Case
CASE.02Route optimisation
A Python route-optimisation model raised route efficiency by 27% and vehicle utilisation by 39% across a national network.
- +27%
Route efficiency (stops per route)
- +39%
Vehicle utilisation
- Context
- Route planning for a national delivery network, with vans, delivery windows and driving-time limits.
- Friction
- Routes were configured by hand, node by node, with no way to test alternatives before changing them.
- Intervention
- A Python model that groups stores into routes and sets the number of routes per node within van, time and window constraints.
- Adoption
- Used to test scenarios before each change, node by node, then built into a standard weekly planning routine.
- Role and scope
- Employee, process engineer for the national network. Built the optimisation model and owned the weekly route planning process.
Deliverables
- Python route-optimisation model
- Scenario testing before each network change
- Standard weekly planning process
- Route KPI dashboards
In short
CASE.02.BRoutes were planned from experience. That worked, but it left capacity on the table, and nobody could test a different configuration without trying it on the road.
The model takes a day’s delivery plan (stores, locations, volumes) and the real constraints (van capacity, delivery windows, hours on the road) and returns the most efficient set of routes and the number of vans needed. AI was used to write and test the code faster; the logic stayed fully owned and understood.
Each node was reconfigured only after its scenarios had been compared. The weekly planning tasks were standardised around the model, with dashboards for stops per route, vehicle and time utilisation, and cost per package, so improvements held after the change.
Evidence
Result achieved as an employee, as stated on Bruno's public CV. Percentages only.
No internal tool names, screenshots, methods or absolute amounts are published.