Snapshot
Who: Regional LTL network (anonymized) — 340 power units, 19 terminals.
Problem: Static routes aged by lunch; dispatchers patched with phone calls; fuel spiked whenever a bridge closed.
What we shipped: Continuous VRP-style re-optimization windows, driver tablet deltas every 15 minutes, and exception playbooks when docks slipped.
Challenge
Drivers distrusted “black box” routing after a failed pilot three years earlier. Transparency and rollback were non-negotiable.
Approach
We shadowed dispatch for six weeks, mirrored their heuristics in software, then beat them on held-out Fridays before any cutover.
Solution components
Rollout timeline
Month 1 · Digital twin
Replayed 11 months of TMS history; found 7 systematic detour patterns.
Month 2 · Shadow Friday
Human dispatch vs. model side-by-side scorecards.
Months 3–4 · Live in two districts
Rollback switch always one tap away for terminal managers.
Month 6+ · Network
Dock APIs added for top 40 shippers by volume.
“Drivers stopped arguing with dispatch when the map showed the same math they were already doing in their heads—just faster.”
— VP Network Operations, anonymized carrier
Results & controls
- Fuel + idle savings annualized to $890K in the first full quarter post cutover.
- OTD +19 pts on lanes with dock integrations vs. control lanes.
- Customer chargebacks for late delivery −31% by dollar value.
- Every re-seq stores constraint snapshot for disputes and insurance.
Representative scenario for marketing purposes. Savings depend on network topology, union rules, and equipment mix.