The plan your best planner would make, before they arrive
Compute-heavy solvers for the decisions that decide your margin: which vehicle, which driver, which order, which route, which carton goes where. Every plan is feasible by construction and explains anything it could not place.
Two half-empty trucks, or one full one
2
Trucks
2
Drivers
43%
Loaded
The day, as it is
- Planners rebuild tomorrow's schedule by hand every evening
- Half-empty trucks leave because consolidation is spotted too late
- Nobody can say whether the plan is good, only that it is done
Agents, engines and your people
Every unit of work passes through the layer best suited to it. Colour shows who is doing the work.
- 01agent
Collect
Orders, fleet, drivers and site windows are gathered from your systems, files and inbox.
- 02engine
Solve
Routing, scheduling, packing and consolidation models search for the best feasible plan.
- 03human
Review
The plan arrives as a proposal with costs, utilisation and reasons for anything unplaced.
- 04agent
Apply
On approval, trips are created and drivers notified, after a final check against live data.
What it does
Vehicle routing
Time windows, capacity in several dimensions, pickup–delivery pairs, breaks and reloads.
Linehaul scheduling
Legs assigned to vehicles and drivers together, respecting hours, docks and cold chain.
Load consolidation
Shipments on similar routes merged onto fewer vehicles, ranked by kilometres saved.
3D load building
Every carton placed and proven to fit before the truck is loaded.
Fleet-mix sizing
The cheapest mix of vehicle types for a set of orders, for planning and for quoting.
Demand-aware planning
Forecasts feed the plan so capacity is ready for what is coming.
The engines doing the maths
Classical optimisation, not guesswork. How it fits together
Vehicle routing
Rich VRP · metaheuristic search (Rust)
Constraint scheduling
OR-Tools CP-SAT · interval model
Fleet-mix optimisation
Integer linear programming · branch-and-bound
3D load packing
Pivot-based 3D packing · knapsack ILP
Route-similarity consolidation
Fréchet-distance matching · parallel (Rust)
Demand forecasting
Gradient-boosted trees · seasonal decomposition
- Why not ask a language model to plan routes?
- A language model can describe a plan but cannot guarantee it is feasible or near-optimal. Fero uses the model to understand your data and a solver to make the decision.
- Does the solver dispatch on its own?
- No. A run produces a proposal. Trips are created only when someone applies it, and the plan is re-validated first.
- What happens to loads it cannot place?
- They come back with the constraint that stopped them, ranked so the most fixable reason is first.
Bring us one lane. We will show you the difference on your own numbers.
A week of your orders, rates or invoices, run through Fero with your contracts and your carriers.