Understanding it
Google sets a threshold for each metric. LCP (Largest Contentful Paint) should stay under 2.5 seconds: the time it takes the largest visible element to render. INP (Interaction to Next Paint) should stay under 200 milliseconds: the delay before the page responds to a click or a keystroke. CLS (Cumulative Layout Shift) should stay under 0.1: it measures layout jumps.
These thresholds are assessed on 75% of visits, separately on mobile and desktop. They are field data, reported by the Chrome browser, not a lab score. INP replaced the older FID metric in March 2024.
Google recommends good results but points out that relevant content comes first. Poor Core Web Vitals rarely sink an excellent page, but they can separate two similar pages and drive visitors away.
A concrete example
Illustrative example: a brochure site that jumps
An architecture firm’s website shows a large photo at the top of the page. On mobile, it takes 4 seconds to appear, and the text suddenly drops when a cookie banner slides in above it.
The team resizes the photo to its display size, converts it to a modern format, loads it first and reserves space for the banner. LCP drops back under 2.5 seconds, CLS under 0.1, and the page no longer moves while people read.
The common mistake
Chasing a 100 score on PageSpeed Insights
The PageSpeed Insights lab score is a diagnostic tool, not the goal. What counts is field data from Search Console and the Chrome report, and passing the thresholds. Gaining 3 points on an already fast page makes little difference.