You’re Probably Using Lighthouse Wrong
summary
Claim: Lighthouse is a useful guide but is widely misused — running audits in local Chrome DevTools yields inconsistent scores influenced by CPU, network and extensions, and the scoring algorithm changes over time. Practical tools and numbers given: use PageSpeed Insights (remote audits plus 30‑day CrUX real‑user data) and enforce checks via Lighthouse CI. Pitfalls demonstrated: you can detect Lighthouse user agent and strip JS to fake 100s; a synthetic 81 score can still fail Core Web Vitals for real users. Amazon’s rule (100 ms ≈ 1% revenue) motivates using real metrics.
AI-written summary · about
chapters (7)
- 0:00You are using Lighthouse wrong
- 10:00How Lighthouse works?
- 10:59It matters A LOT where you run the audit
- 14:30Be suspicious about Lighthouse score - it's easy to trick it
- 15:46When and how does it make sense to use Lighthouse?
- 20:06Lighthouse and real-world data can be very different
- 21:09Thank you
related talks
40mYou’re probably using Lighthouse wrong bt Filip Rakowski
38mYou’re probably using Lighthouse wrong
42mYou’re probably using Lighthouse wrong
27mCommon misconceptions around Google Lighthouse
26mDemystifying Lighthouse, CrUX, and Core Web Vitals
27mHow to Measure Performance Effectively?
35mExploring the Lighthouse
48mLighthouse for Chrome