Login
Speed optimisation tricks, or selling numbers you don’t need
You’ve probably seen it: “we build super-fast websites, 100/100 in the speed test”. Yeah, all green, wow effect. But what’s really behind it? And why do people pay money for things that make their website worse?
Here are the tricks that raise Google’s speed score and still hurt real users, and what to do instead.
Delayed loading. The “delay until interaction” option most optimisation plugins and setups have: hold something back until a mouse move, a tap, a scroll, or a timer, whichever comes first. Google’s test never moves the mouse, so the thing never loads during the test, and the score goes up.
The usual target is the big hero image on mobile. In the test it simply isn’t there. On a real phone, a visitor who taps quickly sees an empty hero until they do. One who just reads sees the image pop in 5 to 10 seconds later, and Google’s real-user data then counts that late image as the page’s main content, so the number Google actually uses for ranking goes down.
What to do instead: if most of your visitors are on slow mobile connections and the image really slows the page, don’t put an image in the mobile hero at all. Everywhere else, serve it at the right size in a modern format, let it load first, and check your real-user data. If it passes, the image is fine.
Delayed scripts. GTM, for example. Same reason: no interaction, no load. But the first interaction is usually a click on your main button, and you lose that event, because the script only starts loading at that moment. And everything you held back now runs on that first tap, which is exactly what Google’s responsiveness metric (INP) measures. If you really have to delay GTM, at least never on the page where your main conversion fires. Or catch the critical buttons and events with a small custom script that pushes them to the dataLayer: GTM picks them up as soon as it loads, so nothing is lost.
Sliders, videos, dropdowns. Slider images load only when the user starts interacting. Is that reasonable every time? The user then waits for them. Same with video. So look at real behaviour: if most people use something right away, like a dropdown menu, it can’t be loaded later at the cost of user experience. For sliders, load the next slide quietly once the page has finished loading, not on click.
And here’s what it looks like in practice. One of our client sites scores 78 in Google’s speed test. At the same time, all of its ~6,400 URLs pass Core Web Vitals in Search Console: the last 3 months in the screenshot, and it’s been like this for 2 years. And it what really matter for you, Google ranking and main for your visitors.
We load every script up front. No tricks.
Real user experience and what Google’s speed test measures can be drastically different things
[link] [comments]
