To become usable for documentation and visualisation purposes, the conceptual schema of a large information system (IS) has to be clustered. [1] But most of the "clustering", "abstraction", or "concentration" rules proposed in [1]-[6] are substantiated only intuitively and have not been validated formally. As a consequence, the application of these rules to large IS schemata leads to inconsistencies (e.g. cyclic references), unnecessary loss of information (e.g. arbitrary object type subsets), and/or impractical results (e.g. combinatorical explosion). In this paper, the NF2 relational model [7] is used to validate the application of clustering operations to conceptual schemata. Based on this validation concept and a critical review of schema clustering literature, some validated, general clustering rules are presented. The schema clustering concept is extended to the clustering of textual documentations. To prove the concept's feasibility, conceptual schemata and textual documentations of SAP's R/3 modules "Production Planning" and "Sales & Distribution" have been clustered