Why off-the-shelf routing was not enough
Standard routing engines optimize for the average trip. Our use case needed temperature windows, service levels, and a much tighter operational model.
That pushed us toward a custom engine that could understand the business rules before choosing a path.
The constraints we had to encode
The router had to care about more than geometry. It needed to factor in delivery windows, traffic volatility, and the operational penalty of a late handoff.
That meant modeling the route as a business decision, not just a map lookup.
- Time windows for sensitive deliveries.
- Live traffic weighting.
- Fallback behavior when conditions change mid-route.
What the engine optimizes
The router balances travel time, live traffic, and delivery constraints in real time. The goal is not only to move quickly, but to move correctly.
That distinction matters in healthcare logistics, where the wrong route can invalidate the delivery promise entirely.
What we learned
The main lesson was that system design and product design are inseparable. The best technical decisions also had the clearest user impact.
Once the engine became understandable to the team, it became easier to operate and easier to explain to customers.
Once the router understands the constraints, the rest of the product can move with much more confidence.