Measuring Errors—Exercises

IMPORTANT
It is extremely important that you set up and work within your own Server ID in order to maintain your data integrity. If you work within any other server id, you will compromise your exercise data as well as the data of other students. Predictable exercise results require that your data be isolated to your own server id.

Basic Data Setup

Defining Schedule

Purpose: The purpose of this exercise is to show you how to set up a schedule for the steps in the forecast cycle.

Windows:
Demand Plan Server Setup/Scheduling
Forecast Server

  1. Set up a schedule for your server ID.
  2. Start the Demand Planning Server.

General exercise for Defining Schedule

Main Exercises

Purpose: The purpose of this exercise is to familiarize you with error measurement methods and other support features in this area.

Windows:
Forecast Client

Measuring Errors—Forecast Graph

  1. On the Windows taskbar, click the Start button, point to Programs, and then click Forecast Client. Log on to the client using appowner as the user ID and password.
  2. In the Flow ID list, select one of your flows.
  3. In the Part No list, click an individual part.
  4. Click Legend On/Off to maximize the Legend area.
  5. The forecast window is a split window. To view the error portion of the graph, place the cursor on the line at the bottom of the Legend area. Click and drag the line upward, and the Error Graph appears.
  6. Click each point in the Legend area to see the various error methods. Look at the measurement methods, and note how they have developed over time, whether the forecast is getting better or worse, whether you are using the right forecast model, etc.

Measuring Errors—Theil's U-Statistics

  1. On the Windows taskbar, click the Start button, point to Programs, and then click Forecast Client. Log on to the client using appowner as the user ID and password.
  2. In the Flow ID list, select one of your flows.
  3. In the Part No list, click an individual part, and note the Theil's U-statistic value in the Theil's field.
  4. In the Forecast Model list, change the forecast model. Note the Thiel's value, which should have changed. What is the conclusion? Is the new forecast model better or worse then the standard one? 
  5. In the Forecast Model list, change the forecast model to EWMA Level and Trend.
  6. In the Alpha and Beta fields, change the alpha and beta values. Look at the Theil's value, and see how it is affected by changes in the alpha and beta.
  7. On the Part menu, click Best Fit Forecast Parameters. Verify that the alpha and beta values have changed and that the Theil's value has improved.
  8. In the Forecast Model list, click Best Fit. Verify that the Theil's value has improved (this will not happen if the selected model is equal to the result of the best-fit process). 

Measuring Errors—Outlier

  1. On the Windows taskbar, click the Start button, point to Programs, and then click Forecast Client. Log on to the client using appowner as the user ID and password.
  2. In the Flow ID list, select one of your flows.
  3. In the Part No list, click an individual part.
  4. Maximize the Legend area.
  5. Click Confidence Interval On/Off to turn on the confidence interval. The dark gray area indicates one times the standard deviation, and the light-gray area, two times the standard deviation.
  6. On the Window menu, click Forecast Table.
  7. In the Adjusted Demand field, look for values colored yellow or red. These colors resemble periodic data elements that are in the dark-gray area or outside the confidence interval in the graph.  

Measuring Errors—Tracking Signal

  1. On the Windows taskbar, click the Start button, point to Programs, and then click Forecast Client. Log on to the client using appowner as the user ID and password.
  2. In the Flow ID list, select one of your flows.
  3. In the Part No list, click an individual part.
  4. Note the value in the Tracking Signal field. In the Delta field, increase the delta value to 0.9. The tracking signal has now increased significantly because the reaction time has been reduced.
  5. Decrease the delta value to 0.1. The tracking signal is reduced to a lower value because the reaction time has been increased.

Copying forbidden: The material in this manual is protected by copyright law, and may not be copied entirely or in part, without the prior permission of Industrial & Financial Systems AB. © 2012 Industrial & Financial Systems AB