Show TOC

 Seasonal Test

Use

The system performs the seasonal test as part of automatic model selection . If the seasonal test is successful, the system selects all forecast strategies that you assigned to the seasonal test in Customizing. For more information, see the IMG for Advanced Planning and Optimization under Start of the navigation path Supply Chain Planning Next navigation step Service Parts Planning (SPP) Next navigation step Forecasting Next navigation step Configure Automatic Model Selection End of the navigation path .

The seasonal test checks the demand history of the previous 24 periods for a location product to see whether it has seasonal demand. If seasonal demand exists, a seasonal model is appropriate for the corresponding location product, such as the seasonal trend model with fixed period groupings .

Features

Prerequisites for the Test

For more information about the prerequisites a location product must fulfill for the system to test it using the seasonal test, see Prerequisites for the Seasonal Test .

Test Data

The system performs the seasonal test with the demand after outlier correction. The system thus corrects the outliers that exceed a demand limit. The system calculates this demand limit according to the following formula:

  • The system calculates the average demand from the average of the demand values over the last 24 periods, excluding the highest value.

  • The system calculates the demand deviation from the average of the standard deviation over the last 24 periods, excluding the highest value.

  • You can define the safety coefficient k in the Outlier Correction Parameter for Demand field on the General tab page of the forecast profile. You get to the forecast profile on the SAP Easy Access screen under Start of the navigation path Advanced Planning and Optimization Next navigation step Service Parts Planning (SPP) Next navigation step Planning Next navigation step Forecasting End of the navigation path .

Test

The seasonal test consists of the following parts:

  • Autocorrelation test

  • t-tests

Autocorrelation test

During the autocorrelation test, the system checks whether the demand history of the current year shows similarities to the demand history of the previous year. To do so, the system calculates the correlation coefficient r as follows:

Year1 is the total demand of the current year. Year2 is the total demand of the previous year.

x i is the demand of the period iof the current year. y i is the demand of the period iof the previous year.

If r is greater than your entry in the Significance Level for Autocorrelation Test in Seasonal Test on the Test tab page in the forecast profile, the seasonal test is successful and the system does not perform any other tests. You get to the forecast profile on the SAP Easy Access screen under Start of the navigation path Advanced Planning and Optimization Next navigation step Service Parts Planning (SPP) Next navigation step Planning Next navigation step Forecasting End of the navigation path .

If you have already performed a seasonal test for the forecast of the corresponding location product, the system proceeds as follows to ensure forecast model stability: It does not compare r with your entry in the Significance Level for Autocorrelation Test in Seasonal Test field but with your entry in the Stab. Signif. Level for Autocorrel. Test in Seasonal Test field.

t-Tests

If the autocorrelation test was not successful, the system uses t-tests to determine whether the average demand in a certain part of the year deviates greatly from the demand during the rest of the year. To do so, the system uses a time window that moves across the demand history of the previous two years. If, for example, the length of the time window is three periods, the system tests the demand of periods one, two, and three against the demand of the rest of the year. It then tests the demand of period two, three, and four against the demand of the rest of the year, and so on.

You can specify the minimum size of the time window in the Minimum Window Size for T-Test Run in Seasonal Test in the forecast profile on the Test tab page. The maximum window size is six periods. The system always performs a t-test for a specific window size. It thereby begins with the minimum window size. If the system can determine seasonal demand using this window size, it does not perform further t-tests with other window sizes. However, if the system cannot determine seasonal demand using this window size, it performs further t-tests with other window sizes.

The system performs a t-test according to the following formula:

n x is the window size. n y is 12 – n x .

s x and s y are the standard deviations.

The seasonal test is successful is the following conditions are met:

  • │t│ of the current year is greater than your entry in the T-Test Run Signif. Level for Current Year in Seasonal Test field in the forecast profile on the Test tab page.

  • │t│ of the previous year is greater than your entry in the T-Test Run Signif. Level for Previous Year in Seasonal Test field in the forecast profile on the Test tab page.

  • The values for t of the current year and the previous year are either both positive or both negative.

If you already use a seasonal forecast model for the forecast of the corresponding location product, the system does not use your entries in the fields above, but instead uses your entries in the T-Test Run Stab. Signif. Level Current Year in Seasonal Test and T-Test Run Stab. Signif. Level Previous Year in Seas. Test fields.

Replace Seasonal Test

To replace the seasonal test with a test you have defined yourself, you can implement the Business Add-In (BAdI) BAdI: Modify Seasonal Test (/SAPAPO/PFCS_FCST_SEASONTEST).