Data modeling best practices for Amazon Quick Sight multi-dataset relationships

This article introduces Multi-Dataset Relationships in Amazon QuickSight, a new capability that enables runtime joins across multiple datasets defined within a Topic, eliminating the need for pre-joined denormalized datasets.

• Define logical relationships between QuickSight datasets and perform runtime joins at query time instead of pre-flattening tables
• Reduces upfront data preparation by keeping each table as its own dataset with declared relationships
• Preserves native granularity of each dataset, avoiding measure duplication across different grains
• Enables reuse across multiple analytical use cases with a single Topic containing defined relationships
• Supports independent refresh schedules per table and row-level security enforcement at runtime
• Uses two modeling layers: physical layer (within datasets) and logical layer (across datasets in Topics)
• Recommends star schema as foundation with dimension tables radiating from central fact table
• Best practices include clean join keys, deliberate granularity management, and rich metadata enrichment for AI accuracy
• Supports star, snowflake, and galaxy/constellation schema patterns with inner join semantics

Multi-Dataset Relationships simplify analytics by enabling flexible, relationship-based data modeling that reduces preparation overhead and improves governance while maintaining data integrity.