ETL Testing: Catching Data Integrity Issues Before Production

Data problems are quiet. A job finishes, the dashboard loads, and nobody notices that a few thousand records were dropped or truncated along the way.

Early in my career I built an automated test suite that compared data across Oracle and PostgreSQL hosts to catch exactly these problems. The lessons still apply to modern pipelines.

What to test

  • Row counts at every stage
  • Field-level comparisons between source and target
  • Differences in SQL dialects and data types
  • Checksums on files in transit
  • Reconciliation reports that people actually read
Scroll to Top