Churn modelling in banking

Studies across a number of industries have revealed that the cost of keeping an existing customer is more than the cost of acquiring a new one. Therefore, it is essential for a company to have a good retention strategy and apply this strategy through marketing and CRM tools. A crucial input to a good retention strategy is modeling the percentage of customers who are more prone to churn. Once these type of customers determined, then the necessary steps can be taken to prevent the churn . This thesis is aiming to provide a churn prediction model for banking sector in Turkey. It proves a bit more challenging to provide a model for banking sector as there are no contractual agreements between a customer and a bank regarding the duration of services. During the implementation of the model, we first converted the raw data into usable and meaningful form so it could be used for our analysis. Later, using data mining techniques and our model, we build a "churn prediction model" for each customer in this data set. Besides the classical datamining techniques, we also use the recently popular "Random Forest" technique in our prediction, and compared the results in terms of prediction ability. Lastly, for the purpose of improving retention performance, we propose some measures and make suggestions regarding the factors that are contributing more to churn according t our model. Keywords: Customer churn prediction, cutomer retention, CRM, data mining, decision trees, neural networks, regression, random forests