Site speed decides how many people ever reach your offer. It is not a developer metric, it is a commercial one: a slow page loses buyers before they see a single product. This is a plain guide to Core Web Vitals for marketers: the numbers Google expects, what actually slows sites down (surprisingly often, video), how to measure your own in ten minutes, and the point where chasing a higher score stops paying off.
Key takeaways
- Google measures three things. LCP, how fast the main content appears, target under 2.5 seconds. INP, how fast the page responds to a tap, target under 200 milliseconds. CLS, how much the layout jumps while loading, target under 0.1.
- In a case study published by Google on web.dev, Vodafone improved LCP by 31 percent in an A/B test and sold 8 percent more.
- The thresholds apply to at least 75 percent of real visits, measured on real devices, not to a score from your office laptop.
- On creative and agency sites, the usual culprits are an autoplaying hero video and uncompressed photography, not cheap hosting.
- You can measure your own site for free in ten minutes with PageSpeed Insights and Search Console. The steps are below.
- Past a certain point nobody notices the difference. A perfect score is not the goal, a smooth purchase is.
Why does site speed decide whether people buy?
Because waiting is the one thing a visitor pays before buying anything. When a page lags, a share of people go back to the results and click a competitor. Nobody complains, you simply get fewer enquiries and never learn why.
The numbers back it up. Google published a Vodafone case study measuring core web vitals for marketers in commercial terms: an A/B test between two visually identical versions of a landing page, where the faster one improved LCP by 31 percent and produced 8 percent more sales, along with better lead and cart rates.
Search adds a second reason. Core Web Vitals are part of the page experience signals. They will not beat relevance, but between two comparable pages they decide against the slower one.
Which numbers does a site have to hit?
Google publishes three thresholds, and each has to be met by at least 75 percent of real visits:
- LCP under 2.5 seconds. That is how long the largest element in the first screen, usually the hero image or headline, may take to render.
- INP under 200 milliseconds. The time between a tap and a visible response. It replaced the older FID metric in March 2024.
- CLS under 0.1. A measure of how much content jumps around. If a button slides away from under a thumb during loading, this is your metric.
The detail most discussions miss: what counts is field data from real people on real phones and real connections, not a lab score from your desk on office wifi.
What slows most websites down?
The order almost never changes on the sites that reach us.
Hero video. Beautiful, effective and the most expensive thing you can do to your loading time. An autoplaying hero clip often weighs more than the entire rest of the page, and it loads exactly while a visitor is forming a first impression.
Full resolution photography. The photographer delivers four thousand pixel JPGs and they get uploaded as they arrived. The browser then scales them down to eight hundred pixels, but it still has to download all of it.
Third party scripts. Chat widget, cookie banner, two ad pixels, a map, a review badge. Each adds a fraction of a second and together they turn a fast site into a slow one.
A theme that does everything. Universal templates ship code for features you never use. On small sites this is the most common reason you cannot get under two seconds without a rebuild.
How do you put video on a site without wrecking it?
This part comes from our own delivery work, because we hand video over to web teams every month. The rules we hold to:
- Never ship the hero clip as a full resolution autoplaying file. If it has to autoplay, use a short silent loop under five megabytes, exported in three sizes for mobile, tablet and desktop.
- Embed long video in a player that loads only after a click. Until then the page should carry an image with a play button, nothing more.
- Match the poster image to the first frame. Otherwise the swap shifts the layout and your CLS suffers.
- Ask the production company for a web version, not the file that went to the presentation. The size difference is often tenfold for a quality difference no phone can show.
When we prepare video for a website, web exports are part of the handover. How to reuse the same shoot across channels is covered in our guide to social media video formats, and what belongs in a short homepage film is in the piece on the company intro video.
When should you stop optimising?
The honest counterargument. Nobody in your audience feels the difference between 1.2 and 0.9 seconds, and closing that gap costs a disproportionate amount of work. A PageSpeed score is a tool, not a target: you can tune a site to ninety points and still have one where nothing can be found.
We called it quits on one project when the last few points meant switching off the animation that carried the brand. The compromise went to the brand, and that was the right call. Speed is a condition, not a product.
So the priority order goes: make it usable on a phone, then get it under three seconds, and only then argue about tenths.
How do you measure your own site in ten minutes?
No developer required:
- Put your most important page into PageSpeed Insights and read only the section with data from real users.
- Note the three values, LCP, INP and CLS, into a spreadsheet with today's date so you have a baseline.
- Open the Core Web Vitals report in Search Console and see how many URLs sit in the red. It covers the whole site, so it also reveals which page template is the problem.
- Repeat for mobile and desktop. The gap is usually large, and mobile is the one that counts.
If the numbers stop making sense, or you need the site and the content solved together, get in touch. What we cover around web and production is listed under our services.
Frequently asked questions
Is site speed a Google ranking factor?
Yes, Core Web Vitals belong to the page experience signals, but they never outweigh relevance. A slow page with excellent content beats a fast page with nothing to say, while between two similar pages speed decides.
Will better hosting fix it?
It helps less than people expect. Hosting affects server response time, which is usually the smaller share of the problem. Most of the seconds come from images, video and scripts, in other words from what you put on the page yourself.
How often should we measure?
Monthly, and always after a bigger change. A new plugin, a new banner or a new tracking tag can spoil the result overnight, and without measurement you notice only when enquiries drop.
Do we need a new website to be fast?
Often not. Compressing images, embedding video properly and cleaning up third party scripts usually move a site into the green. A rebuild makes sense only when the theme or the platform itself is the bottleneck.
