Improving Core Web Vitals: load time and user experience

The three metrics

LCP (Largest Contentful Paint): How quickly does the largest visible content appear? Guideline: under 2.5 seconds.
INP (Interaction to Next Paint): How quickly does the page respond to input? Guideline: under 200 milliseconds.
CLS (Cumulative Layout Shift): Does the layout jump around while loading? Guideline: under 0.1.

How to measure

Use PageSpeed Insights or the “Core Web Vitals” report in Search Console. Both use data from real visitors.

What helps

Images: Modern formats (WebP, AVIF), the right size, and lazy loading for images outside the visible area.
Scripts: Fewer and smaller scripts; remove services you do not need.
Fonts: System fonts or locally hosted fonts.
Reserve space: Images and ads with fixed dimensions, so that nothing jumps.
Server: Fast delivery and caching.

Lab data and field data

PageSpeed Insights shows two worlds. Field data comes from measurements of real visitors over a period of about 28 days and is what counts for Google. Lab data is a single test run under fixed conditions and is useful for finding causes and checking improvements. For pages with few visitors, field data is often missing, and only lab data remains. Important: after a change, it takes weeks for the field data to shift noticeably.

An example of troubleshooting

Suppose a tradesperson’s website has a poor LCP value. The report shows that the largest element is a JPEG header image of several megabytes, embedded as a background image and loaded late. The solution has three steps: reduce the image to the width at which it is actually displayed and convert it to a modern format, embed it as a regular image instead of a CSS background, and do not lazy-load this particular image. For an image below the visible area, on the other hand, lazy loading is the right choice. You will find more on image optimisation under Optimising images.

Typical causes for each metric

Poor LCP: large images, slow servers, blocking scripts or fonts in the head section.
Poor INP: heavy scripts that occupy the main thread, such as chat widgets, tracking packages or elaborate sliders.
Poor CLS: images without width and height attributes, banners that load late, cookie notices that push the content down, fonts that change size while loading.

Order of measures

Measure the most important page types, not just the home page: a service page, an article, the contact form.
Fix the single biggest problem first. It is often one image or one script.
Remove services that nobody analyses any more. Every script costs load time.
Measure again and document the result.

Myth and fact

Myth: A perfect score of 100 in PageSpeed Insights is essential. Fact: What matters is whether the page reaches the “good” thresholds in the field data. The lab score is only an aid. A value in the green range is enough; beyond that, extra effort usually brings little. The foundation remains a clean, mobile-friendly structure, as mobile-first indexing requires anyway.

Check on a smartphone too

The metrics often differ markedly between mobile and desktop, because phones have weaker processors and slower connections. Since many visitors come on mobile, look at the mobile values first. Also test the page on a real, older device over a mobile network: if it still feels smooth there, it will run well on better devices too.

Realistic expectations

Speed is one factor among many, but it noticeably improves the user experience. More under Technical SEO.