Large dbt sources that nothing references

This rule flags a declared dbt source holding at least 10% of the storage catalog.json measured, when nothing in the project references it. That means no model, exposure, metric or semantic model. At 25% or more we rate it high severity. The rule declines when fewer than 10 relations were measured.

Its ingestion and storage are paid for on every cycle, whether anything reads it or not. If a tool outside dbt reads it, that dependency is undeclared. The next person to change the ingest has no way to find it.

Settle what reads the largest unreferenced sources. Confirm what reads the source. Declare the model or exposure that does, and stop ingesting the table if nothing does.

A raw table kept for audit, replay or a retention period is read rarely by design. So is a landing table another tool reads directly. Tell us which applies, and the finding comes out of your report.

manifest.json and catalog.json, which dbt docs generate writes. The adapter must report sizes or row counts: Snowflake and BigQuery do, DuckDB does not. The finding is medium confidence, because a reader outside dbt appears in neither file.

dbt_project_evaluator, rule fct_unused_sources. Ours checks a subset of what it flags; theirs is stricter. dbt Labs' rule flags every source nothing references. This one flags only those large enough to be a material share of the storage the catalog measured, and only where the adapter reported sizes, so theirs finds strictly more.

This means our check corresponds to their rule. It does not mean the publisher reviewed or endorses it.