Windows, sequences, and the system that runs backwards

Your delivery slot is decided long before anyone gets into a vehicle.

A delivery window feels like a statement about a driver's afternoon. It is really a statement about the night before. Overnight, parcels are sorted into runs, and each run is a sequence of stops built from the addresses that happen to be in the system at that moment. Once your parcel is placed in a run, your window is essentially fixed, because it is derived from your position in a sequence that has already been optimised. Very little that happens during the day can move you earlier; things that go wrong can only move you later.

This is why a narrow window is expensive and a broad one is cheap. A broad window lets the planner sequence stops in the order that minimises total travel, which is the whole source of the saving. A narrow window is a constraint on that sequence, and every constraint added to a route makes the rest of the route worse for everyone else on it. When you pay extra for a tight window, you are not paying for urgency. You are paying for the inefficiency your constraint imposes on the run.

The same logic explains why the return journey is a whole second system rather than a rewind of the first. A returned item begins as one object at one door, which is the most expensive possible starting condition, and there is no sequence waiting to receive it. It must be collected, identified, inspected, graded, restocked or written off, and only then rejoined to ordinary freight. Nothing in the outbound network is shaped for this direction: vehicles fill up rather than empty out, sorting happens by condition rather than by destination, and someone has to look at the item and decide what it is now worth. That reverse flow is one of the largest costs hidden inside an ordinary delivery.