Site Reliability Engineering and Observability Maturity Assessment

About this assessment

Most outages are not exotic: a configuration push that went to every region at once, a retry storm against a slow dependency, a page that fired 40 times a night until people muted it, a runbook last touched two years ago, a postmortem whose actions never closed. AI adds new ones: an LLM provider that hits its rate limit at peak, a prompt change that quietly halves answer quality, an agent that restarts the wrong cluster, and a triage assistant whose confident guess becomes the root cause. This assessment follows the reliability practice end to end: tiers, the SRE operating and engagement model, shared ownership and skills; user-centric SLIs, SLOs, error budgets and their use in decisions; telemetry standards, end-to-end traces, debugging, burn-rate alerting, cost and sensitive data; readiness, progressive delivery, flags and configuration, change to incident linkage and AI changes; resilience patterns, capacity, load, chaos, recovery and provider fallback; detection, command, on-call health, runbooks, status, postmortems and AI-assisted triage; toil, self-service, automated remediation, AI detection and agent actions; and KPIs, action follow-through, themes, game days, the cost and value of reliability, benchmarking, prediction and improvement. The model is self-contained: it owns every document it uses, and needs no other assessment. It tracks the DORA measures that bear on stability (change fail rate, rework rate and failed deployment recovery time) and leaves throughput measures (lead time and deployment frequency) to delivery assessments.