Churn or U-turn
Identify customers with high attrition probability using ensembles of models. Find actionable patterns in churner attributes.
To build an effective customer retention program, managers have to come to an understanding of why customers leave and identify the customers with high risk of leaving by accurately predicting customer attrition.
If the success depends on how accurately one can predict customer attrition, an assemble of predictive models would be a good choice since they often outperform a single model. However, using more complex models with more predictive power also means that it can be hard to understand the drivers behind the models. In order to describe the main drivers, and then understand why the customers are leaving, one can use a simple model such as a decision tree, which is very straightforward to understand.
At Innohead we have years of experience using ensembles of models and in managing the tradeoff between model complexity and simplicity.