According to estimates, it is five times more costly to attract new customers than retain existing ones. Preventing customer churn has thus become an essential issue in many sectors, including banking. Using churn prediction models, banks can identify potential leavers and then decide on the right course of action to prevent their departure. Our client–the Polish branch of an international banking institution–also relied on such models but sought ways to improve its existing system.

Working together with our client’s team, we performed a deep refactoring of the churn prediction model. We tackled both the business and technical aspects of the existing solution, extending the scope of the analysed sets with unstructured text data describing transactions made by clients. We deployed Machine Learning techniques in the modelling process. At the end of the project, we also delivered a bespoke workshop for our client’s team to help them build the competencies they needed to make the most of the new solution and to be able to create analogous models without our assistance.

The new model has significantly improved customer churn predictions made by our client–by over 10% compared to the previous one. As a result, our client optimised related processes and is now able to better identify customers at risk of churning and prevent them from leaving.