The statement is correct. In SQL, and by extension in dbt which compiles to SQL, UNION ALL is significantly more performant than UNION. The UNION operator combines the result sets of two or more SELECT statements and then performs a distinct operation to remove any duplicate rows. This de-duplication step is computationally expensive as it often requires sorting or hashing the entire combined dataset. In contrast, UNION ALL simply concatenates the result sets, including all duplicate rows, without performing the costly de-duplication step. Therefore, if duplicate rows are acceptable or if the datasets are known to be mutually exclusive, using UNION ALL will result in a faster and more efficient query.

The "False" option is incorrect because it ignores the fundamental operational difference between the two operators. UNION performs two operations: concatenation and de-duplication. UNION ALL performs only one: concatenation. The additional de-duplication step makes UNION inherently less performant.

1. dbt Labs Official Documentation (SQL Style Guide): The official dbt style guide explicitly recommends this for performance reasons.

Source: dbt Labs

"SQL Style Guide"

Reference: In the "Be specific" section

the guide states: "Prefer union all to union. The union command performs a union all then a distinct. If you don't need to remove duplicate records

union all is much more performant."

Link: https://docs.getdbt.com/guides/best-practices/how-we-style-our-dbt-projects/3-be-specific#prefer-union-all-to-union

2. dbt Labs Official Documentation (dbt-utils package): The widely used unionrelations macro within the dbt-utils package defaults to using UNION ALL

reinforcing this as a best practice for performance.

Source: dbt Labs

dbt-utils Package Documentation

unionrelations

Reference: The macro is designed to "stack a list of relations on top of each other" and by default

it generates a series of select ... from ... union all statements. This design choice prioritizes performance.

Link: https://github.com/dbt-labs/dbt-utils#unionrelations-source

3. Vendor Documentation (Google Cloud for BigQuery): The documentation for major data warehouses

which execute the SQL generated by dbt

confirms this performance principle.

Source: Google Cloud

"BigQuery optimized SQL patterns: part 1"

Reference: In the section on "Set operators"

the documentation advises: "If your data doesn't have duplicates or you don't need to eliminate them

use UNION ALL for better performance. BigQuery must do extra work to remove duplicate rows for UNION DISTINCT (or UNION)."

Link: https://cloud.google.com/blog/products/bigquery/bigquery-optimized-sql-patterns-part-1

4. Vendor Documentation (Snowflake): Snowflake's documentation also highlights the performance advantage of UNION ALL.

Source: Snowflake Documentation

"UNION & UNION ALL"

Reference: The documentation notes: "UNION ALL is much faster than UNION because UNION ALL does not search for and remove duplicate rows."

Link: https://docs.snowflake.com/en/sql-reference/operators-set#union-union-all