1) Your domain has a dns op (order placement) record like an mx record. Your op record points to your online pos system. A standard chunk of data is sent by the customer browser and received by your system.
2) You pay a company for marketing+online menu services. You optionally grant them the right to list/crosspost the same data on competing services. You have the choice which menu-fronts are allowed to use your name and advertise your product.
3) The order comes into your system like any other order. This triggers a separate signal to a p2p dispatch network. The dispatch network load balances all the incoming orders, assigning them to drivers in a logical way that hits restaurants and destinations with minimal extra travel.
4) The delivery driver chooses any client they want, that is compatible with the dispatch network. Tasks are routed to their client.
The order placement, pos, and the dispatch network are totally separate technologies, not proprietary vertically integrated institutions.
Without having the first-hand knowledge of parent comment, my assumption is restaurateurs want more control of both the marketing and the delivery process. I don't think there is enough $ in the pie to pay a third party for marketing and a 3rd party for delivery. I believe the desire is to go back to something they can more directly manage (traditional delivery worked for them), but have an online ordering/POS tool that is easy to use/find from a google search. The consumer does not care if the order is delivered by an employee or someone on a network.
There's no reason they have to use a third party marketing company, if they want to build their own menu and run their own ads.
I wrote a complete system like a Shopify for small retailers integrated with delivery etc. But the code is locked in a private repo for me to get back to if my current startup fails.
You do understand that the problem with food delivery isn't the online bit. It's fairly easy to manage the online bit.
Expectation management is a big issue in food delivery business - that turns people away real easy. Miss delivery timing twice and people are much more likely to drop you. Get swamped with calls - people drop you. It's a hellhole of customer service.
Building out a reliable delivery network - is not easy. The farther the delivery goes out - the more complex it gets.
Out of all - Uber has the best solution to evaluate speed of delivery using a car... and OK for bicycles.
The economics are impossible to sustain, and I've been on the ground first hand talked to over 1000 owners at restaurants, grocers etc. Every single one of them has a timeline to get rid of Eats, Doordash, Instacart etc. They're just waiting for the right time to build out their own systems.
There was an enormous opportunity to help them build their systems out that I might pursue if my current venture fails.