The qualitative review of a statistical prognosis is a primary activity in the forecast process. The ideal would be that each statistical prognosis was revised to obtain the best possible result. However, in many cases this type of review are not possible due to the volume and detail of the information. Ebays opinions are not widely known. In these cases, tools to effectively assist sift predictions and focus analysis on those articles where human attention is more necessary that should be developed. That is an exception report? An exception report compares a value A to value B and list all items whose difference between A and B exceeds a defined threshold. The top example shows an exception report which lists all items where the prognosis for next month (A) has changed by more than 20% compared to the forecast generated a month ago for the same period (B). The report lists six cases where the 20% threshold was exceeded and an exception was triggered.
The main objective of an exception report is to assist the monitoring process to be more effective and efficient. In the above example the report allows to immediately identify those articles where forecasts have changed considerably so that human intervention is necessary. This monitoring helps to keep under control the forecasting process, allowing to identify potential problems before issuing the forecast firm which is a very common this type of reports use. Usually the forecasts are reviewed against forecasts generated with different times in advance or against historical values (for example, against the month immediately previous or annualized is saying the same month a year ago). Another use of exception reports is controlling generated forecasts against what you truly step. This allows to identify articles which this variation above the allowable range which can be an indication of a Article out of control. Some forecasters analyze statistics of the sample as MAPE (mean absolute of the percentage of Error) and the MAD (absolute deviation Media).