Delivery subscriptions and memberships: the discount your P&L cannot absorb

1. What does a delivery membership actually measure before you sign it?

It measures one thing: how many NEW orders the program brings versus how many existing orders it discounts, and that subtraction runs on contribution margin per order, not on GMV.

The aggregator will show you volume growth because that is their metric, and yours is a different one. With most US restaurant traffic already off-premise according to the National Restaurant Association (2025), a permanent 10% discount gets multiplied by sixty transactions a year on every loyal customer you already had. The correct calculation separates the installed base from the incremental one: if 70% of membership orders came from customers already buying quarter after quarter, you did not buy demand, you bought a retroactive rebate on your own clientele and financed it out of cash.

2. The commission ceiling the industry already found through legislation

Fifteen percent is the number regulation put on the table when San Francisco capped delivery commissions at 15% (Restaurant Dive, 2020) and New York made it permanent with 15% per delivery plus 5% for other services (Restaurant Business, 2023). Those caps did not come out of an operators' assembly: they came from auditing what load a restaurant P&L can carry without breaking the unit. Outside those jurisdictions nobody hands you the ceiling, so use it as an internal benchmark. Add effective commission plus membership discount and compare the total against 20%: if your dish runs 30% food cost and 4% packaging, every point above that aggregate 20% comes straight out of the margin that pays rent. That is the boundary, and it is arithmetic, not negotiation.

3. Under 500 thousand USD a year: stay out, or enter with a closed catalog

In the small band the recommendation is DO NOT join the aggregator's subscription program while average contribution margin per delivery order sits below 38%. An operator this size moves between 15 and 40 orders a day through the channel, and lacks the volume to amortize a permanent discount with added frequency; you simply sell the same thing cheaper. If you still want presence, expose a closed catalog of six to eight dishes with food cost between 28% and 30%, well under the 32% maximum set by the Masterestaurant costing rule, and keep anything above 32% out of the program. With nearly 75% of restaurant traffic already off-premise according to the National Restaurant Association (2025), being absent is no option either; being exposed without a filter is an expensive mistake.

4. From 500 thousand to 1 million: the band where GMV rises and EBITDA falls

This is where the trap detonates. An operator this size showed me a dashboard last year with delivery GMV growing 31% while EBITDA fell, and the cause was precisely the aggregator membership: free shipping and a permanent 10% had captured his best customers, the ones already ordering four times a month. Your decision threshold is this: join the program only if you can prove, using the aggregator's own data, that a meaningful share of membership orders come from accounts with no purchases in the prior 90 days. If they will not hand you that number broken out, the answer is no, because you would be funding it blind. Diego F. Parra frames this discussion with the Masterestaurant methodology: unit economics of the order first, scale afterward.

5. Above 1 million: negotiate your own program before someone else's

Past a million in annual revenue you already own a database, and with a database the membership must be YOURS. The call: launch a proprietary program on your direct channel before expanding the aggregator's, with a simple threshold of 1,200 identified repeat customers and an average frequency of 1.8 orders per month. DoorDash generated nearly 60 billion dollars in sales for local merchants in 2024 according to its annual report, and that volume proves the channel works; what it does not prove is who keeps the relationship. A 9-dollar monthly plan with waived delivery on orders above 25 dollars, billed by you, turns the discount into deferred revenue and returns the customer data. Keep the aggregator as pure acquisition, never as retention.

6. Above 5 million: the media-chef operator and the large-format themed house

Two profiles play differently in this band: the media-chef restaurant with a strong personal brand and the large-format themed venue with 300 seats and tourist flow. Neither should enter a percentage-discount membership, because their average check —between 45 and 80 dollars— turns every point into real money, and their demand is not elastic to price but to reservations. For them the right membership is one of ACCESS: table priority, closed tasting menu, seasonal presale. The delivery channel of these houses should run under a satellite brand with a separate catalog, leaning on ghost kitchen infrastructure, which surpassed 20,000 operating locations in the United States in 2023 according to Statista. Blending the flagship brand with a permanent discount destroys pricing power in the dining room.

7. Above 10 million (group or chain): the membership is a data asset

At group scale, subscription stops being a pricing tactic and becomes infrastructure, with one clear operating threshold: launch nothing below 25,000 active members projected over twelve months, because under that figure the platform and CRM cost never amortizes. The scale reference exists and it is brutal, so the question is not whether the model works, but whether you control the data. A group with eight or more units should negotiate tiered commission with the aggregator in exchange for guaranteed volume, and run its own membership in parallel. Two channels, two economics, one identified customer.

8. What happens if the discount fails to generate incremental frequency

Suppose you enroll the full catalog, give away a permanent 10%, and average frequency moves from 2.1 to 2.3 orders per customer per month. That 9.5% lift in orders looks like a win until you cross it against the discount applied to the entire base: if 100% of orders lose 10 points of price to gain 9.5% of volume, and your contribution margin was 40%, absolute margin drops around 12%. Now add that 58% of limited-service operators already carry a larger off-premises sales share than in 2019 according to the National Restaurant Association (2025), which means that degraded channel will be an ever larger share of your business. A discount without incremental frequency is not a weak campaign: it is a structural price cut dressed up as a loyalty program.

9. What separates a profitable program from one that bleeds?

Where the order comes from. A membership that brings guests who never bought from you pays for its own discount; one that merely cheapens your loyal customer transfers your contribution margin straight into the diner's pocket, and the aggregator contributes nothing.

The catalogue you expose. The commission cap. When San Francisco limited delivery commissions to 15% (Restaurant Dive, 2020) and New York made 15% per delivery plus 5% for other services permanent (Restaurant Business, 2023), the industry learned which number an operator can actually carry. Outside those cities, you negotiate for it. Measurement. Without cohorts there is no evidence, and without evidence the board decides on narrative. The nearly five monthly orders DoorDash reports (Restaurant Business, 2024) is a market average, not your baseline; yours lives in your own data. Dependency. A virtual restaurant or dark kitchen living off one app's algorithm carries concentrated territory risk, and the owned channel, the Google Business Profile listing and geotargeted advertising are the risk mitigation no membership plan will hand you.