The problem
Cobalt Freight's planning team discovered shipment delays at the same time as their customers — way too late. Dispatchers were constantly rushing to handle complaints instead of proactively tackling potential problems. Unfortunately, there wasn't a reliable system that combined weather updates, route history, and carrier performance into one easy-to-understand signal.
What we built
We created a custom model using Cobalt's past routing data, carrier info, and weather trends, and we plugged it right into their dispatch system—no extra tools needed for planners! This cool model spots shipments that might get delayed early on, so dispatchers can reroute them or give customers a heads-up, making everything run smoother and keeping customers happy.
How we approached it
Discovery (Week 1–2): Audited Cobalt's existing data — routing logs, carrier records, and weather feeds — and identified what was usable versus what needed thorough cleanup.
Build (Week 3–5): Trained and validated the model against two years of historical shipments, tuning it specifically for Cobalt's lanes rather than using a generic logistics model.
Deploy (Week 6): Integrated directly into their dispatch dashboard, with a simple flag and confidence score dispatchers could act on immediately — no retraining required on their end.
The result
Now, planners receive a helpful heads-up instead of dealing with frustrating customer complaints. The model works seamlessly in the background of their current workflow—no new software to learn and no extra logins needed.
