Amazon Quick Sight now supports seven specific data modeling patterns for Multi-Dataset Relationships, enabling users to combine datasets at runtime rather than relying on pre-flattened tables. Supported schemas include the simple star schema, which relates a central fact table to multiple dimension tables, and the snowflake schema, which uses multi-level chains for normalized data. For complex environments, the platform supports multi-fact tables that share conformed dimensions, allowing users to compare different business processes, such as sales versus returns, through a common bridge.

Developers can also utilize role-playing dimensions, where one physical table serves multiple analytical functions, such as an order table linking to separate 'OrderDate', 'ShipDate', and 'DeliveryDate' roles. The platform also handles multi-fact tables with varying grains, automatically aggregating finer data to match coarser levels during runtime joins. These multi-dataset configurations maintain independent refresh schedules, allowing high-velocity transaction facts to update hourly while slower dimensions update weekly.

Runtime row-level security (RLS) is automatically enforced before joins, ensuring that users only access permitted records based on policies applied to each individual dataset. While the current release is limited to inner joins and restricts circular relationships, users can address complex scenarios by denormalizing redundant paths or flattening recursive hierarchies before importing them as datasets. The architecture supports up to 12 datasets per Topic and requires consistent use of either SPICE (in-memory engine) or Direct Query mode across all related tables. These patterns shift data modeling from pre-processing to dynamic assembly, facilitating more flexible interactions across visuals, calculations, and generative AI interfaces.

Amazon Quick Sight now supports seven specific data modeling patterns for Multi-Dataset Relationships, enabling users to combine datasets at runtime rather than relying on pre-flattened tables. Supported schemas include the simple star schema, which relates a central fact table to multiple dimension tables, and the snowflake schema, which uses multi-level chains for normalized data. For complex environments, the platform supports multi-fact tables that share conformed dimensions, allowing users to compare different business processes, such as sales versus returns, through a common bridge.

Developers can also utilize role-playing dimensions, where one physical table serves multiple analytical functions, such as an order table linking to separate 'OrderDate', 'ShipDate', and 'DeliveryDate' roles. The platform also handles multi-fact tables with varying grains, automatically aggregating finer data to match coarser levels during runtime joins. These multi-dataset configurations maintain independent refresh schedules, allowing high-velocity transaction facts to update hourly while slower dimensions update weekly.

Runtime row-level security (RLS) is automatically enforced before joins, ensuring that users only access permitted records based on policies applied to each individual dataset. While the current release is limited to inner joins and restricts circular relationships, users can address complex scenarios by denormalizing redundant paths or flattening recursive hierarchies before importing them as datasets. The architecture supports up to 12 datasets per Topic and requires consistent use of either SPICE (in-memory engine) or Direct Query mode across all related tables. These patterns shift data modeling from pre-processing to dynamic assembly, facilitating more flexible interactions across visuals, calculations, and generative AI interfaces.