Best Fit Analysis
Best Fit Analysis compares historical demand and uses it to calculate a forecast value for each Forecast Method over a specified time period. When the Best Fit Analysis process is run, it uses rules to determine whether to replace, eliminate, or keep forecast methods in place.
 |
Forecast Methods are made available to the Best Fit Analysis process through the configuration of Forecast Parameters. The forecast method that results in the smallest forecast error over the specified time period is designated by the application as the Best Fit forecast method.
|
Best Fit Analysis may be used at the outset to identify which forecast methods to apply to which pairs. It can be run periodically to ensure that the selected forecast method still produces the smallest error over time. Sometimes a SKU exhibits unstable demand and unstable Best Fit assignments, which result in unstable levels which, in turn, can lead to excess or stock-outs. As a result, leaving typical pairs under the control of Best Fit forecasting is strongly discouraged. Use Best Fit Analysis periodically to confirm the assigned forecasting method instead.
|
|
For best results, three years of demand history (36 slices if counting in months) and a one year forecast window (12 slices if counting in months) are required when using Best Fit Analysis.
|
There are two methods of the Best Fit Analysis process for determining the Best Fit forecast method:
• Auto Approved — This method runs Best Fit Analysis and assigns the resulting Best Fit forecast method to the selected stream automatically. If Best Fit is specified at the stream configuration level, it will run once and then assign the forecast method at the SKU level. This prevents the Best Fit process from running again on the SKUs in the stream.
• Recommended — This method runs Best Fit Analysis and the Best Fit recommendations are posted to the Best Fit page for review and approval.