PageSpeed Insights vs Lighthouse — How to Read Your Site's Real Performance Data

You optimized images, minified CSS, deferred scripts. Lighthouse gives you 100 out of 100. Then you open PageSpeed Insights and see a 45. Or the opposite: users complain the site is slow, but lab tests show it's fast. The culprit? The difference between lab data (Lighthouse) and field data (CrUX, real user data).

At Meteora Web, we deal with these tools daily for clients across industries. The first step to real optimization is understanding that a green Lighthouse score doesn't mean your site is fast for visitors. This hands-on guide explains the differences, how to read them, and what to do to improve the numbers that matter: real user metrics.

What is the difference between lab data and field data in PageSpeed Insights?

Lighthouse runs a test in a controlled environment (a headless browser on a Google server with a simulated connection). It's fast and reproducible, but doesn't represent real devices, networks, and behavior. PageSpeed Insights shows both:

• Lab data: Generated by Lighthouse on the fly, useful for diagnosing technical issues.
• Field data: From the Chrome User Experience Report (CrUX), collected from millions of real Chrome browsers. This is what Google uses for ranking.

The colored score (green, orange, red) in PSI is based on field data, not Lighthouse. That's why the two often disagree.

Why field data is the priority

Google uses Core Web Vitals (LCP, FID, CLS) calculated from field data for search ranking. If CrUX shows poor metrics, your ranking suffers, regardless of your Lighthouse score. Field data accounts for 3G connections, older hardware, slow devices — your real users.

Real example: An e-commerce client scored 95 on Lighthouse. But mobile field LCP was 5.2 seconds. We discovered the server was in Milan, while many users were in Sicily with unstable mobile connections. By optimizing the CDN and server-side rendering, LCP dropped to 2.1 seconds in field data. The light went from red to green.

PageSpeed Insights vs Lighthouse: when to use each tool?

Your choice depends on the goal:

• Use Lighthouse standalone (or in Chrome DevTools) for a quick technical audit: find render-blocking resources, oversized images, unoptimized JavaScript. Great for immediate debugging.
• Use PageSpeed Insights to understand real user experience. The “Discover what your users experience” section (based on CrUX) shows metrics at the 75th percentile. That's what Google evaluates for ranking.

How to run Lighthouse from the command line

To automate audits or test from a server, install Lighthouse CLI:

npm install -g lighthouse
lighthouse https://yoursite.com --view --preset=desktop

For CI/CD, use lighthouse-ci.

How to read PageSpeed Insights without being fooled

When you open PSI for a URL, you see the field data score (if available) and lab data below. Here's what to check in order:

1. “Metrics” section (field data): LCP, FID, CLS. If missing (“Not enough field data”), your site has too little traffic for CrUX. In that case Lighthouse is your only reference, but don't ignore the issue: collect data via the CrUX API or RUM tools.
2. “Opportunities” section (lab data): Lighthouse suggests optimizations. Even if field is good, applying them prevents future regressions.
3. “Diagnostics” section: Shows details like Total Blocking Time (TBT) and First Contentful Paint (FCP). TBT correlates with FID, but doesn't replace field data.

Checklist to avoid misleading scores

• If field score is red (LCP > 4s or CLS > 0.25), act immediately. Lab data is only a support.
• If field is green but lab is low, you probably have a strong server and CDN, but some client-side optimization could improve. Don't relax: users might switch to slower devices in the future.
• Don't obsess over Lighthouse 100. Google itself says beyond a certain point marginal gains aren't worth the effort. Focus on CrUX metrics.

Why field data (CrUX) is more reliable for Google ranking

Google's ranking system uses Core Web Vitals based on field data. Since 2021, they are an official ranking factor. However, not all pages have enough data. If your site is new or low-traffic, Google doesn't have enough samples to evaluate field scores. In that case, you're not penalized, but you don't get the ranking boost.

To get field data you need:

• At least 1000 unique Chrome visitors in the last 28 days.
• Data aggregated by origin (domain) or URL if traffic is sufficient.
• Check field data in Search Console (under “Core Web Vitals”) and in PageSpeed Insights for specific URLs.

Advanced tool: CrUX Dashboard on Looker Studio. Connect CrUX to monitor metrics over time for your whole domain. See the official guide.

How to improve real metrics using lab data

Lab data is useful for diagnosis, but actions must be driven by field metrics. Here's a sequence we use in our projects:

1. Identify the bottleneck from field data: If LCP is high, focus on server response time and rendering of the main content. If CLS is high, check layout shifts from images without dimensions, late-loaded fonts, dynamic ads.
2. Use Lighthouse to simulate changes: After applying a fix, run Lighthouse to verify lab score improves. But don't stop there.
3. Monitor field data for at least 28 days: CrUX updates daily but averages over 28 days. A change can take weeks to reflect in field data.
4. Use RUM (Real User Monitoring): Tools like Google's Web Vitals JavaScript library let you collect real metrics from your users and send them to Google Analytics (GA4). This shows immediate impact.

Practical example: reducing field LCP

A client had a field LCP of 4.8 seconds. Lighthouse showed the largest element as a hero image loaded via slow JavaScript. We:

• Removed lazy loading from that image.
• Optimized the image (WebP, correct dimensions).
• Added <link rel="preload"> for the hero image.

After 4 weeks, field LCP dropped to 2.1 seconds. The PSI score went from red to green.

What to do now

1. Open PageSpeed Insights for your domain and check the “Discover what your users experience” section. If no data, use Search Console to check traffic.
2. Download a CrUX report for your domain via the CrUX Dashboard on Looker Studio.
3. Run a Lighthouse audit and compare recommendations with field data. Prioritize optimizations that impact field metrics.
4. Monitor Core Web Vitals weekly. If you see degradation, act immediately. Don't wait for Google to penalize your ranking.
5. Read related guides: For deeper insights on GA4 tracking and image security, check our Core Web Vitals pillar guide and Firebase Storage — Image Upload and Security.

Remember: a site is measured in revenue, not compliments. If field data is good, your customers are happy and Google rewards you. Lighthouse's green score is just a tool, not the finish line.