Amazon QuickSight multi-dataset Topics unify up to 12 datasets in one semantic layer

Amazon QuickSight has expanded its semantic layer capabilities with the public preview of multi-dataset Topics. Previously, each Topic was tied to a single denormalized dataset, requiring joins to be pre-computed. This limited flexibility for complex, normalized data sources. Now, users can add up to 12 datasets to a single Topic and define explicit relationships between them. The QuickSight chat agent automatically interprets natural language queries, identifies relevant columns across datasets, and constructs appropriate SQL joins on the fly. This means business users can ask questions like 'Show sales by region and inventory levels' without understanding the underlying schema or needing IT to pre-join tables. The feature supports SPICE datasets as well as Direct Query to Amazon Redshift, Athena, S3 Tables, Snowflake, and Databricks, though SPICE and Direct Query cannot be mixed in one topic.

Each dataset within a multi-dataset Topic retains its own independent enrichment—column descriptions, synonyms, semantic types, calculated fields, and exclusions—which improves natural language query accuracy. This metadata can be defined in QuickSight or imported from AWS Glue Data Catalog or Databricks Unity Catalog. The unified semantic layer centralizes governance while allowing normalized data to stay in its source systems. Use cases include retail analytics (e.g., joining sales, inventory, customer data), HR headcount analysis, and any scenario requiring cross-table queries without data duplication. By eliminating the need for denormalization, QuickSight reduces data preparation complexity and latency. The same multi-dataset Topic can power both chat-based Q&A and traditional analysis building. This evolution marks a significant step toward self-service analytics for enterprises with complex data models.