Supply Chain & Logistics
Real-time route optimization & inventory forecasting
Regional 3PL400+ vehicles6 distribution centers
The problem
Routes were planned overnight against demand data that was already stale by dispatch. Trucks left full and returned half-empty; stockouts triggered expedited freight that quietly consumed the quarter's margin. Planners overrode the system so often that nobody trusted its numbers, or knew when it was actually right.
What shipped
- Live demand model fed by streaming vehicle and order telemetry
- Route re-optimization triggered by disruption events, not the clock
- SKU-level demand forecasts wired directly into replenishment
- Planner console with auditable overrides, trust, measured
Problem-to-production blueprint
1DiagnosisMargin-leak decomposition: routing waste vs stockouts vs expedited freight, quantified per lane.
2Architecture & PoCStreaming telemetry plus live demand model, proven on the two worst-performing lanes.
3Production engineeringEvent-triggered re-optimization and forecast-to-replenishment automation across all six DCs.
4StewardshipWeekly forecast-accuracy reviews with planners; override analytics feed the next model iteration.
Outcomes
-18%Cost per route
93%Forecast accuracy, from 71%
-37%Stockout events, YoY
Where the margin leaks
Route and inventory decisions that keep pace with reality.
Streaming data, event-triggered optimization, and forecasts your planners trust.