In most commercial enterprises, information is scattered across a large number of (legacy) data stores. Moreover, it is nearly impossible to obtain funding to replace these data stores with a single integrated database, given tight budgets, time-to-market pressures and past failures of various kinds of large-scale efforts. In this paper, we advocate an approach which hinges on (1) a high-level enterprise information architecture, (2) the identification of the stable elements in this information architecture - essentially identifiers of some core entity types, the sets of legal values of some descriptive attributes - and (3) a central metadata and re f- erence data repository which supports the tracking of the history of refe r- ence data.