slowly-changing-dimensions

Slowly changing dimensions in dbt are tracked using snapshots to capture attribute changes over time, preserving historical context for accurate point-in-time analytics. You configure dbt snapshots to detect source data changes and apply SCD Type 1, 2, or 3 patterns accordingly.

SCD Type 1 overwrites current values for corrections, Type 2 adds a new row for each change to preserve full history, and Type 3 tracks one prior value for quick comparisons. Choosing the right pattern depends on source data characteristics and your analytical requirements.

Point-in-time joins are implemented by querying dbt snapshot outputs, which contain start and end timestamps for each historical record. These timestamps allow downstream analytics to join fact data with the correct dimensional attributes active at any specific historical moment.

Yes, dbt snapshots maintain both current and historical records by automatically tracking effective dating. Validation and testing guardrails ensure correct segregation of current versus historical rows, enabling accurate reporting on current states and past changes.

Use SCD Type 2 when you need to preserve a full history of changes for auditing or point-in-time analysis, such as tracking customer attributes or pricing history. Use Type 1 when you only need the most recent corrected value and do not require historical tracking.

Snapshot configurations require validation tests to ensure historical tracking accuracy and prevent data gaps. Guardrails verify that snapshot updates correctly capture changes, maintaining the integrity of current versus historical records for downstream models.