Skip to main content

 

Planview Customer Success Center

Changepoint Analytics: What-if Analysis Extension

On Friday, November 13, 2020, the What-if Analysis extension becomes available in Changepoint Analytics. This extension builds machine learning models to explore scenarios, test, and evaluate business assumptions. It helps you explore and predict a range of possible outcomes when one or more measures change, which in turn allows you to make smarter and quicker decisions driven by data as well as to plan for the future.

You can use this tool to predict a target measure and answer what-if questions such as:

  • How much I have to sell to hit my target?
  • What if we reduce inventory capacity to a different number?

How to use What-if analysis

The What-if Analysis extension is available for Dashboards in Changepoint Analytics and shows up under native visualizations when you create a dashboard.

To add a What-if Analysis widget to a dashboard:

  1. Under Native visualizations, click the What If icon.
  2. In the visualization, drag a data item from the Source tab to the Target Column slot of the visualization. The target column is the column that you would like to predict.
  3. Drag key drivers (measure columns), or predictors, for the target column into the Input Columns. Key drivers are the main cause of the target column.
  4. If you are not sure what are possible key drivers, you can click on the Autofill input columns button to automatically discover key drivers.

How to use What-if analysis

To start your what-if analysis, click the Create model button. This process takes approximately one minute. After training is completed, it should be able to estimate the target column for any value of input columns.

The prediction strength measures how good is the model in predicting the target column. If the prediction strength is good, the base and prediction bars will have relatively the same values.

If you are not happy with the model prediction strength, you can apply additional training by clicking on the Train model button. The additional training gives the model another chance to better fitting your data to the model and results in better model with higher prediction strength. This process usually takes a few minutes.