Canary analysis
Releasing a change to a small fraction of traffic first and comparing its metrics, error rate, latency, business KPIs, against the unchanged baseline before rolling out further. Canary analysis turns deployment into a controlled experiment: if the canary's signals degrade against the baseline, the rollout halts automatically and the blast radius stays tiny.
Canary analysis is what turns a canary release from a hopeful gesture into a decision. By routing a small slice of traffic to the new version and statistically comparing its error rates, latency, and key metrics against the stable version, it catches regressions before a full rollout, and automates the go/no-go so the call isn't a human eyeballing a dashboard at 2am. Its accuracy depends entirely on having the right metrics instrumented up front.