Count what your dispatcher actually does on a Friday night. An order lands every couple of minutes, and for each one someone has to decide which courier takes it, in what order, and when they leave. Over a shift that adds up to 150+ decisions, most of them made from memory, on a phone, in between arguments about who has to drive to the far zone. A dispatcher who's good at this costs $2,500–3,500 a month. A bad one costs more than that, just in cold food and couriers standing around waiting.
Courier dispatch software is supposed to make those decisions for you. There's a real gap, though, between software that shows orders on a map for a human to sort out and software that assigns them itself. The first kind renames the dispatcher's job. The second kind removes it. This piece is about the second kind and how it works.
Assignment is a scoring problem
The obvious way to build dispatch is a queue: the next order goes to whoever has been idle longest. It feels fair, and it breaks on the very first order, because the courier who's waited longest might be sitting on the wrong side of your delivery area.
A real assignment engine does something else. Several times a minute it scores every free courier against every order that's ready to go. What it weighs:
- Where each courier is, and just as much, where their current drop-off will leave them. A courier finishing a delivery three blocks from the next pickup beats an idle one parked at the restaurant.
- Kitchen timing. Send a courier to an order that still needs 12 minutes on the grill and you're paying someone to lean on a counter. The engine reads kitchen status so couriers show up when the food does.
- Whether two orders can ride together. If they're headed to the same neighborhood a few minutes apart, they should. Software catches those pairs every time; a tired dispatcher catches some of them.
- How loaded each courier is, and what they're driving. The bike courier shouldn't get the 6-mile run, and a courier already holding two orders shouldn't get a third with a tight promise time.
That's the logic running inside our logistics management module: an order assigns itself the moment the kitchen confirms it, and the route re-plans if something changes on the way. The dispatcher's 150 decisions turn into software output, and a person only looks when something odd comes up.
What operators actually save
Three numbers move once assignment goes automatic. Courier idle time drops first: operators who switch tell us their couriers spend a lot more of each shift actually carrying food, so the same fleet handles 15–25% more orders at peak. The wait between an order being ready and a courier claiming it shrinks to almost nothing, because the engine picked someone before the bag was even sealed. And payroll comes down, since a mid-size operation can run its whole dinner rush with nobody sitting in the dispatch seat.
That last saving is where our AI assistant earns its keep. Delly, the delivery assistant built into Dots, watches the whole flow — assignments, delays, courier status — and handles the exceptions that used to need a human: reassigning an order when a courier's phone dies, flagging an address that geocoded to the wrong spot, nudging the kitchen when prep time slips. For an operation that used to pay someone to sit in that seat, that's $3,000+ a month back in the budget.
Bring a typical Friday's numbers. We'll show you what the engine would have done with them.
The map becomes a supervision tool
Automating assignment doesn't take the map away; it changes what you're looking for on it. Instead of deciding every order, a shift manager watches a monitoring panel: which orders are trending late, which courier is stuck, where demand is clustering tonight. You step in on the 5% of orders that need a judgment call and leave the other 95% alone.
Where humans still beat the algorithm
There are three spots where I wouldn't trust the software on its own. A brand-new courier who doesn't know the area yet will do worse than their score suggests, so a manager should ride along on their first few shifts. A big catering order with a strict serve time deserves a courier you picked by hand, not one picked by points. And when a snowstorm cuts everyone's speed in half, a person needs to widen the promise times before the engine starts making commitments it can't keep. Good software makes an override one tap: you keep control of the policy, and it keeps doing the arithmetic.
What to check before you buy
If you're shopping for courier dispatch software for a restaurant, a few things separate the real systems from the map toys. Does assignment sync with kitchen prep status, or does it treat every order as ready the second it's placed? Can it stack orders on rules you set yourself? Does it push live courier tracking to the customer, since the "where's my food" calls dry up when it does? And can orders flow in from your own app and website without anyone retyping them? That last question is why we built dispatch into one platform instead of bolting it on: the order, the kitchen screen, and the courier app share a single state, which is how 3M+ orders have moved through the system without a dispatch army behind them.
A cheap way to size this up before you spend anything: for one week, log every dispatch decision your team makes by hand, and every minute an order sits ready with no courier on it. Multiply those minutes by your average order value and your reorder rate. That number is what manual dispatch already costs you. You may as well spend it on software.
Auto-assignment, courier app, live tracking, and Delly watching the exceptions — launched in about two weeks.