Tokopedia Android Engineering 2020 Review
Article Summary
Rico Harisin from Tokopedia reveals how their Android team shipped weekly releases while building a complete performance infrastructure from scratch. Their secret? Treating CI/CD like a product.
Tokopedia's Android Engineering team spent 2020 'strengthening the core' by building custom tooling for performance tracking, release governance, and data validation. Working remotely during the pandemic, they created an entire suite of in-house tools to support weekly releases at scale.
Key Takeaways
- Built PLT (Page Load Time) metrics with three custom tools: Ch'en, Sherlock, and RUM
- Pre-merge checks validate unit tests, builds, app size, and code risk automatically
- Cassava and TopAds instrumentation tests prevent data tracking regressions in production
- Unit test dashboard and build monitoring track coverage and CI/CD health across squads
Tokopedia created a comprehensive performance and quality infrastructure that enables confident weekly Android releases through automated validation at every stage.
About This Article
Tokopedia's Android team only had basic memory consumption measurements per activity. They couldn't track performance metrics across versions and branches, so regressions went unnoticed.
Rico Harisin's team built three internal tools to fix this. Ch'en detects regressions, Sherlock provides performance insights, and the RUM dashboard tracks trends across versions.
The team defined Page Load Time (PLT) as their standard metric, measuring the time from when a page opens until the first information appears. This gave them the data they needed to optimize performance across the codebase.