In this article we discuss an approach to database optimisation in which a\nconceptual schema is optimised by applying a sequence of transformations. By\nperforming these optimisations on the conceptual schema, a large part of the\ndatabase optimisation can be done before actually sliding down the software\ndevelopment waterfall. When optimising schemas, one would like to preserve some\nlevel of equivalence between the schemas before and after a transformation. We\ndistinguish between two classes of equivalence, one based on the mathematical\nsemantics of the conceptual schemas, and one on conceptual preference by\nhumans. As a medium for the schema transformations we use the universe of all\n(correct) conceptual schemas. A schema transformation process can then be seen\nas a journey (a schema-time worm) within this universe. The underlying theory\nis conveyed intuitively with sample transformations, and formalised within the\nframework of Object-Role Modelling. A metalanguage is introduced for the\nspecification of transformations, and more importantly their semantics. While\nthe discussion focusses on the data perspective, the approach has a high level\nof generality and is extensible to process and behaviour perspectives.\n