Skip to content
Latchkey

What Is a Statistical Significance Gate?

A statistical significance gate applies a hypothesis test to a canary's metric difference and treats it as meaningful only when it passes a confidence threshold. This separates genuine regressions from the random fluctuation present in any sampled metric. The gate fails the release only when the evidence is strong enough.

Why it matters

Without significance testing, small random swings can trigger needless rollbacks and erode trust in the pipeline. The gate keeps the decision grounded in evidence rather than noise.

Related guides

Tired of flaky CI? Latchkey auto-heals failed jobs and retries them for you. Start free →