https://seattle.eater.com/2024/2/21/24079162/tony-delivers-seattle-delivery-app-fees-downtown
Tony Illes was working as an Uber Eats delivery person when an ordinance passed last year by the Seattle City Council came into effect in mid-January. The new rule required app companies to pay workers like Illes a minimum wage based on the miles they travel and the minutes they spend on the job. The apps say that this amounts to around $26 an hour, and both Uber Eats and DoorDash responded by adding $5 fees to every order (even when the customer is outside Seattle city limits) while calling for the law to be repealed. According to a recent DoorDash blog post, the ordinance has resulted in an “unprecedented drop in order volume,” a drop that Illes felt personally. He told Geekwire that “demand is dead” and told local TV station KIRO 7, “I didn’t get an order for like six hours and I was done.”
So Illes had an idea: Who needs these apps, anyway? He printed up signs with QR codes directing people to a bare-bones website with his phone number, promising that he would deliver food by bike in Uptown, South Lake Union, Belltown, and a chunk of the downtown core for $5 a pop from 11 a.m. to 2 p.m. and 5 p.m. to 8 p.m. daily. All you had to do was order the food and send him the screenshot. He called himself “Tony Delivers.”
Allocating a job to a driver is the easy part. It's all the other stuff people expect from a delivery app that's the hard part. Like having an accurate DB of stores and facilitating orders/payments. If you don't do that then people can troll with fake orders and stiff drivers. Plus moderation of drivers who steal food or are convicted burglars/rapists (existing apps already suck at that).
But a federated approach would be immensely more complicated to do well and is a privacy nightmare. You'd need to share buyer's address and drivers' current locations to many different instances to facilitate a buyer on one instance and potential drivers on several different instances. All that data needs to be available (and accurate to the minute) to the instance that assigns the job. Similar privacy/logistic issues pop up when you consider payments.
I think there's also a fair amount of optimization work on the pricing depending not only on supply/demand, and geography but also other environmental conditions, and a notable amount of data to collect in quasi real time to reach a good level of service. Uber-like services are ethically very questionable, but there's a lot of fine tech behind, their tech blogs are often very rich.