Governance · 10 min
Governance & Administration
What should every admin team be able to answer about its own configuration?
Troubleshooting model
An alert did not reach the expected person.
Troubleshoot left-to-right:
- Was the alert created?
- Which integration created it?
- Which team received it?
- Which routing rule applied?
- Which escalation policy applied?
- Which schedule was used?
- Who was on-call?
- Was the notification delivered?
- Did escalation continue?
Traceability rule
Every production operational path should be explainable by an administrator without opening the configuration.
Machine-initiated operational event
Human-initiated incident
If administrators cannot explain the path, the configuration should be reviewed.
Ownership
Who owns:
- Services
- Teams
- Schedules
- Escalation policies
- Integrations
- Alert policies
- Automation
- Incident workflows
Naming standards
Apply consistent naming to
- Teams
- Services
- Schedules
- Escalation policies
- Integrations
- Automation rules
Why it matters
Naming standards improve troubleshooting and long-term maintainability. During an incident, a predictable name is the difference between finding the right escalation policy immediately and guessing.
Avoid hard-coded people
Bad pattern
- Alert
- Bob Smith
Preferred pattern
People change roles, leave organisations, go on vacation and rotate responsibilities. The operational model should not need to be redesigned every time staffing changes.
Architectural Design Principle
- Service
- Team
- Escalation
- Schedule
- People
Build the operating model around stable organizational structures. People change. The service and response model should remain understandable and maintainable.
Periodic administrative review
Monthly / quarterly review checklist
- Review noisy integrations
- Review unused integrations
- Review alert volume
- Review escalation frequency
- Confirm schedule coverage
- Review team membership
- Review services and ownership
- Review automation rules
- Review alert policies
- Review recurring incidents
- Review PIR completion
- Test critical alert paths
Common admin anti-patterns
Hard-coding individuals
Avoid using named individuals as permanent routing architecture.
One alert = one incident
Not every operational signal requires incident coordination.
Recreating every Opsgenie object
Migration does not require duplicating every legacy object.
Too many escalation policies
Reuse consistent response patterns where possible.
Automation solving architecture problems
Fix ownership, services, routing and escalation design before layering automation on top.
Alerts without service context
Operational events should increasingly be associated with understandable service ownership.
No ownership for integrations
Every production integration should have a clear technical and operational owner.