Use backtesting to estimate the performance of your forecast model. 

Backtesting logic

When you create a forecast model, it holds back part of the historical data (actuals). The data withheld is equal to the length of the forecast horizon. The trained forecast model automatically produces predictions for the period represented by the withheld data. Backtesting supports the comparison of forecast against actuals. The accuracy metrics are derived from this comparison.

Example:

You have 12 months worth of historical data and a 3-month forecast horizon. 

  1. The model training process uses the first 9 months of historical data and predicts the last 3 months. The remaining 3 months of historical data are held back from the model training process.
  2. PlanIQ compares the forecast of the last three months with the actuals. This comparison results in advanced accuracy metrics that estimate the performance of the forecast model.
  3. You can import the withheld data to estimate performance at the item level.

The closer the forecasted values are to the actuals, the more accurate the model's predictions are, and the better the estimated performance.