What causes Snapshot Strategy Failure?

The `unique_key` column has duplicate values in the source, violating the snapshot merge strategy. The `updated_at` column used for the timestamp strategy has NULL values. The snapshot table in the warehouse was manually altered or dropped, causing a schema mismatch on the next run. The source data contains more rows than expected and the snapshot query runs out of warehouse resources. A check strategy snapshot has too many monitored columns, causing excessive row churn and slow performance

How do I fix Snapshot Strategy Failure?

Verify the unique_key column has no duplicates in the source: `SELECT , COUNT(*) FROM GROUP BY 1 HAVING COUNT(*) > 1`.. For timestamp strategy: confirm the updated_at column is NOT NULL and is a valid timestamp.. If the snapshot table was manually modified, either recreate it by running with `--full-refresh` or align its schema to what dbt expects.. Reduce the number of columns in a check strategy snapshot by switching to a timestamp strategy if an updated_at field is available.. Add a WHERE clause to the snapshot SELECT to exclude known-duplicate rows at the source level.

High severitydbt

Power BI Refresh Error:
Snapshot Strategy Failure

What does this error mean?

A dbt snapshot failed during execution. Snapshots track slowly-changing dimensions by comparing current source data to the previously stored state. Failures occur when the unique_key or updated_at column is missing, has duplicates, or when the snapshot table schema drifts.

Common causes

1The `unique_key` column has duplicate values in the source, violating the snapshot merge strategy
2The `updated_at` column used for the timestamp strategy has NULL values
3The snapshot table in the warehouse was manually altered or dropped, causing a schema mismatch on the next run
4The source data contains more rows than expected and the snapshot query runs out of warehouse resources
5A check strategy snapshot has too many monitored columns, causing excessive row churn and slow performance

How to fix it

1Verify the unique_key column has no duplicates in the source: `SELECT <unique_key>, COUNT(*) FROM <source> GROUP BY 1 HAVING COUNT(*) > 1`.
2For timestamp strategy: confirm the updated_at column is NOT NULL and is a valid timestamp.
3If the snapshot table was manually modified, either recreate it by running with `--full-refresh` or align its schema to what dbt expects.
4Reduce the number of columns in a check strategy snapshot by switching to a timestamp strategy if an updated_at field is available.
5Add a WHERE clause to the snapshot SELECT to exclude known-duplicate rows at the source level.

Frequently asked questions

Can I recover missing snapshot history after a failure?

Partially — manually insert historical records with the correct dbt_scd_id, dbt_valid_from, and dbt_valid_to columns. Automated recovery is complex and error-prone.

When should I use the `check` strategy versus the `timestamp` strategy for snapshots?

Use `timestamp` when the source has a reliable updated_at — dbt compares only timestamps, so it's faster. Use `check` for sources without an updated timestamp; it's slower since every monitored column is compared row by row.

Does `dbt snapshot --full-refresh` delete my SCD Type 2 history?

Yes — `--full-refresh` drops the snapshot table and rebuilds from current source data. All SCD Type 2 history is permanently lost — only use it if history is no longer needed or you have a backup.

Official documentation: https://docs.getdbt.com/docs/build/snapshots