Skip to main content
Nerve Centre KPIs · Audit Profile · Sentiment Settings The questions every Snowflake operator asks at 9am on Monday: why did credit spend jump 50% week-over-week, are warehouses right-sized or sitting idle, which queries are queueing and blocking analytics dashboards, and how stale is replication lag. Cross-references ecommerce platforms for data-pipeline insights - Fivetran ingest stalled while orders flowing indicates a data warehouse bottleneck costing revenue visibility.

​
What this audit checks

​
Credit Burn & Cost

  • Credit burn +50% week-over-week without proportional query growth (runaway query / scheduled job overrun)
  • Idle warehouse credits >10% of total (auto-suspend not firing; warehouse staying warm idle)
  • Cost per query rising >25% week-over-week (query inefficiency or result-set bloat)
  • Warehouse right-sizing opportunity - sustained low utilization (<30% for >7 days; downsize candidate)

​
Performance & Queueing

  • Query latency p95 >5 seconds sustained 15m (INFORMATION_SCHEMA measurement; 45min+ latency from ACCOUNT_USAGE)
  • Query queue depth >5 queries sustained 10m per warehouse (warehouse undersized; queries backing up)
  • Slow-query rate >5% of total queries (TOTAL_ELAPSED_TIME > 5000ms in QUERY_HISTORY)
  • Query error rate spike >1% in 1h window (syntax errors, permission failures, resource exhaustion)

​
Capacity & Saturation

  • Warehouse saturation >90% (running_queries / max_concurrency_level sustained; need upsize or multi-cluster)
  • Disk usage >90% (database/schema storage approaching quota; table growth outpacing cleanup)
  • Active sessions climbing sustained (connection pool pressure from application retry loops)

​
Replication & Backup

  • Cross-account replication lag >10 seconds (DATABASE_REPLICATION_USAGE_HISTORY measurement; stale replica risk)
  • Time Travel snapshot >72 hours old (retention floor violation; recovery window shrinking)
  • Failed login attempts >10 in 24h (LOGIN_HISTORY spike; brute force or stale credentials)

​
Severity thresholds

​
Data sources

  • GET - Credit burn and warehouse cost trends (45min-3hr latency)
  • GET - Latency percentiles, queue depth, errors, elapsed time (45min-3hr latency)
  • GET - Real-time 7-day query performance; use for p95/p99 over ACCOUNT_USAGE when fresh data needed
  • GET - Per-warehouse credit consumption and idle-time detection
  • GET - Disk capacity, database-level and table-level growth trends
  • GET - Replication lag per database and target account
  • GET - Failed login detection; bursts indicate brute force or credential rotation issues