Recurring signal
The pattern
The strongest transitions do not begin with an app deletion. They begin when a restaurant separates discovery from repeat demand, then gives regulars a direct path back.
Marketplace delivery performs a real job. It puts a restaurant in front of diners who may not know it, supplies delivery logistics, and can turn spare kitchen capacity into revenue. Operator conversations rarely frame that reach as worthless. The recurring problem appears later, when the marketplace remains the default doorway for diners who already know the restaurant.
At that point, one channel is being asked to do two different jobs: acquire a first order and process every order after it. The first may justify an acquisition cost. The repeat order deserves different economics. The shift that moved the needle was not marketplace versus direct as a winner-take-all choice. It was marketplace for discovery, direct for the relationship that followed.
Pull-quotes below are editorial reconstructions of recurring operator accounts. Identifying details are changed.
Discovery and repeat demand are different jobs.
Operators who made the cleanest transition kept the marketplaces visible while changing what happened after a diner discovered the restaurant. A first marketplace order could still arrive with useful reach and convenient logistics. Packaging, receipt language, the restaurant website, and the ordering link then made the direct channel easier to find for the next order.
This sequencing reduced the risk of an abrupt exit. New demand continued to arrive through the marketplace while repeat demand gradually moved toward a restaurant-controlled channel. The test was not whether marketplace revenue fell in isolation. The test was whether total online revenue held while the direct share and contribution margin improved.
The most useful measurement separated channel mix from total demand. Operators watched the direct share, marketplace share, average ticket, repeat-order behavior, and gross dollars retained after channel costs. That view made a healthy transition distinguishable from a simple sales decline. Marketplace volume could fall for the right reason when direct volume rose by the same amount; it could also fall because demand weakened. The P&L alone did not explain which event had occurred.
“The marketplace still did the discovery job. The expensive part was paying the discovery rate again when a regular returned through the same app.”
The statement changed the conversation.
A headline commission can look like one marketing expense. The full statement usually shows a wider operating cost: marketplace commission, delivery service, sponsored placement, promotion participation, packaging, menu-price adjustments, refunds, and staff time spent reconciling channels. The exact mix varies by contract and market, but public merchant plans establish the range. DoorDash currently publishes delivery tiers of 15%, 25%, and 30%; Uber Eats publishes U.S. marketplace tiers of 20%, 25%, and 30%.
Those percentages arrive before food, labor, occupancy, or packaging is paid. The National Restaurant Association describes a typical restaurant as operating near a 5% pre-tax margin, with food and labor each consuming about one-third of sales. A twenty-point channel charge therefore does not trim a large surplus. It competes with line items that already consume nearly the entire sales dollar.
The less visible cost is the repeat relationship. Marketplace terms and data access vary, but operators consistently describe the same practical limitation: the restaurant often cannot treat a marketplace diner like a direct, permissioned customer. Service recovery, win-back communication, and recognition of repeat behavior remain inside the platform relationship instead of the restaurant’s own systems.
“A ninety-day statement showed $3,266 leaving on $14,200 in marketplace revenue. The number became clear only after every charge landed in one sheet.”
The useful lever was the path, not a new promotion.
Direct-order growth often came from existing demand rather than a new advertising campaign. The direct link moved to the first position on the restaurant website and social profile. A short line appeared on receipts and takeout packaging. Staff received one consistent explanation for guests who asked where to order next time. None of these mechanics created a new diner; they removed friction for a diner already intending to return.
Operators also reported that the message worked best when it stayed factual. The restaurant did not need a permanent discount or a campaign that trained regulars to wait for a coupon. A clear direct route, reliable menu availability, accurate pickup timing, and a reason to use the restaurant’s own channel were more durable than a launch-week promotion.
The second direct order mattered more than the first. Once a diner used the direct channel with appropriate consent, the restaurant could support service communication and future outreach through its own system. That did not guarantee retention, but it created the possibility of a relationship that improved instead of resetting at each transaction.
“A $340 pickup order on a rainy Tuesday would have lost roughly $85 to a marketplace commission. The direct order kept that slow service in the black.”
Operational simplicity decided whether the shift held.
A direct channel that created manual re-entry, a second menu to maintain, or another tablet beside the expo line could return saved commission as labor and errors. Operators repeatedly tied durable channel shifts to the kitchen handoff: one current menu, clear throttling rules, dependable modifier mapping, and an order stream staff could trust during the rush.
This is why the POS question and the direct-ordering question recur together without being the same decision. The POS handles service execution across floor and kitchen. The direct-ordering platform handles the restaurant’s online doorway and customer relationship. Strong implementations make the systems cooperate while preserving separate accountability for each job.
Sequencing mattered as much as selection. Operators who changed the ordering path, menu structure, kitchen routing, and staff procedure in one cutover had trouble identifying the source of a failure. A staged rollout made the handoffs observable: menu and modifiers first, test orders next, limited operating hours after that, then broader promotion once service stayed stable. Margin improvement lasted only when the new route could survive the busiest service without special handling.
“Three tablets looked like three revenue channels until the rush. Then they became one labor problem, one ticket problem, and one very crowded counter.”
Transition economics
The math
An illustrative restaurant produces $40,000 in monthly online revenue. At the starting mix, 15% is direct and 85% runs through marketplaces. At the later mix, 55% is direct and 45% remains marketplace volume. Total online revenue does not change in this model; $16,000 simply moves from one channel to the other.
At a 20% marketplace commission, that shift avoids $3,200 in gross commission each month. At 25%, it avoids $4,000. The annual range is $38,400 to $48,000 before direct-platform fees, payment processing, delivery expense, discounts, refunds, implementation cost, or changes in order volume.
The range matters more than a single headline savings number. A restaurant can run the same model at the effective rate on its own statements, then subtract every direct-channel cost on the same time basis. A monthly platform subscription belongs beside monthly avoided commission; per-order processing and delivery belong beside per-order marketplace charges. Keeping the units aligned prevents a flat fee from looking larger than a percentage charged across thousands of orders.
- Revenue shifted
- $16,000/mo.
- Gross monthly range
- $3,200–$4,000
- Gross annual range
- $38,400–$48,000
What the model leaves out
Gross avoided commission is not net savings. A responsible comparison subtracts payment processing, the direct platform’s fixed or variable fee, delivery dispatch, implementation, menu maintenance, and any demand change caused by moving the ordering path. It also separates marketplace orders that represent genuine discovery from repeat orders that could reasonably move direct.
The model becomes useful when the restaurant replaces assumptions with its own statements. The commission leakage calculator keeps the inputs in the browser, while the marketplace fee table records current public plan terms and source dates. Neither replaces the actual contract or P&L.
Context from operator conversations
The tools that came up
These tools recur because they address separate operating jobs. Their inclusion is editorial; the links below are affiliate redirects disclosed at the point of use.
Featured in operator conversations
Toast
Toast addresses floor and kitchen execution — modifiers, courses, payments, and kitchen communication through service. It belongs in this conversation when the channel shift also needs a dependable path into the operating system.
Learn more →Affiliate partner. Full disclosure at /disclosure.
Featured in operator conversations
ChowNow
ChowNow addresses the direct online channel — a fixed platform cost, direct-order economics, and a restaurant-controlled customer relationship. It recurs when the objective is to give repeat demand a route outside the marketplace.
Learn more →Affiliate partner. Full disclosure at /disclosure.
Source notes
Public documents behind the math
- Commission and platform context: Federal Trade Commission food-delivery platform staff report.
- Restaurant fee example: Federal Trade Commission complaint describing Grubhub restaurant commissions.
- Current marketplace tiers: DoorDash merchant marketplace plans and Uber Eats U.S. merchant pricing.
- Restaurant cost structure: National Restaurant Association restaurant inflation analysis.
- Direct-order processing context: ChowNow card-processing fee documentation.
Source review: July 11, 2026. Public pricing changes; current contracts control.
Verified operators only
Add your field notes.
Membership is free and always will be. Applications are reviewed weekly. Vendors are declined.
Request to join →