You’re probably using Lighthouse wrong
summary
Claim: Lighthouse is a useful guide but is often misused because it reports synthetic page quality not real user experience. The talk shows concrete pitfalls: Lighthouse runs in lab conditions that differ by CPU, extensions and Lighthouse version (11 versions so far), PageSpeed Insights uses an emulated Moto G4 and CrUX real-user data, tests can be gamed by sniffing the Lighthouse agent, and CI use cases include Lighthouse CI for regressions. Recommended tools and data: PageSpeed Insights, Chrome User Experience Report and Yellow Lab for actionable fixes.
AI-written summary · about
chapters (16)
- 0:00Introduction
- 2:17Lighthouse scores misleading
- 7:42Lighthouse default tool
- 9:28What Lighthouse measures
- 12:18Lighthouse like BMI
- 14:24How Lighthouse scores calculated
- 16:08Local Lighthouse tests unreliable
- 18:59PageSpeed Insights for results
- 21:25Lighthouse isn’t useless
- 23:30Lighthouse in development
- 24:35Real user vs synthetic scores
- 26:41Lab vs real-world performance
- 27:44Key takeaways
- 30:11Q&A: Score concerns
- 32:16Q&A: Lighthouse in CI
- 33:40Q&A: Fluctuating results
related talks
22mYou’re Probably Using Lighthouse Wrong
40mYou’re probably using Lighthouse wrong bt Filip Rakowski
42mYou’re probably using Lighthouse wrong
27mCommon misconceptions around Google Lighthouse
27mHow to Measure Performance Effectively?
26mDemystifying Lighthouse, CrUX, and Core Web Vitals
16mFrom beginner to expert in web performance
22mUsing Lighthouse in the Real World