Mobile Feature Flags and Experiments at Uber

on YouTube ↗37m

" summary

Uber runs 25,000 feature flags and ~4,000 active experiments across 10+ apps and builds an in-house library with annotation processors that auto-generate parameter APIs and CI jobs that upload a per-OS/app/version parameter directory so backend serves only ~20–25% of parameters per request. Developers get an in-app override screen, synchronous multi-layer caches, prioritized launch fetches, emergency rollback via FCM wrappers, and automated stale-flag cleanup using Piranha structural search-replace to generate diffs and remove dead code.

AI-written summary · about

" chapters (20)

1. 0:00Introduction and setup
2. 1:15Uber’s experimentation scale
3. 2:41Feature flags and parameters
4. 4:28Automated parameter creation
5. 7:35Backend parameter workflow
6. 9:43In-app parameter override
7. 11:49Automated stale parameter cleanup
8. 13:58Fetch triggers and payload optimization
9. 17:55Parameter caching strategies
10. 20:25Logging and data accuracy
11. 23:36Experiment logging and analysis
12. 26:52Emergency rollback mechanism
13. 29:01Targeted parameter rollouts
14. 30:48End-to-end observability
15. 31:52Auto rollout parameters
16. 32:57Dynamic parameter updates
17. 34:00Bundled parameter cache
18. 35:03Parameter constraints
19. 35:45Future platform improvements
20. 37:08Closing and resources