Demand Forecasting Model

The Forecasting model can be accessed only if the the option is checked on the Administration page, Folder Parameters window and if Advanced Level is selected.

The Forecasting module automatically calculates sales forecasts for the future periods in the planning process. This module uses classical extrapolation methods, in particular exponential smoothing models. These forecasts can be used:

•      either to establish the Sales and Operations Plans. In this case, the forecasting module generates monthly forecasts, which can be transferred into the Sales and Operations Plans window. The forecasting module is accessed by the Forecasting button in the Sales and Operations Plans window:

•      or to estimate the short term requirements. In this case, the forecasting module will generate weekly forecasts which can be transferred into the Sales Forecast window. In this case, the forecasting module is accessed via the Forecasts button, either on the item Sales Forecast Maintenance window, or on the Sales Forecast Table window.

 If the distribution process (to the different distribution centers) is taken into consideration (via the DRP procedure), the short-term forecasts are entered for each distribution center (in addition to the direct forecasts for the plant).

The implementation of extrapolation methods requires data concerning past sales: in other words, it is necessary to have historical data of sales. These time series have to be entered.

Then it is necessary to choose the numerical parameters of the forecasting model, clearly in order to minimize the forecast errors. The corresponding forecasting model is then applied to the historical data describing past sales. Three forecast error indicators can be calculated for a forecast procedure:

•      The mean algebraical error, i.e. the average error between the computed forecasts and the corresponding observed sales (over a given horizon).

•      The mean absolute error, i.e. the average absolute error between the computed forecasts and the corresponding observed sales (over a given horizon).

•      The mean quadratic error, i.e. the average quadratic error between the computed forecasts and the corresponding observed sales (over a given horizon).

When these error indicators are low, the forecasting model is an efficient model. It is thus necessary to find the best model and the best parameters in order to guarantee low error indicators. Once a satisfactory model has been designed, the forecasts can be generated and transferred into the associate windows (the Sales and Operations Plans window and the Sales Forecast Maintenance window).