Data mining very large commercial databases requires a careful consideration of both data management issues and data analysis issues. A multi-strategy approach, called the Hot Spots Methodology and involving clustering and rule induction, has been developed to tackle the analysis of giga-byte sized datasets. This has been coupled with the development of a Java-based system for access to collaborator data through the JDBC database- independent API and for the integration of a number of external modelling tools. This seminar will illustrate the hot spots methodology through actual real world case studies and will highlight the advantages of a Java-based system.