Predictions
Note: The Predictions feature is only available to customers who have purchased Oracle Infinity IQ . Learn more about Oracle Infinity IQ.
The Prediction feature allows you to predict the likelihood of visitors exhibiting specific behaviors, and then take action on predicted behaviors, in the moment, to affect conversion rates or objectives.
The Infinity Predictions model trains itself to predict the probability or likelihood that a visitor or customer will exhibit specific behaviors within the next five clicks. Once trained, the model can run against live data streaming through Oracle Infinity's data pipeline and insert probability scores that reflect the likelihood of a behavior occurring.
The benefit of leveraging the Predictions feature is that it enables you to take action on what is predicted to happen, so that you can affect outcomes and increase the chances of your customers reaching a desired result.
Prediction model training
To use the Predictions feature in Actions, the prediction model must first be trained on the presence of specific behavioral events in your historical data. Approximately one week of historical data must be collected and available to train the model effectively.
The prediction model training process looks for the following behavioral events that have occurred on your website, mobile app, or other digital property.
-
Add product to cart
-
Close session
-
Download file
-
Purchase product
-
Submit form
The definitions of these behaviors can be found in the Standard behaviors guide.
Prediction probability value
Once the model is trained, the behavior event data collected by the CX Tag or mobile SDK into Infinity will be processed, and a prediction probability value will be appended to each event collected.
The Prediction probability indicates the probability or likelihood that the behavior will occur in the next five clicks. The value is represented as low, medium, or high.
The prediction probability value will stream through the Infinity pipeline along with all other behavioral event data collected. You can then use this data in Infinity Actions, so that you can engage with visitors who exhibit the predicted behavior.
Example: You want to configure a prediction for a user abandoning a purchase in the next five clicks, so that you can initiate the proper personalized content. To do this, you set the prediction probability value for the Purchase product behavior to low.
Predictions and actions
When a prediction is configured as part of an Action, the action will listen for the following:
-
The defined rules (behaviors and filter conditions) in the action.
-
The presence of the prediction probability, depending on how it is configured in the action.
If the conditions are met, the action can then deliver contextualized information about that visitor to personalize content on the web page or mobile app screen while the visitor is still engaged and affect the outcome of the user's behavior.
To configure predictions in an action, the prediction model must be trained first.