How we work
The same method runs through every engagement. It is deliberately conservative: it favours results that hold up under challenge over results that look impressive.
Build a real road network
We construct a routable model of the actual road network from open mapping data, including turn restrictions, toll costs and measured traffic speeds. Straight-line distance is fast to compute and wrong in ways that compound: it understates rural travel time, ignores the river with no bridge, and produces a plan that does not survive contact with a delivery schedule.
Model demand and cost by activity
Storage, transport and overnight costs are separated rather than blended, because they respond to different decisions. This is what makes it possible to see that a network is spending most of its money standing still, which a single cost-per-kilometre figure conceals entirely.
Solve, rather than estimate
Network configuration and vehicle routing are solved with a constraint solver against explicit, stated constraints — vehicle capacities, driver hours, delivery windows, licensing rules. A solved answer can be interrogated: you can ask why a facility is served from one depot rather than another and get a reason, which matters when a recommendation is challenged.
Compare scenarios, not a single answer
We evaluate a structured grid of scenarios so the client sees the shape of the trade-off rather than one recommended point. Knowing what the second-best option costs, and where the curve flattens, is usually more useful for a decision than the optimum alone.
Hand over the tooling
Where the engagement warrants it we deliver the model itself, as an application the client can run. A study describes the network as it was on the day it was written; a tool lets the client re-answer the question as demand shifts and facilities open.
What it is built on
Network modelling runs on PostGIS and pgRouting over open mapping data. Vehicle routing and network design are solved with a constraint solver. Analysis pipelines are reproducible and the figures in every report are pinned to their source data by automated tests, so a number in a board pack can be traced back to the calculation that produced it.