The three metrics and their thresholds
- LCP (Largest Contentful Paint) — how long it takes for the largest element in the first screen, an image or block of text, to appear. Good — up to 2.5 seconds; poor — over 4 seconds.
- INP (Interaction to Next Paint) — how quickly the page responds to clicks and taps. Good — up to 200 ms; poor — over 500 ms. In March 2024, INP replaced the earlier FID metric.
- CLS (Cumulative Layout Shift) — how much the layout shifts while loading. Good — up to 0.1; poor — over 0.25.
Assessment uses the 75th percentile: a page is considered good if the thresholds are met for at least 75% of visits. Mobile and desktop are assessed separately.
Field data and lab data
This is the key thing to understand, otherwise the numbers will confuse you.
- Field data comes from real visitors using Chrome over the last 28 days — the Chrome UX Report (CrUX). This is what the Core Web Vitals assessment is based on.
- Lab data is a one-off simulated measurement: a typical phone on a typically slow network. It helps you find causes and test a fix right away, but it isn’t what your visitors experience.
So a score of “45 out of 100” in a lab test and green field metrics are not a contradiction — they’re two different measurements.
Step 1. PageSpeed Insights
- Open PageSpeed Insights, paste the page URL and run the analysis.
- At the top is field data: the three Core Web Vitals metrics and whether the assessment is passed. You can switch between data for this page and for the whole site (origin).
- Below is the lab test with a score from 0 to 100 and diagnostics showing what exactly slows the page down.
If the top section says there isn’t enough data, the page has too few Chrome visitors. That’s normal for small sites — then rely on the lab test and the Search Console report.
Step 2. The Search Console report
In Search Console, open the Core Web Vitals report. It covers the whole site rather than one page: how many URLs are good, how many need improvement and how many are poor — separately for mobile and desktop. Google groups similar pages, so a single template problem shows up across hundreds of URLs at once.
If your site isn’t in Search Console yet, start with this guide.
Step 3. Find the cause
- Poor LCP — most often a heavy hero image, a slow server response, or fonts and scripts that delay rendering.
- Poor INP — heavy scripts: chat widgets, tracking codes, sliders, page builders.
- Poor CLS — images and banners without set dimensions, late-loading fonts, ad slots and cookie banners that push content around.
PageSpeed Insights diagnostics point to specific files. I described how we automated these checks in the article on connecting PageSpeed Insights to Claude.
Don’t expect instant results after fixes. Field data builds up over 28 days, so improvements show in reports gradually. The lab test will show the effect immediately.
Common mistakes when checking
- Only the home page is tested, while the problems are on product or article pages.
- Looking at desktop results when most visitors come on phones.
- Chasing 100 points in the lab test instead of green field metrics.
- Drawing conclusions from a single run: lab results vary from run to run.
In short
- Thresholds: LCP up to 2.5 s, INP up to 200 ms, CLS up to 0.1 — for 75% of visits.
- The assessment is based on 28 days of field data; the lab test helps you find causes.
- Check a single page in PageSpeed Insights and the whole site in the Search Console report.
- Look at mobile first.
- After fixes, allow a few weeks for field data to update.
If your metrics are red and you don’t know where to start, we can speed up your site and sort out Core Web Vitals: measure before and after, and fix images, scripts and caching.
