How we handle incidents

Incidents are inevitable; speed, clarity, and learning quality define response quality.

Declaration

Declare when there is customer impact, degraded service, or high likelihood of impact. When in doubt, declare early.

Response model

• Use incident channel as source of truth.
• Assign explicit ownership for coordination and customer communication.
• Triage fast and switch to incident mode immediately.
• Use huddles when text is insufficient.
• Prioritize mitigation before deep root-cause analysis.
• Communicate clearly through status page/direct channels.

Closing criteria

• User-visible behavior is healthy again
• Queues or retries (where applicable) are drained or intentionally controlled
• Alerts or synthetic checks return green
• Status communications are updated or explicitly concluded