Factors that affect the loyalty of mortgage customers : Predicting customer churn using machine learning

Acquiring new customers implies a certain cost for the banks, so there is a problem when new customers decide to leave the bank, shortly after the onboarding process. This thesis explores which factors, and especially what products and services, that affect the loyalty of new mortgage customers. These insights were partly generated through descriptive statistics, but also through a classification model using machine learning. Fixation periods proved to be very important in the short term, as a particularly small share of the customer base left the bank with a fixation period of two years or longer tied to their mortgage loans. In general, all additional products and services had a positive effect on customer loyalty. In the short term, first-time homebuyers left the bank at a greater rate compared to customers who had switched banks prior to leaving. In addition to this, the model also suggests that younger customers churn more frequently than older customers, and that a higher loan amount tends to increase the likelihood of a customer leaving. It also highlights the fact that customers living in the vicinity of Stockholm, appear to leave at higher rates than the rest of the country. However, these conclusions should be treated with some level of skepticism, due to interaction effects between, for example, the customer's geographic region (Stockholm) and other variables. The model performs much better than simply guessing, however, some further improvements are required before the model is put into production