Building a Foundation for Performance
summary
Pinterest serves ~250 million MAU and >1M requests/sec and builds a performance stack that proves results: synthetic regression tests plus RUM, dashboards and experiments tie performance to business metrics, and automated tooling catches regressions. Key specifics: pwt as a custom metric, blocking regression alerts at thresholds (≈250ms/500ms/1s), binary-search bisecting with perf detective, deployment + test fleet called perf watch, experiments that drove ~20% home‑feed PWT reduction and 25% profile PWT.
AI-written summary · about
chapters (17)
- 0:00Intro
- 0:11Building a Foundation for Performance
- 1:33Industry numbers
- 3:43Web Android
- 4:35Ownership
- 5:50Key Performance Metric
- 10:36Synthetic Tests
- 10:56Real User Metrics
- 12:11Experiment Dashboard
- 12:55Ad-Hoc Queries
- 14:20Performance Data
- 17:15Regression Testing
- 19:34Detecting a Regression
- 21:04Investigating a Regression
- 28:55Optimization Strategy
- 34:29Education
- 34:54Documentation
