Define testable NFRs, SLOs, KPIs, workloads, and acceptance criteria with product, architecture, engineering, and SRE teams. Build production-representative workload and capacity models using traffic patterns, transaction mix, arrival rates, concurrency, growth forecasts, and appropriate open or closed workload models.
Use APM, metrics, logs, and traces from tools such as AppDynamics, New Relic, Grafana, Prometheus, Splunk, Elastic, or OpenTelemetry to isolate application and platform bottlenecks.