Amazon QuickSight launches runtime joins to eliminate pre-flattening datasets

Business intelligence analysts traditionally had to pre-join every table into a single, denormalized dataset before building visuals in Amazon QuickSight. That meant deciding the join shape before knowing every question, duplicating measures when fact and dimension tables had different grains, and creating a separate flat dataset for each reporting scenario — a maintenance headache. AWS has now shipped Multi-Dataset Relationships, a paradigm shift that lets you keep each table as its own dataset inside a QuickSight Topic and define logical relationships between them. At query time, QuickSight dynamically assembles the exact join needed for a visual, calculated field, filter, or natural-language Q&A, referencing only the tables actually required.

The runtime approach delivers several concrete advantages. Upfront data preparation drops to a single definition of relationships — no more custom SQL or database views for every use case. Because each dataset maintains its own level of detail, measures no longer inflate when joined across grains. A single Topic can serve dozens of analytical use cases without rebuilding datasets, and governance becomes simpler: permissions, transformations, and business logic live at the individual dataset level. Crucially, you can set independent refresh schedules per table (e.g., hourly for orders, daily for product catalog) based on data volatility, and row-level security rules are enforced during runtime joins, ensuring consistent data access regardless of how tables are combined. The architecture separates the physical layer (merging tables inside a dataset) from the logical layer (relating datasets in a Topic), giving analysts a clean, reusable data model that adapts to questions as they arise.