion in 2024 with a 10.74% CAGR toward 2033 (IMARC Group, 2024), and that is the window. One more rule: the virtual brand shares no dishes with the main one —it shares INGREDIENTS, a very different thing— or the guest spots the trick and you lose both reputations at once.

7. Above USD 5M: automate the back of house before you touch commission

At this volume, an order somebody types by hand into the KDS costs you more than any commission point you might negotiate. Integrate the app into the ticket system and kill the manual entry: fewer dispatch errors, fewer remakes, fewer dead minutes. This band holds the profile we all recognize —the large-format themed venue, or the one backed by a media figure— where delivery usually runs as an appendix of the dining room and ends up damaging the brand through delivery times nobody tracks. My criterion: above five million, delivery either has an owner with a name and a budget, or it should not exist. Asia-Pacific ghost kitchens went from USD 21.73 billion in 2024 toward a projected USD 60.59 billion by 2032, a 12.8% CAGR (Coherent Market Insights, 2024): the scale arrived, dispatch discipline has not.

8. Above USD 10M: govern the channel as a business unit

A group or chain in this band needs formal channel governance: a consolidated P&L plus one per location, a monthly committee reviewing contribution margin by unit, and a written shutdown threshold. Mine: if a location spends two quarters under 18% contribution margin in delivery, with catalog and station already fixed, the channel goes dark at that location. It hurts and it works, because delivery is no strategic obligation: it is a line that must earn its place. Consider what happens if commission rises three points tomorrow —the sector has seen it more than once—: with a channel P&L you know within forty-eight hours which locations and which dishes fall underwater, and you act; without one, you find out next quarter, after the loss. Ghost kitchens ran near USD 70.4 billion in 2024 (Research and Markets, 2024), and that pool hands no margin to whoever fails to measure it.

9. The four differences that move EBITDA

The first one is accounting, and it orders everything else: the traditional playbook runs a single income statement, so delivery either subsidizes or gets subsidized with nobody noticing, while the Masterestaurant architecture opens a per-channel P&L and charges every order its commission, packaging, waste and the hours of whoever packs the bag. With delivery apps still expanding year after year, an uncosted channel is no longer an experiment. It is a blind business line. Menu design is the second gap. Publishing the whole dining-room card looks generous and turns out to be the fastest way to give margin away, because long-execution, high-food-cost dishes travel with a flat 20% uplift that never covers commission plus packaging. We filter on two hard criteria — a 32% food cost ceiling per dish and execution times compatible with two simultaneous channels — and then run channel-specific menu engineering, where combos lift the average ticket without touching unit price. Third comes installed capacity.

10. The four differences that move EBITDA — in practice

A kitchen that bills hard from noon to 2:00 p.m. and again from 7:00 to 9:00 p.m. carries paid hours that produce nothing, which makes a virtual brand on the existing line the cheapest scalability play available today. Cloud kitchens are growing at an 11.65% CAGR between 2022 and 2032 according to Statista/Toast (via OysterLink), with close to 7,606 ghost kitchens already active in the US. Building a dark kitchen from scratch is not the entry ticket. The fourth difference is the one fewest operators work on: data governance. If your read on the channel is the aggregator panel at month close, you decide with stale numbers biased toward revenue. A dashboard that crosses food cost variance, orders per hour against channel break-even and acceptance time by window turns the operation into a decision architecture — and it hands you operational due diligence arguments when you sit down to negotiate with the app or raise capital.