This dissertation describes a tool for generate benchmark data to test schema matching algorithms based on the creation of a set of conceptual schemas, populated with data. The tool, simply called Bench, allows importing any data set with a well-defined conceptual schema and offers facilities to generate variations of the schema through transformations that reflect structural alternatives found in typical conceptual modeling. Such transformations in fact define reference alignments between the original schema and its variations. Bench also permits evaluating the performance of the schema matching algorithm submitted for testing by comparing the reference alignments with those the algorithm under testing is able to find.