Spreading forecast engine results to base periods using the Cycles.Period Measure

If the Spreading Measure for time-phased results parameter is defined and enabled for the cycle period, the period value is prorated using proportions, based on the values defined for each item and locatin of the Cycle.Period measure.

The period interpolation is the value for each item, location, base-level period in the period horizon divided by the sum of the value for each item, location, aggregate period (at the Forecast Engine.Period Level).

Example of the period values:

  • Forecast Engine.Period Level = Months
  • Cycle.Calendar level to store scenario values = Weeks
  • Forecast Engine.Spreading Measure for time-phased results = Saved Statistical Forecast [DPLS_FSTAT_EXT]
The ratio of week FY12 W18 = (Value of DPLS_FSTAT_EXT for FY12 W18 Item and location) / (Value of DPLS_FSTAT_EXT for FY12 M05 Item and Location).
Note: The measure values are defined for only base-level items. Therefore the value of DPLS_FSTAT_EXT for FY12 M05 Item and Location is a representation of the sum value at the aggregate node. The value of DPLS_FSTAT_EXT parameter is derived.

The spreading values from the base-level periods are used to define the scenario values irrespective of the number of levels that exist between the selected Forecast Engine.Period level and the Cycle.Calendar level.

When prorating, these time-phased measures are interpoled using the Spreading Measure for time-phased results parameter:

  • Forecast
  • Model fitting History
  • Online Model Fit
  • Retrospective Model Fit
  • Seasonal Indices
  • Outliers
  • Step Change Exceptions
  • Tracking Signal Exception
  • Upper Prediction Interval
  • Lower Prediction Interval
For forecast engines, the exception measures are time-dependant. Each period with an exception is set to 1. If applicable, the values are prorated to the base level to store the scenario values. You can also view the data at the forecasted item, location and period level.
Note: The Forecast Engine output results that are time-independent are not affected by period interpolation. So, the value of PCONST is considered as normal.