An extract can have one row per person, one row per assignment, or one row per event. Those grains produce different counts. Inspect keys before joining: a repeated employee id in the assignment table can be legitimate, but it duplicates a person when you count joined rows. A missing exit reason is a quality issue only for someone who left. An active employee usually has no exit reason. SQL COUNT(column) skips missing values, while COUNT(*) counts rows. Keep the problem rows available for review rather than silently replacing them. These checks identify records to investigate; they do not prove what the correct value should be.
Compare typed date columns directly and retain the affected records for investigation.
Compare typed date columns directly and retain the affected records for investigation.