Website speed, search rankings and the orders you never knew you lost

Website speed is discussed almost entirely in terms of search rankings, which is unfortunate, because that is the weaker half of the argument. The stronger half is that slow pages lose customers silently, every day, and nobody ever writes to tell you.

What speed does to rankings

Search engines do use page experience signals, including how quickly the main content appears, how soon the page responds to interaction and whether the layout jumps about while loading. Those signals are real and worth attending to.

They are also weaker than most agencies imply. Relevance and quality dominate. A slow page that is genuinely the best answer to a question will usually outrank a fast page that is not. Speed acts more like a tiebreaker between pages of similar merit, which in competitive fields is precisely the situation you are in.

There is also an indirect effect that matters more than the direct one. If visitors arrive from search, wait, give up and return to the results, that pattern is not a good sign for your page. Speed therefore influences ranking partly through what it does to human behaviour rather than only as a measurement.

The commercial argument, which is much larger

Every study of ecommerce speed points the same direction. As pages get slower, completed purchases fall, and the effect is most severe on mobile connections, which is where the majority of visitors now arrive. The visitors you lose this way never appear in any report as complaints, because they simply left.

For a business system rather than a shop, the arithmetic is even simpler and rarely done. If a screen used two hundred times a day takes four seconds instead of one, that is ten minutes of paid time lost daily for every person doing it. Across a small team that is a working week each month, spent watching a spinner.

What to measure

Ignore the single score out of one hundred that testing tools produce. It compresses several different things into one number and encourages chasing points rather than fixing problems.

Watch three things instead. How long until the main content of the page is visible, since that is what a visitor experiences as loading. How long until the page responds to a tap or a click, since a page that looks ready but ignores you feels broken. And whether the layout shifts while loading, which is what causes people to tap the wrong thing.

Measure them on a mid range phone over an ordinary mobile connection rather than on a fast computer on the same network as the server. The developer view of a website is almost always considerably better than reality.

The fixes that actually produce results

Images are usually the largest single problem. Photographs uploaded at the size they came out of a camera and displayed at a fraction of that, uncompressed, without modern formats, and loaded all at once including the ones far down the page. Fixing images alone frequently transforms a site.

Third party scripts come second. Analytics, chat widgets, advertising tags, review embeds, social buttons and heat mapping tools each add weight and each fetch further resources. Sites routinely carry several that nobody remembers adding and nobody uses. Removing them is free and often the fastest improvement available.

Then the server side. A page that takes two seconds to be generated before the browser receives anything cannot be made fast by any amount of front end work. That delay is almost always database queries, usually the pattern where code issues one query per row in a list, and it is fixed by profiling rather than guessing.

Caching follows, since repeatedly generating a page that has not changed is wasted work. Then fonts, where loading four weights of two typefaces delays text appearing. Then moving heavy work such as report generation and message sending into background jobs so nobody waits for it.

What not to bother with

Buying a larger server before measuring. More hardware masks inefficiency and costs money every month forever, and if the cause is a query running fifty times per page you have simply bought a faster way to run it fifty times.

Chasing a perfect score. The difference between a good result and a perfect one is usually invisible to humans and can involve considerable effort. Get the site genuinely fast for real people on real phones, then spend the remaining budget on something that grows the business.

Keeping it fast

Performance decays. Content accumulates, features are added, somebody installs a new marketing tag, a table grows past the point where a missing index mattered. Sites that were optimised once and never checked again are usually slow again within eighteen months. Periodic measurement catches that while the correction is small, which is considerably cheaper than repeating the whole exercise from the beginning.