Forecast Simulation - Sales Statistics
Forecast simulation is a method for comparing different forecast methods.
To calculate the base forecast for the next period, the forecast method that would have given the most correct result in relation to actual demand for the previous forecast period is selected.
The comparison is made when each forecast is calculated. Information on the selected method is saved.
Description
Forecast simulations are used as an alternative to the four forecast formulas used to calculate base forecasts.
These five forecast formulas are standard formulas for forecast simulations:
- 10 - The forecast for the next period is calculated as the average actual demand for the three preceding periods.
- 11 - The forecast for the next period is calculated as the average actual demand for the three preceding periods increased or decreased by the calculated trend.
- 12 - The forecast for the next period is calculated as the average actual demand for the three preceding periods. The average is multiplied by the ratio of the total actual demand for the three periods for the current year and the total actual demand for the same three periods for the preceding year.
- 13 - The forecast for the next period is calculated as the average actual demand of the preceding, same, and following periods from the preceding year.
- 14 - The forecast for the next period is calculated as the average actual demand of the preceding, same, and following periods from the preceding year increased or decreased by the calculated trend.
The formulas are automatically assigned forecast methods F1, F2, F3, F4, and F5, where other forecast parameters can be defined.
For forecast method 1, methods F2, F3, F4, and F5 are automatically created as simulated forecast methods. F1 is the method created automatically in M3.
It is possible to add more simulated forecast methods based on the system’s four forecast formulas. It is also possible to manually enter methods for forecast simulation other than F1.
Example
The following demand history is recorded for a combination of market, customer group, and item:
| Oct 94 | Nov 94 | Dec 94 | Jan 95 | Feb 95 |
|---|---|---|---|---|
| 90 | 97 | 95 | 98 | 98 |
| Oct 95 | Nov 95 | Dec 95 | Jan 96 |
|---|---|---|---|
| 101 | 104 | 98 | 101 |
The forecast for February 1996 is to be calculated. The forecast simulation covering the five standard forecast formulas in the method is used for the forecast. This means that a comparison is made to arrive at the formula resulting in the most correct value for the previous period’s actual value (January 1996).
The base forecast for January 1996 according to each forecast formula is:
| 10: | (101 + 104 + 98) / 3 = 101 |
| 11: | 101 + 2 = 103 |
| 12: | 101 * (101 + 104 + 98) / (95 + 97 + 90) = 108 |
| 13: | (95 + 98 + 98) / 3 = 97 |
| 14: | 98 + 2 = 100 |
After the forecast for January 1996 is calculated, the result is compared to the actual value for that period. According to the forecast formulas above, the result from formula 10 is the closest to the actual value for January 1996, and is therefore used to forecast demand for February.
The base forecast for February is then (104 + 98 + 101) / 3 = 101.