Demand forecasting
Item and location forecasts that drive buying, stock and labour plans.
The problem
Forecasts are built from last year plus a guess, leading to stockouts and overstock.
Data we need
- Sales or shipment history
- Promotions and prices
- Calendar, weather and events
How it works
- Assemble history at item and location level
- Build forecasting models with explainable drivers
- Backtest against recent seasons
- Feed forecasts into replenishment
- Review exceptions with planners
What it changes
Targets are agreed with you before we build, and measured on your data before release.
- Fewer stockouts and markdowns
- Labour planned to expected volume
- Planners focus on exceptions
Time series and forecasting
Demand, load and failure forecasting from sensors, meters and transaction history.
- Temporal CNNs
- Prophet-style models
- Anomaly detection