Credit granting institutions have to face the risk that some amount of money that has been lent will not be repaid. Therefore banks and other lending institutions are concerned with the issue of measuring the risk of each credit default, which is needed to accept or reject credit applications. The aim of this paper is to present an approach for classifying applicants, based on duration data - survival analysis. Including an extra dimension - time - in the analysis allows the lending institution to take into account the profitability of a loan. Moreover, failure models give an opportunity to include in the model time-varying covariates - regressors that change in time of the repayment. They make it possible to use in the estimation data about the credits that are still being repayed - censored durations. This method is also a useful tool to predict the influence of the particular characteristics on the probability and the expected time of exit to different kinds of states - complete repayment on time, default but also e.g. an early repayment.