Abstract: With information infrastructures getting more and more complex, it becomes necessary to give automated support for managing the evolution of the infrastructure. If changes are detected in a single system, the potential impact on other systems has to be calculated and appropriate countermeasures have to be initiated to prevent failures and data corruption that span several systems. This is the goal of Caro, our approach to change impact analysis in large information infrastructures. In this paper we present how we model the metadata of information systems to make a global analysis possible, regardless of the data models used, and how the analysis process works.