Your history becomes the forecast.
Demand, load, volumes or arrivals. A model is trained on your history and your calendar, and returns a forecast with a band around it, scored against your current method on weeks it never saw.
Seasonality is learned, not assumed
The trend, the weekly and yearly cycles and the holidays are read out of your history, and the components can be read directly: what a Sunday does to load, what the July shutdown does to a month.
The trend, the weekly and yearly cycles and the holidays are read out of your history.
The band is part of the forecast
Every point comes with the range it should fall in, so stock and staffing can be planned on the low and high end, not only the average.
Your calendar is an input
Holidays, promotions, shutdowns and price changes enter the model as drivers, so their effect is part of the forecast.
Any grain, any horizon
Hourly to monthly, one series or thousands, per product, site or line, from the same pipeline.
