An information architecture (IA) specification is given for a dutch organization, the CBR (Central Bureau for certificates of driving proficiency). A discussion of the specification process shows that other issues are taken into account than those tackled by the information planning method used. However, this is done in an informal and intuitive manner. We argue that the specification of alternative information architectures can give a more solid basis to this process. In order to generate these alternative architectures, various design strategies are needed. A generalization of design strategies can be found in cluster analysis. The application of cluster analysis calls for an appropriate problem representation. This is achieved by using a model hierarchy as a conceptual model of the data being handled in an information planning process. The model hierarchy also serves as the basis of a decision support system for information architecture specification, which is described briefly. To demonstrate the feasibility of our approach, we compare the original CBR architecture with an information architecture generated by the decision support system. It turns out that the new architecture performs better with respect to the decentralization perspective of the organization.