Why are my freshness checks always late?

If your freshness checks are “always late,” it usually means the checks are tied to a schedule or trigger that can’t keep up with reality. Common causes:

• They run on a fixed interval
  If you check every 15 minutes, anything that changes just after a check can appear stale until the next one.

• Clock or timezone mismatch
  Your source system, scheduler, and monitoring system may not agree on time. Even a small drift can make checks look late.

• Queue or job backlog
  The freshness task may be waiting behind other jobs, so it starts later than planned.

• Slow data arrival upstream
  The check may be correct, but the data itself is landing later than expected.

• Window too tight
  If the freshness threshold is shorter than the real-world latency of the pipeline, alerts will fire “late” or constantly.

• Dependency issues
  If freshness is measured only after downstream steps finish, any upstream delay pushes the check back.

• Sensing the wrong timestamp
  Sometimes checks use load time, event time, or partition time incorrectly, which makes the freshness age calculation misleading.

What to do:

1. Compare scheduled run time vs actual run time.
2. Check source timestamp, ingestion time, and alert evaluation time.
3. Look for scheduler delays/backlogs.
4. Increase the freshness threshold if the pipeline naturally lags.
5. Ensure all systems use the same timezone and synchronized clocks.

If you want, I can help you diagnose it for your specific setup—e.g. dbt freshness, Airflow, SQL checks, or a custom monitor.