Exploring Automated Prediction Options in EPM Planning

EPM Planning has many options available for automated Predictions using statistical methods and machine learning algorithms. These predictions use historical data to predict future periods and can be used as an alternate estimate for the future, or to sense check manual forecasts using advanced methods.

This blog will cover how and when to use the different options, along with providing a key feature comparison of Predictive Planning vs Auto Predict vs Advanced Predictions.

Predictive Planning

Predictive Planning is an interactive prediction, which has been available for many years and is run on demand by end users via a data form. It uses the form context to produce a dynamic prediction, best case and worst case for all members on the form (e.g. for all accounts on the row), but they must be viewed one at a time. The only setup required is the data form layout which must meet certain criteria, so this is a great option to start with.

The predictions produced are univariate, which means they only use the history for a single variable to predict that variable forward. For example, to predict IT Costs, the prediction would only use historic IT Costs as a basis for future.

The results are displayed in a graphical panel at the bottom of the form for the users to view at run-time, which can be a great sense check for users to compare to their manual forecasts. The only downside here is that the results are not stored anywhere within the system, so the predictions can only be seen directly from the data form.

Auto Predict

Auto Predict uses the same statistical methods as Predictive Planning, meaning it’s also a univariate prediction, however the rules can be run for multiple members at once and the results are stored in a selected Scenario/Version combination. This means the results can be retrieved around the system in forms, calculations or reports, rather than being viewed one by one on a data form.

Unlike Predictive Planning, Auto Predict rules are setup in the IPM Configure area by an administrator and then run ad hoc or scheduled as a job via the jobs console. The auto predictions can also be setup to consider Calendar Events, such as one-off or repeated events, which may increase accuracy of future predictions.

More on calendars & events here: https://lydia-maksoud-epm.blogspot.com/2025/08/configuring-ipm-calendars-events.html

Advanced Predictions

Similar to Auto Predict, Advanced Prediction rules are setup in the IPM Configure area and the results can be stored and have the option to incorporate Events. The main difference between Auto and Advanced, is that Advanced Predictions are multivariate, meaning they use multiple variable inputs to influence a prediction result.

For example, to predict Sales Volumes, rather than only using historic Sales Volumes (this would be univariate), a multivariate prediction can use other inputs such as industry volumes, selling price or discounts to influence the final sales volume prediction.

Once the prediction has been run, the Feature Importance tab will display the importance of each driver so users can review the most influential factor. 

The other main difference is that Advanced Predictions use more sophisticated machine learning algorithms to produce the prediction. As part of the setup, users get the option to select the most relevant algorithm for the dataset or use the Oracle AutoMLx which will automatically assign the best fitting model.  

Important Note! Advanced Predictions require an EPM Enterprise license.

To summarise the features available using each prediction method, see below feature comparison table:

	Predictive Planning	Auto Predict	Advanced Predictions
End User Run	Yes	No	No
Method	Via Data Form	Setup & run in IPM	Setup & run in IPM
Option to Schedule	No	Yes, via Jobs	Yes, via Jobs
Stores Results	No	Yes	Yes
Variable Inputs	Univariate	Univariate	Multivariate
Statistical Methods	Non-seasonal methods (e.g. ARIMA) & Seasonal methods (e.g. SARIMA)	Non-seasonal methods (e.g. ARIMA) & Seasonal methods (e.g. SARIMA)	Machine Learning Algorithms (SARIMAX, Prophet, AutoMLx etc.)
Calendars & Events	No	Yes	Yes
Produces Fitted, Best & Worst Case	Yes	Yes	Yes
Feature Importance	No	No	Yes
Licensing Required	Standard or Enterprise	Standard or Enterprise	Enterprise Only