A slow website loses visitors before they even see what it has to offer. Google takes page speed into account when ranking sites, and users simply close the tab. Let’s take a look at how to test your website and what to fix first.
Three Metrics Google Looks At
This set of metrics is called Core Web Vitals. There are three of them, and each one corresponds to a different aspect of whether a site feels “fast or slow”:
| Metric | What it measures | Standard |
|---|---|---|
| LCP | when the main element of the screen appeared—usually a large image or a headline | up to 2.5 seconds |
| INP | How quickly the page responds to actions: clicks, input | up to 200 ms |
| CLS | Does the layout shift while loading? | up to 0.1 |
INP replaced the previous FID metric in 2024: it no longer measures the first interaction, but rather all interactions per visit.
How to check
PageSpeed Insights
The easiest way: paste the address into pagespeed.web.dev. Check out the "Mobile" tab—that's where most visitors come from, and that's where the numbers are worse.
It's important to distinguish between the two sections of the report. The top section contains data from real users over the past 28 days—this is what Google takes into account. The bottom section consists of lab tests using emulation; it's needed to identify causes, but it is not an evaluation in and of itself.
Search Console
The "Key Web Metrics" section shows the status of the entire site, rather than a single page, and groups URLs by issue. If there are few pages, there may not be enough data—this is normal.
The simplest test
Open the website on your phone using mobile data, not Wi-Fi. Count the seconds until you can tap something. This test is more reliable than any score.
What to Fix First
Pictures — Source 80% of Excess Weight
A typical scenario: a 4,000-pixel, 8 MB photo from a camera displayed in an 800-pixel-wide block. The browser will load it in its entirety.
- Scale down to the actual display size.
- Convert to WebP—at the same quality, the file is 25–35% smaller.
- Place
widthandheightIn the code: without them, the page jumps when the image finishes loading, which affects the CLS. - Enable lazy loading for everything below the first screen. But not for the main image at the top—that, on the contrary, should be loaded first.
You can prepare the images right now: Compress and convert to WebP — Up to 30 files at a time, without uploading them to the server.
Third-Party Scripts
Every widget—whether it's a chat box, a map, a review widget, or an ad pixel—is a request to a third-party server. Your hosting may be fast, but the page will still have to wait.
Check to see if you really need all of them and if they’re loading on the pages where they’re actually used. A feedback widget that’s enabled on every page just for the sake of one is a common issue.
Fonts
Host your fonts on your own server instead of loading them from third-party services, and add font-display: swap — Then the text will appear immediately in the system font, rather than after the file has loaded.
Hosting and Cache
If the server takes more than 0.5 seconds to generate a response, optimizing images won't help much. Enable page caching and check your PHP version: the site runs noticeably slower on older versions than on 8.2–8.3.
What Not to Do
Chasing a perfect score. PageSpeed is a diagnostic tool, not a goal. The difference between 85 and 100 points is imperceptible to the user, yet it takes more effort to achieve that than it does to go from 40 to 85.
Install five optimization plugins at once. They conflict with each other, and then it's impossible to figure out exactly what caused the layout to break.
Turn everything off. If you set aside analytics just to boost your metrics, you’ll lose sight of where your customers are coming from. Speed is a means to an end, not an end in itself.
How much does that actually amount to?
Speed is one of many ranking factors, and on its own, it won’t push a website to the top. But it does influence user behavior: how many people waited for the page to load and made it to the form. That’s why the effect is most noticeable with paid traffic—you pay for every click and lose a portion of that traffic even before your offer is displayed.
We've discussed other reasons why visitors don't submit requests separately: Seven places where the funnel flows through.
In Brief
Start with images and third-party scripts—that’s 80% of results for 20% of effort. Test on mobile and look at real-world user data, not lab scores.
If you don't want to figure it out on your own—we We accept websites for promotion, including those belonging to others, and we start with an audit: you receive a list of issues with an explanation of what’s critical and what can wait.
Telegram