Google has paid attention to site speed for years. Core Web Vitals is how it measures it - three numbers that affect the user experience, to some degree your position in search, and whether a client waits for your offer to load at all.
Short answer: what matters most is data from real visits, especially on mobile, because Google mainly evaluates the mobile version of a site. The three metrics are LCP under 2.5s, INP under 200ms and CLS under 0.1. The biggest single win is usually images; the biggest loss comes from scripts nobody uses.
The three numbers to know
| Metric | What it measures | |
|---|---|---|
| LCP | under 2.5s | How long until the main content appears |
| INP | under 200ms | How fast the page answers a click |
| CLS | under 0.1 | Whether content jumps while loading |
- LCP (Largest Contentful Paint) - how long until the main content appears. Target: under 2.5s.
- INP (Interaction to Next Paint) - how fast the page answers a click. Target: under 200ms.
- CLS (Cumulative Layout Shift) - whether content jumps while loading. Target: under 0.1.
Worth knowing: INP replaced the older FID metric in 2024 and it is stricter. FID measured the response to the first click; INP measures the worst response across the whole visit. A site that passed before may have stopped passing since, with no change to its code.
What speed does to sales
A 2020 Google and Deloitte study found that a 0.1-second mobile speed improvement was associated with conversion rising 8.4% in retail and 10.1% in travel. It is a correlation within a specific group of brands, not a universal rule, but the direction is clear: a slow site costs clients, and with a premium client it also undermines the brand.
~8%
conversion lift per 0.1s saved
40-60
typical mobile score for a theme
2.5s
Google's threshold for a good LCP
53%
of mobile visits are abandoned when a page takes over 3s to load (Google, 2016)
Speed is not a technical feature. It is the first impression, a ranking signal and a conversion lever at once.
Lab data versus field data
This distinction explains most of the confusion around PageSpeed. The score you see immediately is a lab test: one simulation, one connection, one device. Below it, if the site has enough traffic, sits a second section - real user data from the last 28 days.
Google ranks on the second one. That is why a site can score 95 in the test and still fail Core Web Vitals, because too large a share of real visitors arrive on a phone with a weak signal - the assessment uses the 75th percentile of visits. And why, after optimising, you wait a few weeks before the assessment moves: field data covers the last 28 days.
Where the seconds actually come from
On company websites the order of wins is almost always the same:
- Images. Usually the heaviest part of the page. A 4MB photo straight off the camera instead of 200KB of WebP is three seconds on a phone.
- Scripts nobody uses. Carousels, animation libraries, four analytics tools, a chat widget nobody answers. Each blocks the main thread.
- Fonts. Four families at three weights, loaded from a third-party server, with the default behaviour that hides text until they arrive.
- Images without dimensions. That is the whole of CLS: the browser does not know how much space to reserve, so content jumps when the photo lands.
- The server. Shared hosting answering in 800ms eats a third of the budget before anything renders.
The first two can be fixed without rebuilding anything, and are usually the right place to start.
Why templates lose this race
WordPress themes and site builders load dozens of scripts just in case - carousels, animation libraries, trackers. That ballast often cannot be switched off easily. Which is why the median mobile score of WordPress sites in HTTP Archive data is around 40, while a site written from scratch, without unnecessary scripts, has a much easier path to a good score - although the technology alone does not guarantee it.
The full comparison of both approaches, with five-year costs, is in WordPress or custom code.
How to check your site in 30 seconds
Go to pagespeed.web.dev, enter the address and start with the section showing data from real visits, if it is available - that is what matters to Google. The score below it is a lab test: good for finding problems, not for predicting rankings. A low mobile score is a signal to look at images, scripts and fonts.
If you run a shop or a site heavy with photography, the piece on what an online store costs covers how performance affects running costs as a catalogue grows.
FAQ
Is 100/100 in PageSpeed necessary?
No. Passing the Core Web Vitals thresholds in real user data is necessary. A hundred points in a lab test is pleasant, but it is not what ranking is based on.
We optimised and nothing changed. Why?
Because field data is a 28-day average. A change deployed yesterday starts showing in a few weeks and settles after a month.
Does speed really affect ranking?
It does, but less than agencies selling optimisation suggest. It is a tie-breaker between pages with comparable content - it will not beat somebody answering the question better. Conversion, on the other hand, it affects directly and immediately.
Where do I start with no budget?
Compress the images and remove the scripts you do not use. That is two hours of work and usually the biggest single jump in score.
Summary
Core Web Vitals asks three questions: is the content visible, does the page respond, and does anything jump. Measure them on the mobile score and on real user data, not on a single lab test. Start with images and unused scripts - that is where most of the lost seconds are.