Problem context
- Severity models often rely on informal tribal knowledge instead of documented criteria.
- Escalation timing becomes inconsistent when response windows are not tied to severity class.
- Post-incident reviews are weak when override reasons are not captured in a structured way.