Predictive Forecast

The Predictive Forecast process uses historical data already in your models to forecast future results. Typically this is done once or twice per year, to aid in planning; Predictive Forecast can be an excellent and faster alternative to the InfoFlex process (though whichever method you choose, you will likely still make manual adjustments to your plan).

The process creates a forecast by taking the source leaf-level actuals of a specific time frame and then applying time-series-based algorithms. The resulting forecast data then populates the specified target version and time period. If you have seasonally consistent data, this gives you a starting point for future planning.

 Tip:  Predictive Forecast is best used for specific accounts/expenses that have historical seasonality in the data set. In other words, the more historical data, the more reliable the forecast.

General tab

  • Description: Optional. Up to 250 characters.

  • Model: The model where the data resides.

Source/Target Members tab

Your selection from Model in the General tab determines the dimensions available in the Source Data section and the Version dimension in the Target Members section.

  • Source Data

    • Select the historical data that you want to base the forecast on, from at least one dimension.

       Note:  You must choose leaf-level data only.

    • Select the member of Version that you want to draw the historical data from.
    • Use Start Period and End Period to indicate the time frame for the historical data.
  • Target Members
    • Select the member of Version that you want the forecast data written to.

Forecast Properties tab

In the Forecast/Properties tab, you define the kind of forecast you want.

 Note:  The two statistical algorithms used by the Predictive Forecast process have many variations; therefore, if you run your data set in another forecast tool that uses these algorithms, you will almost certainly get different results.

  • Forecast Model: Select the time-series-based forecasting algorithm you want to use:
    • Best fit: (Default) Runs both ARIMA and Holt-Winter's algorithms, and then uses machine learning to compare the results and determine which is the statistically most accurate, based on previous actuals.
    • ARIMA: A version of the ARIMA (Autoregressive integrated moving average ) forecasting algorithm.

    • Holt-Winters: A version of the Holt-Winters forecasting algorithm.

       Note:  To learn more about the specific versions we chose, click the following links (separate browser tabs will open): ARIMA; Holt-Winters.

  • Forecast Start Date: Select the time period the forecast data should start at.

  • Forecast Duration: Set the number of periods you want in the forecast.

     Note:  The type of period is always based on the leaf-level of your Time dimension.

Data tab