This paper analyzes the importance of using proper techniques for differentiating mortgage customer qualities. This paper builds on the mortgage literature and compares several statistical tests, such as statistical description, discrminant analysis and logistic regression to measure customer credit risk. This study uses a unique data set with Taiwan consumer mortgage data that contains extensive financial and personal information on both good and bad customers. The results show that logistic regression model is better than the rest because its predicting ability outperforms others. This study can have potential implications for banking industry on risk management and pricing strategy.