Measuring and enforcing app health at scale
summary
Trainline runs 30–40 Android engineers across 12 teams, 3 million LOC and 580+ modules and measures app health with Play Console vitals, Crashlytics + Firebase Performance, macro/microbenchmarks and CI. Enforcement uses lint/Consist checks, detekt custom rules that fail the build for forbidden patterns, Copilot-powered PR rules and AI “vibe coding” to generate checks. A public health dashboard on GitHub Pages shows module scores; automation catches regressions and scales reviews, improving cross-team consistency and delivery predictability.
AI-written summary · about
chapters (10)
- 0:00App health at scale
- 3:19Defining app health
- 6:08User-facing health metrics
- 9:55Measuring performance
- 13:27Health before production
- 16:37Code health standards
- 19:06Golden paths in architecture
- 22:35Automating health enforcement
- 25:22Teaching health standards
- 29:39Key takeaways & Q&A
related talks
25mElevating Android App Resilience: Incident Management Strategies
43mThe Engineering Playbook for Growing Android Teams
39mMobile Developer Productivity Panel
36mApp quality on Google Play
41mBuilding Mobile Apps At Scale
26mDevelop Rock-Solid Apps with Firebase
37mTackling Android Development at Scale
47mGamifying Reduction of Technical Debt