What Is a Good LCP, INP and CLS Score?

Search Console has flagged a batch of your URLs as "needs improvement", or PageSpeed Insights has shown you three acronyms with coloured bars. Nobody on the team can say what number counts as a pass. Without that line, every fix is a guess and every developer estimate has nothing to anchor it.
A good Core Web Vitals score is an LCP of 2.5 seconds or less, an INP of 200 milliseconds or less and a CLS of 0.1 or less. Google judges each one at the 75th percentile of real visits over 28 days, separately for mobile and desktop. A page passes only when all three clear the line.
What counts as good, needs work and poor?
Each metric has three bands. These are the thresholds Google has used since INP joined the set in March 2024:
| Metric | Good | Needs improvement | Poor | | --- | --- | --- | --- | | LCP (Largest Contentful Paint) | 2.5 s or less | 2.5 s to 4 s | Over 4 s | | INP (Interaction to Next Paint) | 200 ms or less | 200 ms to 500 ms | Over 500 ms | | CLS (Cumulative Layout Shift) | 0.1 or less | 0.1 to 0.25 | Over 0.25 |
PageSpeed Insights and Search Console use these same bands for their green, amber and red. You can check any number in this article against the tool you already have open, which is the right habit with platform thresholds. They are Google's to change, and plenty of guides online still describe an older metric set.
Three things about the table matter more than they first appear to.
Amber is not a near pass. A page with a 2.5 s LCP and a 0.1 CLS but a 260 ms INP does not pass the assessment. The verdict is all-or-nothing across the three metrics, so treat "needs improvement" as a failure you have a head start on.
CLS is not a time. LCP is in seconds and INP in milliseconds, but CLS is a unitless score. A CLS of 0.1 is not a tenth of a second, and comparing it to the other two numbers tells you nothing.
The threshold is a floor, not a target. If your 75th-percentile LCP sits at 2.4 s, one heavier hero image or one new script will tip you into amber. Aim for headroom. A page that only just passes this month is one release away from failing the next.
Why does Google judge the 75th percentile?
Your score is not one page load. It is a distribution of thousands of real visits, collected from Chrome users and reported through the Chrome UX Report. Google reads that distribution at the 75th percentile. In plain terms, at least three in four visits have to be at or under the threshold.
The choice is deliberate. Judging at the median would let a page pass while half its visitors had a poor experience. Judging at the 95th percentile would let a handful of ancient phones on bad connections dominate the result, whatever you did to the page. The 75th percentile sits between the two. It is strict enough to cover most of your audience and stable enough that you can actually move it.
Two consequences follow for anyone trying to improve a score.
First, the data is a rolling 28-day window. A fix you ship today does not show its full effect for up to four weeks, because the old, slower visits age out gradually. Teams often ship a real improvement, check the next morning, see no change and assume it failed. Wait out the window before you judge.
Second, mobile and desktop are assessed separately, and they rarely match. The same page loads on slower processors and patchier networks on mobile, so the slow tail of visits is longer and the 75th percentile lands further out. A page can be comfortably green on desktop and amber on mobile without a single line of code differing. If mobile is where you keep failing, it is worth working through the broader question of whether your site is too slow on phones before you chase individual metrics.
What does each number actually measure?
Knowing the threshold is only useful if you know what moves it. On marketing sites, each metric tends to fail for a small, predictable set of reasons.
LCP: when the main content appears
LCP is the time until the largest visible element finishes rendering. That element is usually the hero image or the main headline. When LCP fails on a landing page, the usual cause is an oversized hero image that the browser does not know to fetch early. Slow server response is the other common cause, because it delays everything behind it.
INP: how fast the page responds
INP watches every tap, click and keypress during a visit and measures the delay until the screen visibly responds. It then reports roughly the worst of them. Poor INP almost always comes down to JavaScript tying up the browser's main thread. On marketing sites, much of that JavaScript is ours: chat widgets, tag managers, consent tools and the tracking pixels stacked inside them. Every tag added for measurement has a cost here, and the cost compounds.
CLS: how much the page jumps
CLS scores unexpected movement. Each shift is weighted by how much of the screen moved and how far it moved. The reported figure is the worst burst of shifts during the visit. Movement in the moment after a user taps or clicks is excluded, because the user caused it.
The causes are images and embeds without set dimensions, web fonts that swap in late, and ads or banners injected above existing content. A cookie banner that loads late and pushes most of the page down by a quarter of the screen height can exceed the whole 0.1 budget in a single shift.
Why your lab score can disagree with this
Everything above describes field data, meaning measurements from real visitors. The big performance score at the top of PageSpeed Insights comes from Lighthouse. Lighthouse runs one simulated page load on a throttled connection, and that is a different measurement entirely.
A site can score in the high 90s in Lighthouse and still fail Core Web Vitals in the field, and the reverse happens too. Lighthouse also cannot measure INP at all, because no real person interacts with a simulated load. Total Blocking Time is the closest lab proxy, and it is useful but not the same thing.
So when someone asks whether your scores are good, the answer that matters is the field data, read at the 75th percentile against the bands above. Lab tools are for diagnosis: they show you which image, script or font is causing the problem. If you are staring at a red lab score on a site that feels quick, there is a whole separate question of why PageSpeed says your site is slow but it feels fast and which one to believe.
Which score should you fix first?
With the thresholds in hand, the triage is straightforward:
- Read field data, mobile first. Open the Core Web Vitals report in Search Console, or the field section of PageSpeed Insights. Mobile is usually where the assessment fails, and for most sites it is where most visits come from.
- Fix red before amber. A metric in the poor band is further from passing and usually has a single dominant cause. Fixing it often brings a larger shift than polishing three amber metrics.
- Match the fix to the metric. For LCP, prioritise and resize the hero image. For INP, audit your tags and remove what nobody uses. For CLS, reserve space for every image, embed and banner before it loads.
- Start with the pages carrying paid traffic. You have already paid for every click that lands on those pages. A slow landing page wastes budget you committed before the visitor saw anything.
- Recheck after 28 days. Judge each change on a full window of field data, not the morning after release.
Conclusion
A good Core Web Vitals score is concrete: LCP at 2.5 seconds or less, INP at 200 milliseconds or less and CLS at 0.1 or less. Each is judged on real visits at the 75th percentile, over 28 days, with mobile and desktop counted separately. The numbers are the easy part. The work is reading the right data, fixing the worst metric first, and leaving enough headroom that next month's new script does not undo this month's fix.
Frequently Asked Questions
No. Core Web Vitals are a ranking signal, but a modest one next to relevance and content quality. A fast page with thin content will not outrank a slower page that answers the query better. Treat passing scores as protection against losing ground and as a conversion benefit for visitors, rather than as a lever that will move rankings on its own.
Field data needs enough real Chrome visits over the 28-day window to report reliably. New pages and low-traffic pages often fall short, so PageSpeed Insights shows no field result or falls back to data for the whole site. In that case, use lab tools to catch obvious problems and wait for traffic to build before judging the page against the thresholds.
No. INP replaced First Input Delay as the responsiveness metric in March 2024. FID only measured the delay before the browser began handling the first interaction. INP measures how quickly the page visibly responds across every interaction in a visit. Any guide or audit that still reports FID as a Core Web Vital predates the change and should be read with that in mind.
Get a complimentary audit to uncover AI opportunities hidden in your data.
Put these strategies to work


