The problem for restaurant-employed delivery staff is nearly the same as the customer-employed delivery staff mentioned above. The driver sits around in the restaurant parking lot twiddling his thumbs and then 10 lunch orders come in over the course of an hour, most of which while the driver’s out delivering the first order. The last order ends up taking 2 hours to get to the customer who is not at all pleased with cold, soggy food long after the lunch break ended.

The food delivery app business works like the insurance business: the aggregate drivers form a risk pool [1] to protect restaurants from the variability of demand. This allows a single restaurant to be able to accept 10 food delivery orders in a matter of minutes just as easily as they would for orders coming in from the tables in their dining room. The app would dispatch up to 10 drivers to handle those orders and even automatically batch them according to proximity of destination.

Of course the app can also handle multiple restaurants in a similar area in the same way so that drivers can be dispatched most efficiently to handle all the demand for an entire city. The more drivers, restaurants, and customers centralize on a single delivery app, the more efficient the system can be (assuming the app developers know how to optimize the transshipment problem [2]).

[1] https://en.wikipedia.org/wiki/Risk_pool

[2] https://en.wikipedia.org/wiki/Transshipment_problem

Well, you can also - shock horror - drive (or bike, if possible) over to the restaurant and pick up your order yourself! That's of course assuming that the restaurant has another means of placing orders than through the delivery apps (e.g. a phone number)...