The ability to predict small businesses' future loan performance based on submitted loan applications is crucial for Indonesian rural banks. The small capacity of these particular banks requires an efficient approach to extract knowledge from structured (quantitative) and unstructured (qualitative) type of credit information. The eXtensible Markup Language (XML) is used to organize this complementary credit data from an Indonesian rural bank. The credit performance evaluation application presented utilizes a mapping approach to preserve structural aspects of data within a format on which wider selections of data mining techniques are applied. Results from decision tree and association rule mining algorithms demonstrate the potential of the approach to generate reliable and valid patterns useful for evaluation of existing lending policy.