Dark green Relative Marketing Group cover graphic headlined "Lighthouse Score vs. Real-User Data" with a diagram of lab and CrUX field data leading to a pass/fail verdict

A familiar handover goes like this. A new website launches, the developer shares a PageSpeed Insights screenshot with a performance score in the 90s, and everyone agrees the site is fast. A few weeks later, the Core Web Vitals report in Search Console lists a batch of mobile URLs as “Poor.” Nobody made a mistake with the numbers. The score and the report measure two different things, and only one of them is the thing Google describes as a Core Web Vitals result.

This post explains what each number is, why they disagree so often, what Google’s own documentation says they do and do not do for search, and a short check you can run today to find out which number you should be acting on.

Two measurements on the same screen

PageSpeed Insights shows two separate results for the same URL, and the difference between them decides what you fix. The Core Web Vitals assessment at the top of the report comes from Chrome User Experience Report (CrUX) field data: real Chrome users’ visits over the previous 28 days, judged at the 75th percentile. The 0 to 100 performance score further down comes from a single Lighthouse lab run on a simulated device and connection. A page passes the Core Web Vitals assessment only when its 75th-percentile Largest Contentful Paint (LCP) is 2.5 seconds or less, its Interaction to Next Paint (INP) is 200 milliseconds or less, and its Cumulative Layout Shift (CLS) is 0.1 or less. The lab score plays no part in that pass or fail verdict.

Google’s PageSpeed Insights documentation sets out both halves. Field data is powered by CrUX and covers “the previous 28-day collection period.” Lab data is produced by Lighthouse, which simulates a page load on one device under one fixed set of network conditions. The documentation states that a lab score of 90 or above is considered good, and then adds the caveat that matters: good lab data does not necessarily mean real-user experiences will also be good.

The thresholds themselves are published on Google Search Central’s Core Web Vitals page: LCP within 2.5 seconds, INP under 200 milliseconds, and CLS under 0.1. The 75th percentile is the value that 75 percent of recorded visits matched or beat, which means the slowest quarter of real visits is what sets your result.

Why the lab score and the field verdict disagree

When we audit a site where the two numbers point in opposite directions, the cause almost always falls into one of these five categories.

  • No one clicks anything in the lab: A Lighthouse page-load test loads the page and stops, so it cannot observe the taps, clicks and keystrokes that INP measures; it reports Total Blocking Time as a stand-in, and a page can look fine on that proxy while real interactions stall.
  • One simulated device versus every real one: The lab run emulates a single mid-range phone on a throttled connection, while field data includes your actual visitors on older handsets, congested mobile networks and browsers with a dozen extensions running.
  • The 75th percentile, not the average: Field results are judged at the point where a quarter of visits were slower, so a page that is quick for most people can still fail if one visitor in four has a poor experience.
  • A 28-day memory: CrUX reports a trailing 28-day window, so a fix deployed this morning is blended with nearly four weeks of visits to the old version, and a regression introduced yesterday will not show up in full for weeks either.
  • Things that only happen to real visitors: Tag-manager scripts that fire after a consent choice, chat widgets that load on scroll, A/B testing scripts and personalized content often behave differently, or not at all, in a cold lab run.

What we see in real audits

Three patterns come up repeatedly in our technical SEO work, across very different kinds of businesses.

The first is a strong lab score paired with a failing INP in the field. The usual culprits are third-party scripts that wake up when someone interacts with the page: tag-manager triggers bound to every click, live-chat and booking widgets, form-validation libraries and heavy navigation menus. None of them runs during a page-load test, so the lab score never sees the cost. The fix is rarely “optimize images”; it is auditing which scripts listen for clicks and deferring or removing the ones that do not earn their place.

The second is the low-traffic site that has no page-level field data at all. PageSpeed Insights only shows URL-level results when there are enough real visits to include that page in CrUX. When there are not, it falls back to origin-level data covering every page on the domain. For many small and mid-sized sites, the “field data” on a service page is really the whole site’s data, so a slow blog template can drag down the verdict shown on a fast landing page. Read the label in the field-data section before drawing any conclusion about a specific page.

The third is the redesign judged on launch day. Because of the 28-day window, a new build cannot be fairly assessed in the field until roughly four weeks of real traffic have accumulated. We set that expectation before launch, because a report showing the old site’s problems in week one causes unnecessary alarm.

The inconvenient truth in all three: chasing a lab score of 100 is often wasted effort. Google’s page experience documentation says plainly that trying to get a perfect score just for SEO reasons “may not be the best use of your time.”

What Google says Core Web Vitals do for search

The same page experience documentation confirms that Core Web Vitals are used by Google’s ranking systems, and in the same answer cautions that good results in Search Console or third-party tools do not guarantee a top position. It also states that Google will still show the most relevant content even when its page experience is sub-par. In practice, that makes Core Web Vitals a tiebreaker among comparably helpful pages and a real user-experience problem in their own right, not a lever that outranks relevance.

One date is worth knowing. According to the Google Search Central blog, Interaction to Next Paint replaced First Input Delay as the Core Web Vitals responsiveness metric on March 12, 2024. If an audit or agency report you received this year still grades your site on First Input Delay, it is working from a retired metric, and its responsiveness findings should not be trusted.

A ten-minute check you can run today

  1. Open PageSpeed Insights, paste the URL of the page that matters most to your business, and stay on the Mobile tab.
  2. Read the field-data section at the top before anything else. Note whether it is showing data for this URL or for the whole origin.
  3. Write down which of LCP, INP and CLS is failing at the 75th percentile. That metric, not the lab score, is your priority.
  4. Only then scroll to the Lighthouse diagnostics, and look for the opportunities that relate to the failing field metric.
  5. In Search Console, open Experience, then Core Web Vitals, then the Mobile report, and check whether the failing page belongs to a group of similar URLs. A shared template problem is fixed once, not page by page.
  6. After you ship a fix, put a reminder in your calendar for 28 days later and re-check the field data then.

Where the effort usually belongs

Once you know which metric fails in the field, the work narrows quickly. A failing LCP points to the hero image, the server’s response time, or render-blocking CSS and fonts. A failing INP points to JavaScript, especially third-party scripts and long tasks on the main thread. A failing CLS points to images and embeds without reserved dimensions, late-loading banners and web fonts that swap in at a different size. Most of these are decided when a site is built, which is why performance budgets belong in the web design brief rather than in a cleanup project after launch.

If you want the basics of finding these reports in the first place, our earlier guide on how to check Core Web Vitals on a small business site walks through each tool. And if your field data is failing and you are not sure which scripts are responsible, that is exactly the kind of diagnosis our SEO services team runs as part of a technical audit. The goal is not a perfect score. The goal is that the slowest quarter of your real visitors still has a good experience.

×
✦ Meet MIA – Your AI Marketing Assistant

Get Instant Marketing
Answers + Book Strategy Call

Talk to MIA and get real answers, custom insights, and expert help to grow your business.
💬
Answer your marketing questions

Get instant expert answers to help you grow.

📈
Analyze your website or ads

MIA can audit and analyze performance in seconds.

💡
Get growth strategies

Receive tailored recommendations.

📅
Schedule a call with our team

Book a strategy consult instantly.

Start Voice Chat with MIA Click to talk
📅 Book a Strategy Call Schedule a consult
⚡ Get Quick Answers Fast responses
🔍 Request Free Audit Analyze my site
MIA Bot

MIA

● Online
Listening...
🎙️

Click the mic to start speaking. We only use your voice to understand your questions.

Skip to content