In general insurance, the expected duration of a customer relationship is of great interest since retaining an existing customer is much cheaper than acquiring a new one from an administrative point of view. The expected acquiring cost of a customer can be spread over all the years a customer is expected to stay as a customer in the insurance company. Hence, a customer with higher expected duration is cheaper for the insurance company than a customer with lower expected duration. The problem with analysing the expected duration of a customer relationship is the incomplete data. For previous customers the duration of the customer relationship is known, but for the existing customers only the entry date is known. Because the termination date is still unknown, or censored, it is not possible to compare these two groups without adjustments being made. Event history analysis is specially made to handle this problem. It is also of interest for an insurance company to evaluate the effects of the variables on the customers. Cox regression is a much applied method for situations like these. This feature enables prioritization of different customers. Such a regression is of interest in itself. But if the expected duration of a customer relationship can be connected to whether or not that customer is profitable, management can set apart the desired customers from the not so desired ones. The profitability model can be developed from a logistic regression, where a fair profitability criterion is applied. In this master's thesis, a Cox regression model is developed for the customer duration and a logistic regression model for the customer profitability. In order for the insurance company to prioritize the different kinds of customers, the models are combined and a combined score is suggested. The results showed a poor customer duration model. Fortunately, the customer profitability model was better. Furthermore, there seemed to be a relation between the models. The suggested combined model was also developed from logistic regression. Due to the poor customer duration model, the combined model was dragged down. All in all, the results could have been better.