Logistics software fails on the exceptions — the split shipment, the failed delivery, the customer who changes the address while the van is already moving. The happy path is the easy ten percent.
Every logistics system models the straightforward shipment well. The ones that survive contact with an operation are the ones whose data model could represent what actually happened.
A data model that can represent partial shipments, multi-leg journeys, consolidations, and returns — not just one order becoming one delivery.
Planning that accounts for vehicle capacity, driver hours, time windows, and the reality that the plan changes twice before lunch.
Events captured where they happen — scan, geofence, or driver action — so status reflects the world rather than someone's memory of it.
Stock by location and condition, with cycle counting and the reconciliation that keeps the system and the shelf in agreement.
An integration layer new carriers plug into, covering modern APIs and the EDI documents much of the industry still runs on.
Failed deliveries, damaged goods, and address changes routed to someone with ownership and a clock, rather than to a whiteboard.
Ask an operations team what breaks and you'll get these four. Ask the software what it can represent and you'll often find it can't represent any of them.
Most of this work is systems agreeing with each other — see data integration and business process automation. The carrier and retailer side of it is EDI integration.
Describe the operation and where visibility breaks down today. We'll come back with what to build first and what it changes for the people doing the work.