Background
A 300M+ RMB/year packaging & printing manufacturer with a complex mix of SKUs, frequent rush orders, and two senior planners carrying the entire schedule in their head.
The Pain
- **Schedule errors** averaging 30-40% when planners were absent.
- **Cost estimates** based on rough rules of thumb, leading to either lost bids or margin erosion.
- **Quality drift** caught too late, after hundreds of units had been produced.
What We Built
A private labor-time prediction model trained on 14 months of historical work orders. Delivered as a FastAPI endpoint integrated into the planning tool. Models retrained monthly on the latest 3 months of data.
Outcomes
|--------|--------|-------|
| Daily planning time | 4 hours | 2.5 hours |
|---|---|---|
| On-time delivery | 78% | 94% |
| Incoming defect rate | baseline | -42% |
The factory reduced its dependency on two key individuals and lifted OEE by reducing changeover wait time.