Skip to main content

Severity Classification Decision Matrix

SEV-1 — Critical​

Definition: Production is down or there is imminent risk of data loss.

Examples:

  • All gateway regions returning 5xx errors
  • Postgres database corruption detected
  • Audit-log hash-chain broken (tamper detected)
  • License expiry affecting all orgs simultaneously
  • mTLS certificate failure across all provider connections

Response:

  • Acknowledge within 15 minutes
  • All-hands war room via Slack #incidents
  • Status page update within 30 minutes
  • External notification to SOC 2 customers within 72 hours (if data breach)
  • Post-mortem within 5 business days

SEV-2 — Major Degradation​

Definition: Significant portion of service degraded but not fully down.

Examples:

  • One region out of three unavailable (SG/IN/EU failover active)
  • Latency > 5x normal for all model providers
  • Single model provider unreachable with no fallback configured
  • Cache layer (Redis) failure causing increased DB load

Response:

  • Acknowledge within 1 hour
  • On-call + backup engineer engaged
  • Status page update if customer-visible impact
  • Post-mortem within 5 business days

SEV-3 — Minor Degradation​

Definition: Non-critical issue affecting limited users or features.

Examples:

  • Single model provider failing (others operational)
  • Admin dashboard unavailable (API still functional)
  • Slow query performance on non-critical endpoints
  • Telemetry/monitoring data gap

Response:

  • Acknowledge within 4 hours during business hours
  • On-call engineer assigned
  • Ticket created in Jira with severity label
  • Root cause documented for SEV-3

SEV-4 — Cosmetic / Non-Urgent​

Definition: Non-functional issues, typos, or minor bugs.

Examples:

  • Dashboard UI spacing issue
  • Documentation typo
  • Minor UX inconsistency
  • Log message formatting

Response:

  • Next business day acknowledgment
  • Standard Jira workflow (no on-call required)
  • Root cause optional