Using an intuitively understandable user interface, marketers and campaign designers can oversee trends in the data and identify most influential customer attributes for a given campaign objective (such as customer retention or upsell). They can run what-if scenarios and get automatic recommendations for optimizing the campaign revenue based on up-to-date customer behaviour data.
In order to facilitate intuitive interpretability of prediction results, the underlying model is based on a Bayesian Network approach, which allows to track the influences of input variables to the target variable. The closed-loop approach allows for the predictive model to automatically adapt to changing customer behavior. The system is set up on a big data stack using SPARK streaming for online data processing.