Logistics systems

Accuracy is a human problem wearing a technical costume. The database says a pallet is in aisle four because someone scanned it there; if the scan is slow, awkward with gloves on, or requires a device that is charging in the office, it will not happen, and the model diverges from reality. Every design decision in a warehouse system should be measured against whether it makes the scan more likely — because a system nobody scans into is an expensive way of being confidently wrong.

Route optimisation is more constrained than the textbook version. Real routes carry delivery windows, vehicle capacities, driver hours, one-way streets, a customer who can only receive before eleven, and a load that has to come off in a particular order. Optimality is not the goal — a route that is six percent longer and that drivers actually follow beats a theoretically optimal one they override, and the measure that matters is what happened rather than what was planned.

Connectivity cannot be assumed anywhere in this domain. Warehouses have dead zones between racking, delivery vehicles pass through areas with no signal, and a handheld will be used in a basement. Every device application has to work fully offline and reconcile when it reconnects, which means conflict resolution is a design decision made early: two people scanned the same item, the server has since moved it, and something has to decide what is true.