Blog

If you’re reading this post then you’re also likely aware that IBM Planning Analytics (IBM PA) already provides a univariate forecasting capability to automate the planning cycle by predicting future outcomes of data that is subject to seasonal fluctuations.

Leverage Multivariate forecasting to improve forecast accuracy

With the latest release of IBM PA however, users can now also employ multivariate forecasting to predict data points that are correlated to other variables within the model which may not be exposed to seasonality themselves. This new functionality enables business users to leverage predictive forecasting for a much wider range of use-cases than ever before!

Multivariate forecasting is user-friendly for immediate deployment

While multivariate forecasting might sound sophisticated, it is actually presented in a very user-friendly manner so that it can be quickly put into play by business managers and regular users in 15 minutes or less by following four simple steps:

Step 1) Setup the Multivariate Forecast:

Select row(s) to forecast, open the Multivariate Forecast sidebar, and choose the variables within your model which are most relevant to the data being projected. Users have the ability to include future values of variables, forecast at consolidated level elements and apply a proportional or relative proportional spread to the results, as well as the ability to save the predicted values to another element such as a different version than where the historical data resides.

Step 2) Preview the Forecast Output Chart:

The system automatically creates visualizations for values that are the target of the forecast, as well as the variables that contribute to the forecast calculation. For example, if you create a forecast for Gross Margin that considers Units Sold and Sales Price as influential factors, you'll get one visualization with historical and forecast values for Gross Margin, and separate visualizations for Units Sold and Sales Price that can help you understand how these variables influence the Gross Margin forecast.

Step 3) Preview Statistics:

Review predictive accuracy, correlations for each variable, and statistical measures of error.

Step 4) Calculate the Multivariate Forecast:

Sit back and relax while IBM PA does the heavy lifting, then review the results.

To Learn More

Click the link provided here for a 2-minute IBM video of Multivariate AI Forecasting https://mediacenter.ibm.com/media/1_2np7852f

For more details, you can review multivariate forecasting and all of the new features introduced in IBM PA 2.0.94 at https://www.ibm.com/support/pages/node/7130162

IBM continues to add capabilities and functionality to the IBM Planning Analytics platform at regular intervals. If you have questions, we’re happy to discuss, or we can even provide a demo if you’d like to take a closer look. You can reach us any time at https://www.acgi.com/contact-us-meeting