Gaining Insights on Future Customer Behavior with Predictive Studio

Predictive models help you gain insight about the future behavior of your customers. Using the Predictive Studio app, you can easily create, train, evaluate and publish predictive models for an either preconfigured or custom predictive score (predictive scenario). The key components in predictive model management are scenarios, models, and training.

A predictive scenario consists of one or more data sources and assigned implementation methods, both of which are tailored to your use case. You can create new predictive models to calculate scores based on preconfigured predictive scenarios.

A predictive model is based on a preconfigured predictive scenario and one or more model fits. It can also include additional details, such as a training set. You can change and enhance predictive models with Model Fits and Scope. A predictive model is used to identify the best fit and publish it to the business scenario.

Training your predictive models is a key step in predictive model management. You can create and adjust model fit and select predictors or use Infinite Insight for automatic selection. You assess fitting quality by means of cross-validation, Lorenz curves, and other methods.

The process to create predictive scores is as follows:

1. Create a predictive model, select a predictive scenario, and define the details needed for the scenario in the Predictive Studio app.
2. Create one or more model fits for your predictive model.
3. Check the quality of your model fits and select the best model fit.
4. Activate your predictive model. The best model fit is used for calculating the predictive score.
5. Use it in Segmentation.

SAP Marketing Cloud provides out-of-the box, predefined predictive scores, such as Consumer Buying Propensity. The buying propensity of a consumer is calculated based on their interactions within a given time period and other attributes. The output score indicates how likely it is that a consumer is going to buy the specified product.

Using these scores, you can easily build, train, and publish predictive models for various targets and contexts. Besides, they can run and manage predictive models directly in the system on their own without the involvement of IT, a developer, or a data scientist.

The graphic below displays the available out-of-the box scores:

To learn more about delivered scores, go to our documentation page.