Manufacturing Demand Forecasting Engine
An industrial manufacturer managing 2,000+ product-customer demand streams in spreadsheets deployed a hierarchical forecasting engine that covers a 16-week horizon, in 12 weeks.
The challenge
Weekly demand for thousands of product-customer combinations was managed in spreadsheets. There was no visibility into a 16-week horizon. Bottom-up and top-down forecasts never reconciled. The planning team spent most of their time on data preparation, not strategic decisions. External factors like seasonality and market indices were not captured.
What we built
We configured our Core Forecasting Engine for hierarchical forecasting at three levels: SKU-customer, SKU, and aggregate. Feature engineering captured seasonality, customer-level variability, and market signals. Automated data pipelines ensured repeatable forecast generation. Performance evaluation used MAPE, RMSE, and weighted accuracy by volume.
Outcomes
Forecasted simultaneously across SKUs and customers
Covering 4 months of production planning
Hierarchical reconciliation across aggregation levels
From kickoff to operational MVP
“We went from updating spreadsheets weekly to having automated forecasts that our buyers actually trust.”