Outlier Adjustment
A method of handling slow-moving or intermittent demand history. With demand being probabilistic, you can set a target confidence level for an upper control limit and adjustment limit. The forecasting process utilizes a probability distribution function and calculates the upper control limit and adjustment amounts. Since slow or intermittent SKUs will never have a negative spike in demand, the forecasting algorithm uses a lower control limit of zero (0).
Outlier calculations are based on raw demand history, while the history average and history standard deviation values displayed on the Demand page are based on history + adjustments. This means that you cannot use these displayed numbers to calculate the Outlier Adjustment values.
Outlier Adjustment Calculation Fields
These are the fields that are used by the Outlier Adjustment calculation.
The following fields, on the Demand Detail page and the page, are used by Outlier Analysis:
Field | Description | Associated Page | ||
|---|---|---|---|---|
Upper Control Limit | The maximum value of demand for a SKU to be considered as not a spike in demand. Any demand greater than this limit is considered a spike and is flagged as a detected outlier. | Forecast/Parameter Metrics container on the Forecast tab of the Interactive Planner Worksheet | ||
Lower Control Limit | The minimum value of demand for a SKU to be considered as not a spike in demand. Any demand lower than this limit is considered a spike and is flagged as a detected outlier. | Forecast/Parameter Metrics container on the Forecast tab of the Interactive Planner Worksheet | ||
Upper Control Limit Confidence Level | The percentage of the time that actual demand is observed to be less than the forecasted demand upper limit. The default is set to 99.73%, with the ability to store up to 3 decimal places.
| Forecast/Parameter Metrics container on the Forecast tab of the Interactive Planner Worksheet Forecasting tab of the Forecast Parameters | ||
Correction Confidence Level | The percentage of the time that the actual demand is observed to be less than the forecasted demand. The default is set to 99.73%, with the ability to store up to 3 decimal places.
| Forecast/Parameter Metrics container on the Forecast tab of the Interactive Planner Worksheet Forecasting tab of the Forecast Parameters | ||
Use Confidence Level for Outlier Adjustments | The method for calculating outlier adjustments. Valid options are: • All — Calculates outlier adjustments using Probability distributions for all pairs. • Slow or Intermittent — Calculates outlier adjustments using Probability distributions for only slow and intermittent pairs. This uses the standard deviation multiplier based calculations for all other pairs. • None — Uses the standard deviation multiplier based outlier calculation for all pairs.
| Forecast tab of the Forecast Parameters |
Outlier Adjustment Rules
Some important things to remember when using Outlier Adjustments:
• Outlier Adjustments are only run on top most parts in a replacement chain. If a part is a replaced part, all Outlier Adjustments are automatically removed.
• Outlier Adjustments can be applied by running a manual AutoPilot job as well as part of a scheduled AutoPilot process.
• The setting of the Std Deviation - Use de-seasonalized and de-trended history field on the Forecasting tab of the Forecast Parameters page determines what is used in the outlier calculation:
◦ When set to Yes, then the de-seasonalized/de-trended history is used in outlier calculations.
◦ When set to No, then the raw demand history is used in outlier calculations.
• Outlier slices are excluded from the history average and history standard deviation calculation when the OUTLIER_USE_NON_OUTLIER_SLICES global setting is set to True.
• Intermittent SKUs are calculated by comparing Calculated (Periods Between Demand) > Periods Between Demand set on the Best Fit tab of the Forecast Parameters page.
 | If the Periods Between Demand in Forecast Parameters is not specified then the PERIODS_BETWEEN_DEMAND global setting value is used. |
• The combination of the OUTLIER_IGNORE_INTERMITTENCY_CHECK global setting and the Use confidence Level for Outlier Adjustments on the Forecasting tab of the Forecast Parameters page determine the calculation method as shown in the following table:
OUTLIER_IGNORE_INTERMITTENCY_CHECK | Use Confidence Level for Outlier Adjustments | Result |
|---|---|---|
True | None | Outliers are calculated for intermittent pairs using older functionality. |
True | All or Slow or Intermittent | Outliers are calculated for intermittent pairs using the Probability Distribution Functions. |
False | None | Outliers are not calculated for intermittent pairs. |
False | All or Slow or Intermittent | Outliers are calculated for intermittent pairs using probability distribution functions. |
• Outlier Adjustments will not adjust any SKUs that have the following conditions:
◦ The annual demand is less than the value set for the OUTLIER_LOWER_TRIP global setting. The default value is 3.
◦ The history slices are less than the value set for the OUTLIER_MIN_HISTORY_SLICES global setting. The default value is 3.
◦ When the OUTLIER_IGNORE_RECENTYEARCOUNT global setting is set to false, Outliers are not adjusted until the first demand record is at least 12 months or 52 weeks prior from the current date. This includes leading 0 demand records, when the Ignore demand before first non-zero slice is set to No on the Forecasting tab of the Forecast Parameters page. In this case, the calculation will not make Outlier Adjustments unless the GS OUTLIER_IGNORE_RECENTYEARCOUNT global setting is set to true.
| | The default value of the GS OUTLIER_IGNORE_RECENTYEARCOUNT global setting is false. |
◦ The Forecast Method of the SKU is set to Leading Indicator or Crostons.
Outlier Adjustment Configuration
Use these recommended steps to configure and get the most benefit out of using Outlier Adjustments.
1. Configure the following global settings:
2. Set the Use confidence Level for Outlier Adjustments field on the Forecasting tab of the Forecast Parameters page to one of the following settings:
◦ All
◦ Slow or Intermittent
3. Set the Periods Between Demand field on the Best Fit tab of the Forecast Parameters page to determine intermittent SKU. If the calculated Periods Between Demand is greater than the value set on the Forecast Parameters page, then the SKU is considered intermittent.