Skill: post-launch-monitoring

Skill: post-launch-monitoring

Purpose

Define a structured monitoring plan for the 72 hours after a release — with clear metrics, alert thresholds, escalation paths, and a stabilization criteria.

When to Use

• During launch planning
• As part of the launch checklist
• After every production deployment

Framework: Post-Launch Monitoring Plan

Step 1 — Define Monitoring Window

• Hyper-care period: First 72 hours after launch
• Extended monitoring: Days 4–14 for gradual rollouts

Step 2 — Key Metrics to Monitor

Metric	Normal Baseline	Alert Threshold	Escalation Threshold
Error rate	< 0.5%	> 1%	> 2%
p95 Response time	< 300ms	> 500ms	> 1000ms
Throughput	1000 req/min	< 700 req/min	< 500 req/min
CPU / Memory	< 60%	> 80%	> 90%
Failed logins	< 5/min	> 20/min	> 50/min
Support tickets	Baseline	2× baseline	3× baseline

Step 3 — Monitoring Cadence

Period	Check Frequency	Who
First 2 hours	Every 15 minutes	PDM + On-call Dev
Hours 2–24	Every hour	On-call Dev
Day 2–3	Every 4 hours	On-call Dev
Day 4–14	Daily	PDM

Step 4 — Escalation Path

• Alert triggered → On-call Dev investigates (15 min)
• Cannot resolve in 15 min → Escalate to Tech Lead
• P0 confirmed → PDM notified, Incident declared
• Business impact → Stakeholder communication within 30 min

Step 5 — Rollback Decision Criteria

Initiate rollback if:

• Error rate > 2% for > 10 minutes
• Any P0 data loss or security incident
• Service unavailable for > 5 minutes

Step 6 — Stabilization Sign-Off

Release is stable when:

• 72 hours with no alerts triggered
• Error rate back to baseline
• No open P0/P1 bugs from the release
• Support tickets normalized

Output Format

1. Monitoring metrics table with thresholds
2. Monitoring cadence schedule
3. Escalation path
4. Rollback decision criteria
5. Stabilization sign-off checklist
6. Suggested next command: /lessons-learned or /delivery-report