Show TOC

 Defining Optimal Amount of Historical Data in DMA Model

Use

For the forecast with the dynamic moving average (DMA) model, the system determines the optimal amount of historical data before it calculates the forecast. The system does so in various ways, depending on whether it is defining historical data for the order item forecast or for the average demand size per order item (demand/item) forecast.

Features

Optimal Amount of Historical Data for Order Items

If there were no order items in the last 26 weeks, the system sets the forecasted number of order items to zero.

If this is not the case, the system executes t-tests to determine the optimal amount of historical data.

The system tests the current 13 weeks against the previous 13 weeks. If there is a significant difference, it tests the current six weeks against the previous seven weeks. If there is a significant difference, the system uses the current six weeks for the forecast. Otherwise it uses the previous 13 weeks. If the first test does not show a significant difference, the system tests the current 26 weeks against the previous 26 weeks. If there is a significant result, the system uses the current 26 weeks. Otherwise it uses the previous 52 weeks.

Each of these t-tests (13 against 13, 6 against 7, 26 against 26) has varying acceptance limits for the significant difference. You can specify these on the SAP Easy Access menu under Start of the navigation path Advanced Planning and Optimization Next navigation step Service Parts Planning Next navigation step Planning Next navigation step Forecasting Next navigation step Forecast Profile End of the navigation path on the Model Parameter tab page in the following parameters:

  • Positive Significance Level T-Test 6:7 for Item in DMA Model

  • Positive Significance Level T-Test 13:13 for Item in DMA Model

  • Positive Significance Level T-Test 26:26 for Item in DMA Model

If t is less than zero, the following acceptance limits apply. You also specify these in the forecast profile on the Model Parameter tab page:

  • Negative Significance Level T-Test 6:7 for Item in DMA Model

  • Negative Significance Level T-Test 13:13 for Item in DMA Model

  • Negative Significance Level T-Test 26:26 for Item in DMA Model

The system performs the t-tests according to the following formula:

  • x and y are the average values of the two tests.

  • s x and s y are the standard deviations of the two tests.

  • n x and n y are the tests sizes.

  • (n x + n y – 2) are the degrees of freedom of the test.

Optimal Amount of Historical Data for the Average Demand Size per Order Item

If there was less than four order items in the last 52 weeks, the system uses all the 52 weeks to calculate the demand/item.

If there were four or more order items in the last 52 weeks, the system checks the following conditions:

  • The optimal amount of historical data for the forecast of the order items was 6 or 13 weeks and the following tests are successful.

  • The demand forecast per month is greater than the Param1: Relevant Periods for Demand /Item in DMA Model parameter in the forecast profile on the Model Parameter tab page.

  • The increase of the forecast order items of last month’s forecast is a least as large as the Param2: Relevant Periods for Demand /Item in DMA Model parameter in the forecast profile on the Model Parameter tab page.

  • The number of order items during the relevant historical periods for the order item forecast is at least a large as the Param3: Relevant Periods for Demand /Item in DMA Model parameter in the forecast profile on the Model Parameter tab page.

If all conditions are met, it uses the same time span of historical data for the demand/item as for the order items.

If not all conditions are met, the system proceeds as follows:

  • If there was a minimum number of order items in each of the last two periods over 13 weeks, the system performs a t-test for these periods to determine whether the difference of the number of order items in these two periods is significant. If the difference is significant, the system uses 13 weeks as the optimal amount of historical data.

    • You can specify the minimum number of order items that must be available for the system to perform a t-test for the last two periods over 13 weeks in the forecast profile on the Model Parameter tab page in the Param4: Relevant Periods for Demand/Item in DMA Model field.

    • A difference is considered to be significant is it is greater than the Positive Sign. Level T-Test 13 :13 for Item in DMA Model parameter or if it is smaller than the Negative Sign. Level T-Test 13:13 for Item in DMA Model . These parameters are in the forecast profile on the Model Parameter tab page.

  • If the system has not calculated 13 weeks as the optimum amount of historical data, the system checks whether there was a minimum number of order items in either of the two last periods over 26 weeks. If so, it performs a t-test for these periods to calculate whether the difference of the number of the order items in these two periods is significant. If the difference is significant, the system uses 26 weeks as the optimal amount of historical data, otherwise it uses 52 periods.

    • You can specify the minimum number of order items that must be available for the system to perform a t-test for the last two periods over 26 weeks in the forecast profile on the Model Parameter tab page in the Param5: Relevant Periods for Demand/Item in DMA Model field.

    • A difference is considered to be significant is it is greater than the Positive Sign. Level T-Test 26:26 for Item in DMA Model parameter or if it is smaller than the Negative Sign. Level T-Test 26:26 for Item in DMA Model . These parameters are in the forecast profile on the Model Parameter tab page.