Lighthouse Performance
Lighthouse audits web pages in a lab environment, producing a 0-100 performance score from LCP, TBT, CLS, and Speed Index. Use its diagnostics to fix issues, but rely on field data for real-world outcomes.
Lighthouse is an open-source, automated tool from Google for auditing web page quality. It runs against a page and produces scored reports across performance, accessibility, best practices, and SEO. The performance category is the most cited: it loads the page in a controlled lab environment and computes a 0 to 100 score from a weighted set of metrics, accompanied by concrete recommendations for improvement.
What It Measures
The Lighthouse performance score is a weighted blend of lab metrics. Largest Contentful Paint (LCP) and Cumulative Layout Shift (CLS) mirror Core Web Vitals. Total Blocking Time (TBT) measures how long the main thread is blocked during load and serves as a lab proxy for interactivity, correlating with the field metric INP. First Contentful Paint and Speed Index capture how quickly content begins and visually completes rendering. The overall score is the weighted geometric mean of these metrics, with TBT and LCP carrying the largest weights in current versions, reflecting their impact on perceived speed.
Methodology
Lighthouse runs in Chrome DevTools, as a command-line tool, as a Node module, or via PageSpeed Insights. It loads the page under simulated or applied throttling, by default emulating a mid-tier mobile device on a slow 4G connection, to approximate constrained real-world conditions rather than a fast developer machine. It records the metrics, applies version-specific weightings, and computes the score. Because it is synthetic, results are reproducible for a given configuration but depend on the test machine, CPU load, network throttling settings, and the Lighthouse version, so consistency requires controlling those factors.
How to Interpret Results
Treat the 0 to 100 score as a diagnostic summary, not an absolute grade: roughly 90 and above is good, 50 to 89 needs improvement, and below 50 is poor. More valuable than the number are the specific metric values and the actionable opportunities and diagnostics Lighthouse lists, such as unused JavaScript, render-blocking resources, oversized images, or excessive main-thread work. Use Lighthouse to find and fix causes during development and in CI, and use field data (CrUX or your own real-user monitoring) to judge actual outcomes for real visitors.
Limitations
Lighthouse is lab-only: a single simulated device and network that can diverge sharply from your real user base, so its score should never override field data. Scores vary run to run due to environment noise such as background CPU activity; run multiple times and compare medians rather than a single result. Default mobile throttling can surprise teams expecting desktop numbers. Weightings change between versions, breaking long-term score comparisons across upgrades. Use Lighthouse for diagnosis and CI regression checks, paired with real-user monitoring for ground truth.