Pick KPI tables first
Start with marts leadership opens every Monday — not every staging model.
Learn · Reliability
Data freshness monitoring
How to know a table went quiet before a dashboard lies.
The problem
Dashboards still render when the source table stops updating. The failure is quiet — until a meeting goes sideways.
The outcome
Set per-table freshness SLAs on BigQuery last-modified metadata, alert once when lag crosses the line, and show blast radius to downstream KPIs.
SLA hours
You define per table
Metadata
INFORMATION_SCHEMA
1 alert
Per root cause
What you get
Set warning and critical
e.g. warn at 12h, critical at 24h — numbers you can put in a runbook.
Group by root cause
One stale source should not page for every child dashboard.
Prove it in demo
Replay a 3-day-stale orders table in /app.
- ✓Product: /vectordata/freshness
- ✓Learn also: /learn/choosing-tables-to-monitor
- ✓Security: metadata only — /security
- ✓Plans from $99/mo
Try VectorData, then decide
Open workspace, or book a Data Stack Health Audit credited toward Managed.