Evaluating
Forecast Performance With the PEER Planner Performance
options, you can review forecasts made for the item in the past and compare its
performance with historical values (actuals).
The numbers in the lowest diagonal represent one-step ahead forecasts beginning at
Start period for the duration of the Horizon period. In the grid, for example, forecasts
are made on a monthly basis; one-step ahead forecasts, therefore, are made one month prior
to the Horizon date. The numbers in the second off-diagonal are the two-step ahead
forecasts (made two months ahead), and so on.
You can see from the grid that the forecast for May, 1997 changed over time. In the
February view (#1), the forecast for May is 140,808 units. In the forecast made the
following month (#2), the forecast for May is 143,979, and in April (#3), the forecast for
May is 131,315. The actual value for May is 155639 units. From the Waterfall chart, you
can see the difference between forecast and actual values for each period, as well as how
forecasts for each period change over time.
Displaying forecast error, percent error or absolute error. You may also wish to know
statistically how accurate your forecasts are. The buttons in the Scale panel let you
change the forecast values in the Waterfall chart to percent or absolute deviation
relative to actuals or forecasts. These measures let you see results in terms of accuracy
or variance from forecast.
Displaying forecast errors over Lead-Time. Forecast errors can also be displayed over
lead times specified in the database or modified on the screen.
Graphing the performance results. Another way of examining errors is to view the graph
of the values in the grid. While the Performance screen is active and the Waterfall chart
is displayed, press Graph on the Main menu. If you would like to see a different graph,
select the numbers for the grid first. While the graph is displayed you can change its
appearance. Graphical displays will show the same data that are on the grid display.
Bar chart of Forecast Error (Actual minus Forecast) for one-step ahead forecasts
|