This chapter provides an overview of PeopleSoft Demand Planning control groups and discusses how to create and maintain control groups.
A control group determines the major forecast process options, such as which forecast models to use, how to process outlier deviations, and what accuracy statistic to use for evaluating forecasts. The system creates a forecast for each item in a forecast view based on the parameters that are defined in the control group. You associate a view with a control group.
The control group manages the forecast development and tracking for each item in the group. You use control groups to define the level of control and the processing that you want to apply to specific groups of items.
Control groups manage these types of controls:
Item reasonable growth factors.
These factors specify the minimum and maximum allowable projected annual percentage increase for a forecast item. Reasonableness checks trend and projected annual growth to make sure that forecasts are realistic.
Tracking factors that might indicate when a forecast may not be valid.
Trend elimination techniques that help you control long-range forecasts that experience period-to-period quantity increases or decreases over a period of time.
Reset controls for items.
When an item is outside the parameters that you define, the system automatically resets the item and creates a new forecast during the next forecast calculation.
Best-fit analysis in which you select the forecasting techniques to use in determining the model that works the best for the items in the group.
You select the seasonal model and the accuracy statistic to apply to the model.
Outlier processing that defines how the system processes data points that are significantly different from other data in the forecast.
Simulation models that you can use to determine the best fit for a forecast.
Note. Use PeopleSoft Demand Planning control groups to manage forecast items while you use PeopleSoft Inventory Policy Planning control groups to manage policy items. The two sets of controls are not related.
To link control groups and forecast views:
Create forecast views and their levels.
Create a demand planning control group.
Modify the template item for each level of the view and assign a demand planning control group.
The system assigns items to the default control group that is associated with the relevant template item when it creates the items. You can reassign items to other control groups either through manual maintenance, mass maintenance, or item simulation.
Note. An item can belong to only one control group at a time.
Analyze initial item control group parameters to identify exceptions and subgroups of items. One way to group items is by an ABC classification. Items with an A classification can have control group parameter settings that control them more tightly and demand a higher level of user involvement in all phases of the forecasting process. Items with B and C classifications can have settings that make it possible for you to apply less control and enable the system to handle more situations automatically.
Another way to group items is by level. Items at the lowest level of the view can have so much variability that the process generates excessive Work Queue alerts that might not be needed for review.
The issues that you must consider in determining the number of control groups include:
Items with extreme seasonality.
These are items that are sold for only a few of the same periods each year. You might need to adjust the suspend periods value.
Items with known model types.
This prevents items from using models that might yield a slightly lower statistical error but result in an unpredictable forecast pattern. Examples of model types include seasonal versus nonseasonal models.
Items with high relative errors (error ratios).
This ratio represents the maximum acceptable value of the ratio of the deviation and the volume of demand for the item (forecast deviation/demand average). When the item exceeds this value, the system resets forecast model parameters automatically. The higher the value, the less likely it is that the system will reset the parameters. Normally, an error ratio of 1.00 or lower is appropriate for most items.
Items that are in a state of rapid change.
For example, new and dying items might require a forecast analyst to be more reactive than usual.
Items at different levels of the view.
Items that are controlled by different forecast analysts.
Items with different Work Queue reporting requirements.
Note. You should get some experience with the system before you decide on a final classification. In a relatively straightforward environment, one control group per view is a good starting point.
To define control groups, use the Control Group component.
You must have at least one control group before you can add forecast items to a forecast view. You associate an item with its control group by using the Control Groups page when you maintain forecast items.
You can copy an existing control group to another control group with a new name and change the new control group parameters. You can then associate the new control group with a forecast item. You can change the details for an existing control group at any time. When you make changes, the new parameters are used for the associated items the next time the system calculates the forecast or performs a simulation.
You can delete only those control groups that are not being used in the system, meaning that you cannot associate a control group with a forecast item and then delete the control group.
This section discusses how to:
Create control groups.
Define calculation parameters.
Set up simulation model controls and periods.
Define forecast tracking parameters.
Control forecast item reset data.
Manage control group messages.
View items that are assigned to a control group.
See Also
Setting Up Forecast Item Model Controls
Page Name |
Object Name |
Navigation |
Usage |
DP_CONTROLGRPS_TBL |
Demand Planning, Define Forecast Elements, Control Groups, Define, Define Control Groups |
Create control groups. |
|
DP_CONTROLGRP_3 |
Demand Planning, Define Forecast Elements, Control Groups, Define, Calculation Parameters |
Define calculation parameters for a control group. |
|
DP_CONTROLGRP_4 |
Demand Planning, Define Forecast Elements, Control Groups, Define, Model Controls |
Set up simulation model controls and periods that you want the system to consider when it checks for the best-fit forecast model. |
|
DP_CONTROLGRP_1 |
Demand Planning, Define Forecast Elements, Control Groups, Define, Tracking |
Define forecast tracking parameters that determine the way in which the forecast is developed and tracked. |
|
DP_CONTROLGRP_2 |
Demand Planning, Define Forecast Elements, Control Groups, Define, Reset |
Control forecast item reset data. |
|
DP_CONTROLGRP_5 |
Demand Planning, Define Forecast Elements, Control Groups, Define, Work Queue |
Manage control group messages. |
|
DP_CTLGRPFCITEMS |
Demand Planning, Define Forecast Elements, Control Groups, Where Used |
View items that are assigned to a control group. |
|
DP_COPYCTRLGRPS |
Demand Planning, Define Forecast Elements, Control Groups, Copy |
Copy the attributes from one control group to create a new group. Define the control group ID and its description when you copy the group. The system copies control parameters from the existing control group to the new one. You must assign items to the new control group when you maintain them through item maintenance or item simulation. |
|
DP_DELCTRLGRPS |
Demand Planning, Process Deletions, Control Groups |
Remove a control group. If current forecast items or seasonality profiles use information that is contained in a control group, the system prevents the deletion and displays a message. Note. The system displays information that is associated with the group to help you understand the contents of the group. The deletion is permanent. |
Access the Define Control Groups page.
Use the Define Control Groups page to create a control and enter its descriptions.
Access the Calculation Parameters page.
Best-fit optimization automatically checks through various forecasting models to find the one that best fits the specific data set.
Outliers are data values that are exceptional values within a dataset. These values, such as a sudden spike or drop in demand, do not follow the trend of the remaining data. To process outliers, the system uses the standard deviation limit and forecast model that you enter to determine the statistical method that produces the least error.
Model Defaults
Define the best-fit model and accuracy statistic to be used in determining the model to select. The default zero demand model is used to calculate models for forecast items with six periods or less of demand.
Note. When calculating the forecast for items with zero demand in Holt-Winters models, the system uses the Final Level and Final Trend values from the forecast item. If these values are not defined, then the system checks for these values for the template item at the same level in the forecast view. If these values are not defined, then the system checks for these values for the control group. If the system finds no values defined for these fields, then it uses the first historical period’s adjusted demand as the Final Level and 0 for the Final Trend.
Outlier processing determines how the system processes outliers within forecast models. Outliers are conditions that exist outside established standards for forecast.
Outliers Standard Deviation |
Enter the number of standard deviations that you want the system to use in determining which data point values are outside the range that is considered valid for determining forecasts. The default value is 2 standard deviations. Values are 1 to 3. |
Outliers Ratio |
Enter the adjustment ratio that the system applies to the forecast before it determines outliers. The default ratio is 1.0. |
Outliers Forecast Model |
Select the forecast model that you want the system to use to determine outliers. The default value is Exponential Smoothing. This forecast model uses historical and fitted data to create the next forecast. Recent values are given relatively more weight in forecasting than are older observations. |
Replace Outliers Required |
Select to replace data that lies outside a certain standard deviation. The system realigns data, making adjustments for any outliers that it finds. |
Access the Model Controls page.
Include Simulation and Calculation Models
Select the forecast models that you want simulation routines to use in determining the best fit for a forecast. The system uses only the models that you select during simulation for items in a view that are associated with this control group. Select at least one simulation model.
See Simulating Item Forecasts.
Select to use an equation to analyze the relationship between two or more quantitative variables to predict one from the others. The model measures the relationship between two variables, X and Y, where X is the independent variable and Y is the dependent variable. A particular observation of Y depends on X and an additional random error. |
|
Select the average of a data series over a specific number of preceding periods. When a new value is added, the system drops the last period from the calculation, so that the specific number of preceding periods remains constant. During simulation, the system uses the average to determine the minimum error moving average model. The moving average model does not effectively process significant trends in data, and the system must store all historical data to create the moving average. In the case of time series data, use this model to eliminate unwanted fluctuations, thereby smoothing the time series. The appropriate number of preceding periods should be determined by selecting the number of periods that yield the least amount of error. |
|
Select to replace the oldest moving average with the most recent moving average instead of replacing the oldest observations within the model. The system uses weights with dates in this model and places more importance on dates with higher weights. For example, if the most recent date has more weight, then the most recent date value has more influence on the weighted moving average. The forecast model helps overcome the strong effect of extreme values within a time series by assigning current data more weight than older data. The start and history parameters are the same as those in moving averages. |
|
Select to use historical and fitted forecast data to generate the next forecast. Fitted data is data created by a line along a series of data points. The system creates the fitted line by the forecasting technique. The system also uses the fitted line as a base line for the forecast. The underlying concept of exponential smoothing is that recent values are given relatively more weight in forecasting than older observations. |
|
Select to process sporadic data. For example, suppose that a manufacturer has inventory that is constantly in sporadic demand; that is, there are no precursors to when orders are placed or the demand is often zero, even though the average demand may be for several units. For example, in a three-period moving average, the February, March, and April actual values evenly determine May's forecast. Depending on the level that is given to the series of data, May's forecast heavily relies on April's actual values and fitted values. Exponential smoothing is most effective as a forecasting model when cyclical and irregular influences compose the main effects on the time series values. |
|
Select for a refinement of the seasonal-decomposition model. Census X11 seasonally adjusts and decomposes forecast data through a series of predefined steps: seasonal, trend, cycle, and random or irregular. A seasonal component of a time series occurs regularly, such as an annual holiday, while a cycle's duration varies from cycle to cycle. Economic variables are cyclical and a depression occurs at irregular intervals. |
|
Select another way of decomposing a time series. Box Jenkins is a multistep model-building strategy for analyzing and forecasting time series data by looking for an adequate model in the group of models that are known as auto regression moving average and auto regression integrated moving average processes. PeopleSoft Demand Planning optimization automatically selects the best model that fits the data. |
|
Select a derivative of exponential smoothing. The difference is that with adaptive exponential smoothing, the alpha value changes systematically from period to period to allow for pattern changes in the historical data. Exponential smoothing uses one level (alpha value) to create the forecast. In this model, the level changes during the life of the forecast to adapt to changes to the data. |
|
Select an exponential smoothing technique that incorporates growth and seasonality into the forecast by producing seasonal lift factors for each seasonal period. This model is similar in principle to simple exponential smoothing, where it calculates alpha to measure the level of trend in the forecast. However, the Holt Winters model also adds the parameter, gamma, to create a linear trend in the forecast, as well as the parameter, beta, for seasonality. The model has these components:
|
|
Select to create a linear equation. The model performs two simple exponential smoothing forecasts and then adjusts for the linear trend in the data. Double exponential smoothing-Brown is similar to double exponential smoothing, because the goal is to create a linear trend, but it does so without adding additional parameters to the equation. Because forecasts can be expressed as a function of the single- and double-smoothed constants, the procedure is known as double exponential smoothing. This model is most appropriate for data that shows a linear trend over time. |
|
Select to calculate alpha to measure the level in the forecast. This model is similar in principle to simple exponential smoothing. Double exponential smoothing-Holt also adds the parameter gamma (the trend component) to create a linear trend in the forecast. This equation is similar to a linear regression line and is useful when a product experiences an exponential growth or decline while not exhibiting recognizable seasonality. If the data is dynamic and does not change due to seasonal factors, then this model is beneficial. |
|
Select to create a linear equation. The model performs three simple exponential smoothing forecasts and then adjusts for the linear trend in the data. Triple exponential smoothing-Brown is similar to double exponential smoothing, because the goal is to create a linear trend. This model is most appropriate for data that shows a linear trend over time. |
Contains five values that simulations and forecast optimizations attempt during the process of determining the minimum error when using the moving average model. A moving average forecast is the average of a data series over a specific number of preceding periods. When a new value is added, the system drops the last period from the calculation, so the specific number of preceding periods remain constant.
Values are anything equal to or less than 12. The default values are 3, 5, 7, and 0.
Access the Tracking page.
Enter the number of recent consecutive periods of zero demand that the item should have before the forecast item is suspended. When the system suspends a forecast item, the item's forecast becomes zero and the system no longer provides forecasting for the item. Values for consecutive periods are between 1 and 9999. The default is 6 monthly periods. Adjust this value if the periods that the view references are weekly periods. The presence of nonzero forecast adjustments, active direct nonzero demand adjustments in the most recent suspend forecast periods of history, or a future forecast effective date prevents this suspension. You should provide input that is based on the characteristics of the forecast items. Typically, six months is a reasonable setting for items that die gradually; however, for items with extreme seasonality, the minimum value should be the length of the season, 12-month periods. For this reason, items that exhibit extreme seasonality, including some zero periods, usually require their own control groups. Note. The actual length of the decline is sometimes difficult to set, and unless you have a very large number of items to manage, it might be better to inhibit manually those items that you do not want to forecast. |
|
Select this check box to indicate that you want the system to depromote the actual demand automatically in periods with promotions. The value that you select appears in the field by default. When depromoting, the system removes the effect of a promotion from the actual demand in promoted periods. As a promoted period, which is indicated by the presence of a promotional adjustment to the forecast, moves into history, the system creates a corresponding adjusted demand entry. The adjustment is set equal to the prorated forecast, and the adjusted demand reason is set to promotion. This applies to both promotion adjustments, including manual adjustments that are made through the workbench and those adjustments that the system automatically adds through creating and maintaining event management instances. |
|
Enter a growth percentage that represents the lowest limit for growth. The reasonable growth factor specifies the minimum and maximum allowable projected annual percentage increase for a forecast item. Reasonableness is expressed as a percentage and is used to check the projected annual growth to ensure that the forecast is realistic. If forecast values are outside of either the minimum or maximum percentages, the system automatically alerts the forecast analyst with an entry in the Work Queue. Growth is calculated by using this formula: Statistical forecast for the next year / adjusted demand for the past year * 100. The normal range of settings is 50 percent and 150 percent. Values are between 1 percent and 500 percent and the low percentage must be less than the high percentage. |
|
Select to indicate that the system should adjust trading days during forecast processing. The system uses trading days in conjunction with Census X11 models that use the ForecastX™ system. Trading days are adjustment factors that the system applies to any calendar day when the forecast daily weight is greater than zero. The X11 variant enables you to test and adjust for trading day variability in the model. The system uses weights that are established in calendars to adjust totals depending on the number of respective trading days on the calendar, so that you can account for trading days in one step. This produces more accurate forecasts when trading days are important. |
|
Enter a growth percentage that represents the highest limit for growth. You use this field in conjunction with the Low Growth % field. Make sure that the value that you enter is greater than the low growth value. |
|
Enter the number of periods during which an increasing or decreasing trend is reduced to zero. Valid values are between 1 and 9999. The default value is 18 periods. Trends are increases or decreases in a data series that persists over an extended period of time. Technically, it is defined as the rate of change in the base. Reasonability is expressed as a minimum and maximum change in the trend. Trends that fall outside the band generate a warning alert in the work queue. If the automatic adjustment flag is turned on, the system also adjusts the demand to the top or the bottom of the band. |
|
Select the signal type for evaluating structural errors that exist in a forecast. Using tracking signals, you can identify situations where a forecasting model is over or under forecasting. When these situations occur, it is possible to adjust the forecast to obtain additional accuracy. Generally, the only time that a forecast should be adjusted is when the tracking signal is at + or - 0.5. Note. Positive numbers indicate underforecasting, and negative numbers indicate overforecasting. You can leave this field blank or select one of these values:
|
|
Enter a value for the smoothing weight if you selected the tracking signal type exponential smoothing. The weight measures the influence of the smoothing in the forecast. Smoothing averages the high and low quantities of a forecast. Using the smooth weight with standard smoothed average deviation (SAD)/ mean absolute deviation (MAD), you skew data to the front of or back of the data set. Values are 0 to .5. A .2 weight skews the data to the front of the dataset. |
|
Enter a value that replaces the tracking test for life profile items only. The value sets the limit that causes a tracking signal to activate. The lower the value, the more likely it is that the system sets a tracking signal for the life profile item. |
Access the Reset page.
When one of the item's forecasts exceeds the parameter that you define for the fields, the system automatically resets the item. You set the parameters by using the Model Control page when you maintain forecast items.
Also when you define an item to be reset, you receive a Work Queue alert.
Select the type of forecast that you want the system to use for determining errors that might cause an item to be reset. Errors can be the difference between actual values and forecasted values or values that exceed certain defined limits. For example, when an item's forecast exceeds the error coefficient limit that you enter, the system automatically resets the item's Calculation Type field to Best Fit. Then, the next time that you calculate the forecast, all items that are reset accordingly have their forecasts recalculated. Values are Statistical, Prorated, and Adjusted. |
|
Enter a value that represents a relative error threshold that causes the system to reset a forecast item. The higher the value that you enter, the less likely it is that the system resets the item. |
|
Enter a value that causes the system to trip a tracking signal. Bias is overforecasting or underforecasting on a consistent basis. The bias ratio is the ratio of the SAD to the MAD. The lower the value, the more likely it is that a tracking signal is set. The normal setting is 0.6 for B and C items, and 0.3 to 0.4 for A items. You can be more sensitive to developing bias for A items with the relatively low setting. A measurement procedure or estimator is biased if on the average it gives an answer that differs from the truth. The bias is the average (expected) difference between the measurement and the truth. |
|
The bias signal limit indicates the number of consecutive tracking signals that must be set before the system resets an item. Occasional erratic demand swings can cause tracking signals to be set or cause bias to be indicated even when there is none. Higher values of the bias signal limit reduce the sensitivity of the system and the possibility of false alarms. The normal bias signal limit setting of 2 is appropriate for most items; however, if the system gives some false indications of bias, you might want to set problem items up in a control group that uses a higher bias signal limit or bias test. |
|
Enter a percentage for a life profile item that indicates the threshold outside of which the system should generate a Work Queue alert when a change to either the life profile volume or weight results in a change to the expected volume. The system generates a Work Queue alert if the change is more or less than this percentage. |
|
Enter the number of periods that must elapse between model changes. The default setting is 6. This value applies to monthly periods. If you use this control group in a view with weekly periods, then this value might be much higher. Making model changes more frequently than this can introduce a degree of instability or unpredictability. A value of 99 suppresses model changes altogether. |
Indicate the forecast types for which you want to reset history. The history reset options control which historical forecast data series are reset or overwritten when you perform a simulation or calculate the forecast in Reset mode. Historical data includes statistical, adjusted, and prorated data.
Indicate which types of evaluated forecast you want to reset or overwrite when you perform a simulation or calculate the forecast in Reset mode. Evaluated forecasts include statistical, adjusted, and prorated forecasts.
Access the Work Queue page.
Work Queue messages are system-defined messages that produce alerts when exceptions occur during the generation of a forecast. The system produces the alerts when errors, warnings, or changes in status occur. Using this page, you can override default values to indicate to the system whether or not to use a message to produce an alert for the control group with which you are working. For example, you can remove a message at levels of the view where the message is not relevant or when you are in a start-up mode and do not want to see alerts for all new forecast items in the view.
The system determines whether messages are created during forecast generation based on whether the Record Error check box is selected. If you remove the selection, the system does not include the messages in the Work Queue. When copying a control group within a view, the system carries the message settings to the copied control group.
See Also
Using PeopleSoft Demand Planning Work Queue Messages
Access the Control Group Items page.
Items on this page are display-only. The Level, Forecast Item, and Short Description columns display items at each level of the control group that you select.