* UML-style declaration of relationships (instead of 'has_many', 'belongs_to', etc.) * efficiency through fine control of collaboration with the DBI layer (prepare/execute, fetch into reusable memory location, etc.) * improved API for SQL::Abstract (named parameters, simplified 'orderBy') * clear conceptual distinction between - data sources (tables and views), - database statements (stateful objects representing stepwise building of an SQL query and stepwise retrieval of results), - data rows (lightweight hashrefs containing nothing but column names and values) * joins with simple syntax and possible override of default INNER JOIN/LEFT JOIN properties; instances of joins multiply inherit from their member tables * named placeholders at the ORM level * nested, cross-database transactions