From “What Broke?” To “What Changed?”: Unlocking Feature Flag Observability
Talkat OpenFeature Summit NA 2025 · on Monday, 2025-11-10· at 15:25· Building B | Level 2 | B213-214·
summary
Argues that adding feature flag context into observability cuts Mean Time to Recovery by turning “what broke” into “what changed,” using OpenFeature hooks to emit flag evaluations into OpenTelemetry so traces can be sliced by trace-level observability and variants compared. Shows the OpenTelemetry Demo e‑commerce stack, describes fields like flag set ID, feature flag context ID and version, demonstrates progressive rollouts and near‑instant rollbacks, and warns about pitfalls like a customer doing 4,000 flag evaluations per request.
AI-written summary · about
chapters (18)
- 0:00Feature flag observability intro
- 1:50Impact of flag changes
- 3:13OpenTelemetry demo example
- 5:48Observability gap in monitoring
- 6:33Level 1 flag change visibility
- 7:16Level 2 manual event mapping
- 8:20Level 3 automatic event mapping
- 9:26Level 4 trace-level observability
- 11:36Trace-level granular insights
- 12:16Open Feature & OpenTelemetry standards
- 13:21Semantic conventions for flags
- 15:27Advanced flag attributes
- 16:50Standards in practice
- 17:33Why standards matter
- 18:14Progressive delivery benefits
- 20:19Observability rollout benefits
- 21:43Challenges & future opportunities
- 22:45Call to action
related talks
39mFun with Flags: How OpenFeature Solves Your Feature Flag Heada...
29mCooking up o11y w/ Feature Flagging
33mSupercharging Observability with Feature Flagging
26mBuild Progressive Delivery and Safety Into Your Apps with Feature Flags
30mA Tale of Two Feature Flags
24mOpenFeature’s Positive Impact on Confidence
11mFeature flag use cases you haven't heard about yet
50mThe Story of One Feature Flag