If you use dbt, you probably have some tests. A few not_null checks, maybe unique on your primary keys, possibly some accepted_values on status columns. But be honest: how many of your columns actually have tests? 10%? 20%? The rest are untested. Not because you don't care, but because writing test YAML for 200 columns across 40 models is tedious work that never makes it to the top of the sprint. The gap in dbt testing dbt tests are rule-based. You write a rule, it checks that rule. If you didn't write a rule, nothing gets checked. This creates three blind spots: 1. Drift goes undetected. Your email column had 0.1% nulls last month. Today it's 12%. No dbt test catches this because you never wrote one that says "null rate should stay below X%." You find out when a PM asks why the marketing numbers look off. 2. Structural changes slip through. A column gets dropped upstream. A type changes from integer to text. dbt won't tell you unless you wrote a test for that specific column.…