Despite the manifold utilities of monitoring credit default rates, little attention is usually devoted to the underlying default definition. This paper proposes working simultaneously with different default severities, related to several past-due ranges, by means of transition matrices (to be named default matrices). In this way, default, as well as recovery, is depicted in a multidimensional flow, with the purpose of avoiding missing relevant information. The challenge lies on performing comparisons between default matrices, for which new metrics are proposed. In this paper, default matrices are built to measure consumer credit delinquency at four large Brazilian banks, allowing a detailed comparison of their credit migration experiences. The study is also able to draw relevant information from comparisons between estimations techniques (discrete and survival approaches) and between default criteria.