Website speed: why it decides things and how to improve it

How speed turns into money

The mechanism works two ways. Directly: a slow site loses people before they see the offer. Indirectly: search engines account for speed in ranking, and ad platforms account for it when pricing a click.

This is especially visible on mobile connections. What opens in a second on office Wi-Fi can take five seconds on the move — and a large share of your audience is exactly there.

Which metrics actually matter

An overall «speed score» is useful as a reference point, but chasing a hundred out of a hundred is pointless. What matters more is that a person on a phone sees the offer in a couple of seconds and can tap the button right away.

What slows sites down most often

1. Images

The number one cause on almost every site. A 4-megabyte photo displayed in a 300-pixel-wide block is routine. Fixed by compression, modern formats and correct dimensions.

2. Render-blocking scripts

The browser reaches a script and stops building the page until it loads. A few of those and the first screen appears a second later.

3. Third-party widgets

Chat, callback, counters, maps, promo popups. Each pulls its own files from someone else's servers. Five widgets easily add several seconds.

4. Fonts

Loading fonts from external services and downloading full sets instead of the characters you need. A separate problem is text that stays invisible until the font arrives.

5. Hosting

The server thinks for a second before it starts sending the page. If the server itself is slow, no optimisation will save you.

6. A heavy theme or template

Universal templates pull styles and scripts for every possible design variant, including the ones you never use.

The order to fix things in

1. MeasureWith a tool, not by eye, and always in mobile mode. Record the numbers before you start.
2. Deal with imagesCompress, convert to a modern format, serve at the block's size, enable lazy loading below the fold.
3. Remove unnecessary widgetsAnswer honestly which of them anyone actually uses. Usually half can go without any loss.
4. Sort out scriptsEverything not needed for the first screen should load after it.
5. Set up caching and compressionOn the server side. Done once, works forever.
6. Check the hostingIf the server responds in more than half a second, the problem is the hosting, not the site.
7. Measure againCompare with your starting numbers. Without this you do not know whether it helped.

What not to do

• Chase a perfect score at the expense of features the business needs.
• Turn off analytics for speed: without data you are blind.
• Measure only on a desktop with fast internet.
• Optimise without recording the starting numbers.
• Assume speed fixes a weak offer. A fast site priced above market is still a site priced above market.

What to do about images in detail

Since this is cause number one, it deserves a closer look. The work has four steps, and each gives a noticeable gain.

1. Correct dimensionsAn image should not be wider than the block it is displayed in. A 4000-pixel photo in a 400-pixel block is ten times more data than needed.
2. Compression75–85% quality is visually indistinguishable from 100% and weighs several times less.
3. A modern formatWebP or AVIF instead of JPEG and PNG — usually 30–50% lighter at the same quality.
4. Lazy loadingImages below the first screen load when the person scrolls to them, not immediately.

Separately: every image should have width and height set. Without them the page «jumps» while loading, which is a complaint search engines raise on its own.

Fonts: the delay nobody notices

Until a font loads, text is either invisible or shown in a fallback and then jumps. Both options are bad.

• Host fonts yourself instead of pulling them from an external service — one connection fewer.
• Load only the weights you use: regular and semibold, not nine of them.
• Trim the character set: if the site is in Russian and English, hieroglyphs in the file are dead weight.
• Tell the browser to show text in a fallback while the font loads — otherwise the first screen is blank.

How to test properly

1. Measure in mobile mode, not on a desktop.
2. Not just the homepage: inner pages are often slower.
3. Take several measurements — the numbers fluctuate.
4. Look at real user data where available, not only the lab test.
5. Compare with the competitors in the top: what matters is your position relative to them, not the absolute number.