Lighthouse
An open-source Google auditing tool measuring performance, accessibility, SEO, best practices, and PWA readiness. Produces lab-based scores, not field data.
§ 1 Definition
Lighthouse generates a 0-100 score across five categories. The Performance score is a weighted composite of lab metrics (FCP, LCP, TBT, CLS, Speed Index). It simulates a Moto G4 on a throttled network. A 100 Lighthouse score does not guarantee good Core Web Vitals in the field.
§ 2 How It Works
Runs a controlled simulation on a Moto G4 with throttled network. The performance score uses a log-normal distribution. Provides actionable recommendations.
§ 3 Common Misunderstandings
Lighthouse scores are NOT real-user metrics. They are lab benchmarks. A site can score 100 in Lighthouse but fail CWV for real users. Use Lighthouse for CI regression detection, not as a performance SLA.
§ 4 When to Use
In CI/CD pipelines to catch regressions. For reproducible debugging. In Chrome DevTools for iteration. Always validate with CrUX data.
§ 5 Note
§ 6 In code
npx lighthouse https://example.com --view --preset=desktop§ 7 Common questions
- Q. Can I pass Lighthouse but fail CWV?
- A. Absolutely. Lab vs field data often disagree.
- Q. Does Lighthouse measure INP?
- A. No. It reports TBT (Total Blocking Time), which correlates with INP.
- Lab auditing tool, not field measurement.
- High score does not guarantee good CWV.
- Best use: CI regression detection.
- Always complement with CrUX or RUM.
Atomic Glue uses Lighthouse for CI regression testing and diagnostic debugging. Get in touch.
Get in touchAn open-source Google auditing tool measuring performance, accessibility, SEO, best practices, and PWA readiness. Produces lab-based scores, not field data.
Definition
Lighthouse generates a 0-100 score across five categories. The Performance score is a weighted composite of lab metrics (FCP, LCP, TBT, CLS, Speed Index). It simulates a Moto G4 on a throttled network. A 100 Lighthouse score does not guarantee good Core Web Vitals in the field.
How It Works
Runs a controlled simulation on a Moto G4 with throttled network. The performance score uses a log-normal distribution. Provides actionable recommendations.
Common Misunderstandings
Lighthouse scores are NOT real-user metrics. They are lab benchmarks. A site can score 100 in Lighthouse but fail CWV for real users. Use Lighthouse for CI regression detection, not as a performance SLA.
When to Use
In CI/CD pipelines to catch regressions. For reproducible debugging. In Chrome DevTools for iteration. Always validate with CrUX data.
Note
Scores are not comparable across major versions. Always use the latest version.
In code
npx lighthouse https://example.com --view --preset=desktop
Common questions
Q: Can I pass Lighthouse but fail CWV?
A: Absolutely. Lab vs field data often disagree.
Q: Does Lighthouse measure INP?
A: No. It reports TBT (Total Blocking Time), which correlates with INP.
Key takeaways:
Lab auditing tool, not field measurement.
High score does not guarantee good CWV.
Best use: CI regression detection.
Always complement with CrUX or RUM.