Overview
Outlier detection is an automatic way to improve the forecast. The process scans, identifies outliers in the sales history and corrects the data point when calculating the forecast.
The Outlier sales period is in line with the forecast type i.e. A monthly forecast type will scan for monthly outliers and a weekly type forecast will search for an outlier weekly outliers.
Outlier Detection
An outlier is a data point that is an unusually large or small data point and falls outside of the expected range and can have a significant impact on the forecast.
The process of screening the historical data for outliers and replace them with more typical values prior to generating forecasts is referred to as outlier detection and correction.
Outlier correction can often improve the forecast. However, if the outlier is not truly severe, correcting for it may do more harm than good. When outliers are corrected, the history is rewritten to be smoother than it was which will change the forecasts and narrow the confidence limits. This can result in poor forecasts and unrealistic confidence limits when outlier detection is not necessary.
Outlier detection should be performed sparingly, and it should be reviewed to determine whether a correction is appropriate.
An automated algorithm detects and corrects outliers as follows:
Each time a forecast is calculated, the fitted errors are generated and their standard deviation is calculated.
If the size of the largest error exceeds the outlier threshold, the point is flagged as an outlier and the historic value for the period is replaced with the fitted value.
The procedure is repeated using the corrected history until either no outliers are detected or the specified maximum number of iterations is reached.
Outlier Deviation Threshold
The sensitivity of the outlier detection algorithm is defined by the standard deviation threshold that is set to four.
This means that if a given fitted error exceeds this defined threshold and it is the largest error detected during the current iteration, it will be flagged as an outlier.
Maximum number of iterations permitted during outlier detection are three. Since only one outlier is identified per iteration, this defines the maximum number of outliers that can be detected for a given item.
Where can I enable the Forecast Outlier Detection?
To Enable the Outlier detection navigate to Settings > Forecast > General > Forecast outlier detection = Find and correct > consider the "Outlier deviation threshold" set to 4 standard deviations as default.
Searching for settings can also be done via the Settings search window
Finding items that have been corrected by Outlier detection
In items and reports you can add a filter "Outlier detection" and select "Yes" then you filter on all items/articles that have been impacted by the "find and correct" outlier detection. you can also add a column in items and reports called "outlier detection" to have a yes/ no indication in line when reviewing or analyzing Items / articles
Outlier Detection is available in the Planning Pro Package subscription