You're gradually rolling out a new feature to your users. But what's actually happening? Are users in region A seeing faster checkouts while region B is quietly timing out? Is version B of your pricing page converting better, or are error rates creeping up in ways your dashboard doesn't show? Most teams can tell you a flag is on. Fewer can tell you what it's actually doing to their system and users.

OpenFeature gives you a vendor-agnostic way to dynamic feature flag evaluation. OpenTelemetry gives you vendor-agnostic observability. Both are CNCF projects, and when you wire them together, flag evaluations show up in your traces, metrics, and events. You can go from "it's live" to "here's the P95 latency impact per region, per variant, per user cohort".

This talk walks through that setup end to end: instrument an app, roll out a feature, then use the telemetry to understand the real impact. We'll also get into traces vs. metrics vs. logs. They each have a sweet spot, and picking wrong costs you either in cardinality or in blind spots.