Snowflake Row-Volume Anomaly Watch

Detects silent data-volume anomalies that freshness checks miss. For each watched Snowflake table it counts rows loaded in the last 24 hours, compares that to the median of the prior 14 daily loads, and flags any table that loaded far more or far fewer rows than usual, which usually signals a duplicated load, a partial extract, or a dropped source.

When a table updates on time but with the wrong amount of data, like a partner feed that delivered half its rows or a job that double-inserted. Freshness looks green while the numbers are quietly broken. Use this to catch volume drift before it reaches reports.

1. 1A daily schedule fires once loads complete.
2. 2The flow queries Snowflake for last-24h row counts per watched table.
3. 3It pulls the trailing 14-day daily counts and computes the median baseline.
4. 4A logic step flags tables where today's count deviates beyond the configured percent band.
5. 5Flagged tables post to Slack with today's count, the baseline, and the percentage delta so an analyst can confirm or dismiss.