Performance
Lighthouse
Lighthouse is Google's open-source lab tool that scores performance, accessibility, SEO, and best practices. It runs in Chrome DevTools, PageSpeed Insights, and the command line.
How it is measured
The performance score is a weighted mix of metrics: FCP, Speed Index, LCP, Total Blocking Time, and CLS. Each is scored against real-world data curves and combined into 0 to 100. LCP and TBT carry the most weight.
A run uses simulated throttling by default. Variation between runs is common, so use the median of several. It reports opportunities and diagnostics, with estimated savings in ms.
Worked example
A page scores 62 with LCP 3,800 ms, TBT 640 ms, and CLS 0.02. Deferring a chat widget cuts TBT to 180 ms and moves the score to 79. Preloading the hero lowers LCP to 2,400 ms and the score reaches 91.
On the same page, field INP is still 320 ms because a filter interaction is slow, which Lighthouse never clicks.
How it differs
Lighthouse gives lab results. PageSpeed Insights wraps Lighthouse and adds CrUX field data. Lab data in general is repeatable, and Lighthouse is its best known tool. A score of 100 does not mean real users are fast.
Common errors
Chasing 100 at the cost of features. Comparing scores from different machines. Running in a browser with extensions. Ignoring the field data section. Treating the score as an SEO ranking factor.
In practice
Run Lighthouse in an incognito window on your key templates, take the median of three, and look at diagnostics rather than the score. Compare against field data.