Why your website feels slow (even though the speed test passed)

A client sent us a speed test showing 94 out of 100 and asked why their site still felt slow on a phone. Both things were true. Scores are measured in a lab, on a good connection, on a page that has already been warmed up. Users are not.

Three things decide whether a page feels fast, and none of them is the score.

How long until something appears

Largest Contentful Paint measures when the biggest thing on screen — usually a hero image or headline — finishes rendering. Under about two and a half seconds feels immediate. Past four, people start leaving.

The usual culprits, in the order we find them:

• A hero image served at full resolution to a phone
• Web fonts blocking text from rendering until they download
• Render-blocking scripts in the head that could load later
• A slow server response, so nothing can start until the HTML arrives

Fixing the hero image alone often moves this metric more than every other optimisation combined. Serve modern formats, size them for the device, and mark the one above the fold as high priority so the browser stops guessing.

Whether the page moves while you read it

Cumulative Layout Shift is the one users complain about without knowing its name: you go to tap a link, an ad or an image loads above it, and you tap something else. It is intensely annoying and it is almost always caused by elements that do not reserve their space before loading.

The fix is boring and reliable — give images explicit width and height, reserve space for anything injected later, and never insert content above existing content after load.

Whether it responds when you touch it

A page can paint quickly and still ignore you for a second afterwards, because the main thread is busy running JavaScript. On a mid-range Android phone this is dramatically worse than on the laptop the site was built on.

Ship less JavaScript. Split it so a page loads only what it uses, defer third-party tags, and be suspicious of any script whose value nobody in the room can articulate. Analytics, chat widgets and tag managers are usually the heaviest things on a marketing site, and they are rarely audited.

Test on a real mid-range phone on mobile data at least once. It is the most clarifying twenty minutes available to anyone who owns a website.

Measure what users get, not what the lab says

Lab tools are useful for catching regressions before release. But the number that matters is field data — what actual visitors on actual devices experienced. If your lab score is excellent and your field data is poor, trust the field data and go find out which devices and connections are having a bad time.

The short version

Right-size the hero image, reserve space so nothing jumps, and cut JavaScript you cannot justify. That is the majority of the win on most sites we are handed, and none of it requires a rebuild.