Core Web Vitals

Core Web Vitals are three numbers with which Google tries to make a feeling measurable: how bearable is it to use this page. The name sounds technical, the three questions behind it are not. When is something usable visible? How quickly does the page react to a click? And does something jump away while you are aiming at it?

The first value is called Largest Contentful Paint, LCP for short. It measures after how many seconds the largest visible element has loaded — usually the main image or the headline. Up to 2.5 seconds counts as good. The most common reason for a poor result is an oversized, uncompressed photo right at the top of the page.

The second value is called Interaction to Next Paint, INP for short. It measures how long it takes for the page to answer an input visibly — a click, a tap, a key press. 200 milliseconds or less counts as good. This value replaced an older one called FID in March 2024, because that one only looked at the very first input and was therefore too generous.

The third value is called Cumulative Layout Shift, CLS for short, and it describes things sliding about. Anyone who has ever aimed at a button and clicked an advert that loaded in at the last moment knows the problem. Up to 0.1 counts as good. The usual culprits are images without fixed dimensions, fonts loaded in afterwards, and embedded elements that only claim their space once they load.

What matters is where the numbers come from, and here lies a misunderstanding that costs a lot of money. The values Google uses for its assessment come from real visitors on real devices with real connections, collected over 28 days and evaluated at the 75th percentile — so three out of four visits have to be good, not the average. A single test on a fast office computer says little about that.

That is why results from testing tools vary considerably. We measure every address several times and state the range alongside, instead of picking out one flattering number — the score from such a tool can differ by several points between two runs without anything on the page having changed. Anyone who gives you a speed figure without a range and without a date has either measured once or picked the most favourable run.

How strongly these values feed through to position is one of those things that has to be answered honestly: they are one factor among many, and Google itself describes them as a tiebreaker between pages that are otherwise similarly good. A slow site with fitting content still beats a fast one without. The tangible gain lies elsewhere: whoever sits in front of a white screen on their mobile for three seconds never gets as far as calling you.

For the rebuild that means a manageable list: images in contemporary formats and at the size they are displayed. Fixed heights and widths, so that nothing slides about. Fonts delivered with the site instead of fetched from someone else's server. And everything that is not needed immediately loaded later. Which of these actually applies in your case is shown by a measurement — which is why it comes before the work with us, not after. The connection with what visitors do next is described in the entry on conversion.