The measurement of forecast accuracy normally compare actual data with corresponding forecast data at a specified lag, typically 2 months. For example, the demand forecast for March 2012 made in January 2012 would be compared with the actual data for March 2012. See Measurement of Forecast Accuracy for details for reporting options.
Forecast Reference Files
The above type of forecast accuracy analysis requires the maintenance of a forecast reference file for each year. This file would contain data as follows:
|January||November from prior year|
|February||December from prior year|
|March||January from current year|
|April||February from current year|
A suitable specification for such reference files in IFP would be as below:
Copying Data into Forecast Reference File
The forecast for each month must be copied into the forecast reference file from the relevant forecast file, e.g. the 2011 March forecast made in January 2011 should be copied into March in the forecast reference file.
Note that for January the November forecast file would be used (this January 2011 forecast would have been created in November 2010.
Using Variable Lags
It is a simple matter to create analyses using lags that vary from one product to another, e.g. if some products have long delivery lead times it may be necessary to use a lag of 3 or 4 months instead of the normal 2.
This is achieved by simply adjusting the forecast used for these exceptional products, e.g. for a lag of 3 months the forecast created in January would be used as the forecast reference for April.
It is essential to lock the forecast reference file to the relevant month after copying data for each month. For example, after copying March data the forecast reference file must be locked to March.