About Error Measures
The Error measure graph is useful for seeing the accuracy of your forecasts. It contains the number of records at each measure of a specific error measurement. Below the graph are statistics for the median, mean, and standard deviation of the error measurement across all records.
All of the error measures perform calculations using forecasts and actuals (the actual historical data, such as sales data).
While there are several error measures available, the following are used most frequently by most users.
Error Measures
| Error Measure | Description | For Use... |
|---|---|---|
| Last Absolute Deviation Z-Score (LADZ) | The difference between the last period’s forecast and its actual divided by the standard deviation of residuals. At the beginning of each forecast cycle, you can use this measure to analyze your forecast against actuals to find any anomalies, errors, or expected differences (discontinued product, for example). A high Z-score indicates your forecast was far off from actuals. In the Top Exceptions table, you can sort by the Z-score to locate and examine the differences in cases of a high Z-score. Answer the following questions: - Is the data correct? - If the answer is yes, did an unusual event or a new reality occur? - Did CCH Tagetik Supply Chain Planning correctly determine what occurred? A low Z-score indicates there is little difference between what was forecast and actuals. | Each time you begin a new forecast cycle (monthly, for example) |
| Quality Score (QScore) | A measure of the quality of the forecast. The QScore measure is most useful when you are in the early stages of using CCH Tagetik Supply Chain Planning. After you have accumulated at least two months of historical data, you can use the QScore measure to see how your data is performing. It is not intended for use frequently as the information does not change often once you determine your instance is built correctly. The QScore is based on a 0-100 scale that works similarly to a classroom grading scale (10-point increments). A score of 90-100 is a very good Quality Score (A), 80-90 is good (B), 70-80 is fair (C), and so on. A low QScore indicates your system may not be built correctly. In this case, contact your Wolters Kluwer representative to discuss details of your setup. | In the early stages of using CCH Tagetik Supply Chain Planning |
| Forecaster Performance | A comparison of the forecast versus actuals (compares forecast with manual overrides against forecast using the CCH Tagetik Supply Chain Planning automatic option). See Use the Forecaster Performance Error Measure. | Occasionally |
The following error measures are available for those users who require them:
Error Measures
| Error Measure | Description |
|---|---|
| Symmetric Mean Absolute Percent Error (SMAPE) | Forecast minus actuals divided by the sum of forecasts and actuals. This measure is superior to MAPE in that it does not result in high error measures when actuals approach zero. |
| Mean Absolute Percent Error (MAPE) | Forecast (F) minus actuals (A) divided by actuals. This is a popular measure, but it is not as useful as SMAPE or Bias Percent in cases where actual values approach zero. |
| Mean Absolute Deviation (MAD) | The average of the absolute difference between the forecast and actuals. |
| Mean Absolute Deviation Percent | MAD divided by the average of the actuals. |
| Bias Percent | The sum of the forecast divided by the sum of the actuals minus 1. This measure examines forecasts to see if the selected forecasting method has applied smoothing to a degree that generates a bias in the forecast. In a CCH Tagetik Supply Chain Planning record, in the Method section, the option Remove bias lets you remove any bias. See About the Data Smoothing Method. |