Advanced Feature Flagging: It's All About The Data
summary
Feature flags combine with telemetry, attribution and a stats engine to become a lightweight experimentation platform: user IDs are hashed to buckets 0–99 so assignments are both random and sticky, telemetry must include that key for attribution, and a t‑test compares distributions. Start at employees then 1–2% or 5% ramp, use 50/50 for power, and auto‑calculate do‑no‑harm metrics to detect regressions. Large orgs (Walmart, LinkedIn, Booking) wire alerts to detect and shut off in one second when anomalies appear.
AI-written summary · about
chapters (6)
- 0:00Introduction
- 3:00How You Measure Matters
- 4:42Randomized Controlled Trial
- 7:33Components Guidelines
- 11:14Stats Engine
- 12:59Examples
related talks
27mJust Enough Feature Flagging
05mFeature Flags
05m5 Life Hacks Using Feature Management
26mBuild Progressive Delivery and Safety Into Your Apps with Feature Flags
59mFaster & less risky releases with feature flags
15mFeature Flags
38mFun with Flags: How OpenFeature Solves Your Feature Flag Headaches
41mFeature Flags: A Boon or a Brake for a Team