Maritime · Planning
AI for demand and planning support
Can AI improve our forecasting?
The challenge
Forecasts are built in spreadsheets, disagree with the ERP, and errors are only visible once stock is wrong.
Prerequisites
- Reconcilable demand history
- One agreed definition of a unit sold
The approach
- Reconcile demand history across systems before modelling anything
- Forecast assistance that flags anomalies for the planner rather than replacing them
- Backtest against the last four quarters before it influences a purchase order
- Measure forecast error, not model sophistication
What makes it work
If the data will not reconcile, we say so and stop — reconciliation becomes the project.
What we would measure
Forecast error
MAPE against the prior method
Stockouts
On A-class items
Working capital
Tied up in buffer stock
Risk position — Limited risk. Purchasing decisions stay with the planner.
Baselines are established in your business during the assessment. We publish no borrowed benchmark numbers.
Assess this for our businessRelated use-cases