Chasing performance issues methodically
summary
Shows a practical methodology: add production observability with Firebase Analytics and BigQuery, define app launch spans and mark impacted sessions >5s, selectively enable tracing via Firebase user properties and upload traces with debug API, import flame graphs into Android Studio profiler, and prioritize fixes by measured gain versus effort. Uses a staging quality gate and replayed network to reduce noise, and reports concrete wins — simple changes (Dagger lazy, move SDKs to background, DexGuard exclusions) yield 8–10% launch improvement using mobile.dev and Firebase + BigQuery.
AI-written summary · about
chapters (13)
- 0:00Introduction and context
- 1:00Pitfalls in performance engineering
- 3:30Challenges with platform teams
- 6:35Identifying the right metrics
- 8:22Production observability setup
- 11:35Creating a baseline for fixes
- 13:19Handling benchmarking noise
- 17:51Quality gates and regression alerts
- 19:16Extracting impacted sessions
- 21:03Segmenting users for tracing
- 23:52Analyzing production traces
- 25:58Fixing identified issues
- 27:40Conclusion and revisiting patterns
related talks
39mPerformance for Conversion! Use Distributed Tracing to Pinpoint Bottlenecks
35mIn Search of Performance: the Subtle Art of Android Profiling
43mPerformance in a Kotlin World
44mEfficient Android app monitoring
20mOptimize app health with Firebase Performance Monitoring and Crashlytics
30mLearn from developers: Lyft and Google Maps share their experiences optimizing app performance
24mMake the Most of Android Vitals
29mCrash-reporting taken to next-level