Business Context

NimbusOps runs a B2B workflow automation platform used by 2,400 enterprise customers. Over the last 30 days, several customers reported intermittent outages even though the engineering dashboard still showed 99.95% uptime, so leadership wants a clearer framework for measuring service availability and overall system health.

Metric Scenario

In the past month, NimbusOps processed 48M API requests, 1.8M scheduled jobs, and 620k user login sessions. There were 3 major incidents: one 22-minute API outage in us-east-1, one 47-minute database degradation that increased p95 latency from 280 ms to 2,400 ms, and one authentication issue that caused login failures to spike from 0.6% to 8.9% for 90 minutes. Customer success says 14 strategic accounts were impacted, while the current executive report only includes a single uptime number.

Requirements

1. Define the core metrics you would track for service availability and service health.
2. Distinguish between customer-facing availability metrics and internal health/early-warning metrics.
3. Explain how you would calculate each metric and what should count as downtime or degradation.
4. Propose a metric hierarchy for executives, engineering managers, and on-call teams.
5. Identify likely decomposition cuts to diagnose whether issues are isolated or systemic.

Data Available