RiverBase Software Advisory
When a ‘successful’ extract loads zero rows
Why ETL jobs that exit cleanly with empty extracts hurt month-end reporting, and how staging checks catch them before the warehouse load.
An extract job that finishes with exit code zero and zero rows is easy to miss. Orchestrators paint the task green. The warehouse load that follows may truncate a target table and insert nothing, or skip the insert and leave yesterday’s figures looking current.
What we look for in audits
During Pipeline Health Audits we ask three questions of every extract:
- Does the job distinguish “source truly empty” from “path missing / credential expired / query timed out”?
- Is there a minimum row-count threshold tied to day-of-week patterns?
- Does staging retain the empty file or result set long enough for a human to inspect?
Teams that only alert on hard failures discover empty loads at month-end when finance reconciles.
A practical threshold pattern
For daily retail sales extracts, we often set a soft alert below the 10th percentile of the prior eight weeks for that weekday, and a hard fail below a much lower floor that still allows public holidays. The numbers belong to the business; the pattern belongs in the job.
What to do next
If your green jobs sometimes deliver empty facts, request a Pipeline Health Audit or describe the job family on our contact form.