Data models should be filtered and normalized. This is particularly important if datasets are sourced from high data volume repositories such as enterprise data warehouses.