Where Do Your Web Performance Metrics Come From? – Matt Zeunert
summary
Different testing tools implement network throttling differently, so identical latency/bandwidth settings still yield different metrics. Packet-level throttling (WebPageTest, DebugBear) slows TCP/DNS/SSL at the OS level and preserves request priorities; simulated throttling (Lighthouse/PageSpeed Insights) models dependency graphs and can inflate LCP (example: 3.4s measured → 21s simulated); Chrome DevTools applies a simple minimum request delay and can report much slower LCPs (example: 30s vs packet 10s). Pick packet-level for accuracy and interpret others with their limitations in mind.
AI-written summary · about
chapters (8)
- 0:00Web performance metrics intro
- 1:06Content load waterfall
- 3:13Collecting performance data
- 5:38Synthetic tests and throttling
- 7:21Lighthouse throttling
- 9:25Throttling limitations
- 10:27DevTools throttling drawbacks
- 11:29Future DevTools throttling
related talks
55mI Feel the Need for Web Page Speed
30mMeasuring Web Performance - Metrics & Tools
49mDemystifying web performance tooling and metrics
35mImprove Your Web Application Using Progressive Web Metrics
18mWebsite performance from user perspective
33mDemystifying Web Performance Tooling
55mLet's make the web faster - tips from trenches @ Google
21mState of the Union for Speed Tooling