How to increase restaurant sales on Rappi: the volume myth and the honest alternatives for 2026

Order-level unit economics rule: run the numbers before you chase volume

Before you touch the photo of your signature combo, work out what each order actually leaves you, because everything else depends on that figure. The math fits on a napkin: commission of 18% to 30% depending on your category and tenure on the platform, packaging at 4% to 7% of the ticket, and the food cost of the delivery version, which has to stay under 32% per the Masterestaurant operating ceiling. Add those three blocks against a reference ticket: with 27% commission, 6% packaging and 31% food cost, you keep 36 gross points to cover kitchen labor, utilities and profit. Stack a 20% promotion on top and those 36 points drop to 16, which means you are working for Rappi. The grill house in Bucaramanga billing 61 million pesos a month landed at 9% contribution margin through exactly that stacking, and no ranking lever fixes math that was broken at birth.

A delivery menu is not your dining-room card with photos; it is a different product

What sells in the dining room rarely survives the app, and the costliest habit owners repeat is publishing all 46 dishes and hoping the algorithm picks for them. Delivery rewards a short menu: 12 to 18 items, built from dishes that hold up through a 22-minute ride without losing texture and whose food cost stays under 32%. Pull the delicate fried items, the creams that split and the salads that arrive wilted, because every complaint costs you twice — the refund and the rating. Lift the ticket with combos that pack into a single box rather than loose add-ons that multiply packaging; average U.S. delivery tickets run between USD 20 and 35 per order (Lightspeed 2025), and your target is the top of your category's range through more food per box, not more boxes. Three variables inside the app are yours, and a fourth one is bought.

Position, photography and prep time: the levers you genuinely control

First comes declared prep time: cutting it from 30 to 18 real minutes lifts conversion because diners filter for speed, though declaring a number you cannot hit will wreck your rating within two weeks. Second is photography; a 40-shot session runs 600,000 to 1.5 million pesos and usually pays for itself. Third, a rating held above 4.6 carries more weight in the ranking than any one-off promotion. The fourth lever is the platform's paid placement, and here you stay cold: buy position ONLY when your contribution margin per order clears 30 points, since advertising on 12 points of margin is buying losses wholesale. Rappi reported 35 million active users in 2024 (Rappi operating report 2024) — the traffic is real, the question is what you pay for it.

When Rappi stops being enough: the number that gives it away?

There comes a point where the platform stops being a channel and becomes an expensive partner that contributes no capital, and one figure gives it away:

once more than 55% of your total sales come through a single aggregator and your average contribution margin per order sits below 20 points, you no longer run a restaurant, you run a contract kitchen. The second symptom is your customer base, because on Rappi the diner belongs to Rappi, so three strong years can leave you with zero phone numbers of your own. Third comes regulatory asymmetry: New York caps commissions by law at 15% for delivery plus 5% for other services (Restaurant Business 2023) and San Francisco at 15% (Restaurant Dive 2020), while Latin American operators pay 27%. When all three show up together, the conversation is no longer about selling more inside the app. An owned ordering site with a payment gateway offers the best economics of any alternative, and it is the one people launch far too early.

Your own channel: for the operator who already has customers by name

It costs between 3.5 and 8 million pesos in year one counting domain, platform and advertising, with a two- to three-month learning curve until the team stops improvising dispatch. In exchange it leaves 34% to 48% contribution margin, since you only pay 2.9% to 4.5% in gateway fees instead of a double-digit commission. Who is it for? The restaurant that ALREADY has 400 to 800 identified repeat customers with a phone number or an email. Opening it without that base means building a store in the desert and paying to bus pedestrians in. One encouraging data point: 46% of U.S. diners prefer third-party apps and order nearly five times a month (DoorDash via Restaurant Business 2024), which leaves the other 54% available to you. A virtual brand run out of your existing kitchen costs almost nothing — photography, registering the name and a new menu — and matures in two weeks; a dark kitchen built from scratch runs 60 to 180 million pesos and takes six to ten months to break even.

Virtual brand versus dark kitchen: two roads people confuse and should not

Confusing them gets expensive. The virtual brand exists to capture categories your main brand cannot touch without confusing the diner: a home-cooking house launches wings using the same chicken, the same oil and the same cook. Its limit is physical rather than commercial, and it shows up once the kitchen passes 70% occupancy at peak hour and the two brands start fighting over the fryer. A dark kitchen suits the operator who already validated demand and needs capacity; Statista counted over 20,000 operating ghost kitchen locations in the United States in 2023, and the survivors opened on proven demand, not on a hunch. Signing up with a second local aggregator usually costs nothing upfront and reduces dependence without demanding new skills, though it rarely moves past 15% to 25% of your current volume because traffic concentrates around the dominant app.

Secondary aggregators and AI over WhatsApp: the cheap channel almost nobody organizes

More interesting is WhatsApp ordering with an AI agent that takes the order, confirms the address and charges through a payment link: implementation runs 1.2 to 4 million pesos with a monthly fee of 150,000 to 450,000 pesos, and the curve lasts three to five weeks until the bot stops fumbling add-ons. A consultant's warning belongs here: if your operation has no written dispatch protocol, the AI merely automates the chaos and multiplies your errors. As Diego F. Parra insists during Masterestaurant audits, you fix the process first and lay technology on top of it afterwards, never the other way around. Sometimes the right answer is to change nothing, and saying so out loud is part of the job. If your contribution margin per order on Rappi clears 30 points, the platform accounts for under 40% of sales and your kitchen runs below 60% of peak-hour capacity, opening your own channel now will drain management attention for at least six months before it returns any cash.

When NOT to switch: staying put can be the profitable call?

Turn it around: what happens if Rappi shuts you off tomorrow and you live on it?

You lose 60% of revenue in 24 hours, spend three months building the owned channel and two more filling it — five months of full payroll against dining-room sales alone. That is the real argument for diversifying, not the commission. And the commission does hurt, though it also buys traffic you would not know how to buy yourself: nearly 75% of U.S. restaurant traffic already happens off-premise (National Restaurant Association 2025). Run the margin-per-order math this Friday using last week's real ticket, and decide on that number. DIRECT CHANNEL with an ordering site and payment gateway: USD 850 to USD 2,000 in year one counting domain, platform and media, with a two to three month learning curve. It returns 34% to 48% contribution margin because it only pays 2.9% to 4.5% in gateway fees.

Five honest alternatives, with cost and intended owner

This suits the restaurant that ALREADY has 400 to 800 identified repeat customers; launching it without a customer base means building a store in a desert. VIRTUAL BRAND out of the same kitchen: near-zero investment —photography, registration and a new menu—, a two-week curve, and it captures categories your main brand cannot touch without confusing the guest. A home-style restaurant launching a wings brand uses the same chicken, the same oil and the same cook. The limit is real: if the kitchen already runs at 80% of peak capacity, a virtual brand adds no sales, it cannibalizes ticket times. DARK KITCHEN FROM SCRATCH: USD 11,000 to USD 44,000 depending on city and equipment, a six-month curve to break-even, and it only makes sense once delivery volume already justifies a dedicated kitchen. Weighing dark kitchen vs physical restaurant without proven demand is the costliest mistake in this segment: a dark kitchen does not create demand, it serves existing demand better.

Five honest alternatives, with cost and intended owner — in practice

SECONDARY AGGREGATORS —iFood in Brazil, Didi Food, PedidosYa depending on the market—: zero entry cost, a one-week curve, and their value lies in negotiation. Sitting on two platforms lowers effective commission because the second one gives you leverage with the first. The risk is operational: two tablets, two prep clocks, two rating algorithms and one expediter. AI-DRIVEN WHATSAPP ORDERING: USD 45 to USD 220 monthly depending on conversational agent volume, a three-week curve, and it offers the best effort-to-return ratio for a mid-sized restaurant. The agent takes the order, suggests the side, charges by payment link and leaves the phone number in your database. Here comes the uncomfortable part: a badly configured agent that answers in 40 seconds loses more orders than it wins.