Articles · Stefan Ylimäki · 13 September 2026

A slow website costs customers, but what does speed really mean?

Nobody has ever said “great, the page loads slowly”. Even so, speed is almost always left undiscussed when ordering a new website, because it sounds like a technical detail. It is not a detail, and you can measure it yourself in five minutes.

What slowness costs in practice

When a visitor taps a link in the search results on their phone, they have given you a few seconds. If the screen shows only blank space or a spinning circle in that time, they press the back button and open the next search result, which is your competitor. They will not send you feedback, and they appear in your statistics only as one visit that ended immediately.

This is most common exactly where a small business gets its customers: on the mobile network, on the move, in a hurry. On a desktop with a wired connection every website feels fast. That is why you should not judge your own website's speed from your own computer.

Measure it yourself, do not believe the sales talk

Google offers the measurement for free. Go to pagespeed.web.dev, enter your website's address and choose the mobile view. You get scores for four areas on a scale of 0 to 100, plus a set of metrics. The same tool works for your competitors' websites, which makes comparison easy. Measurement results for the sites I have made are on the references page.

Three metrics are enough. They are Google's Core Web Vitals, and they describe a page's user experience from three different directions. The rest is the builder's worry.

LCP: How quickly the page's main content becomes visible, the moment when the visitor sees what they came for. 2.5 seconds or less is considered good.
INP: How quickly the page responds when the visitor taps something. It measures the delay between the tap and a visible reaction. 200 milliseconds or less is considered good.
CLS: How much the content jumps around while loading. You notice this yourself when you press the wrong button because the page moved just as you tapped. A value of 0.1 or less is considered good.

The thresholds follow Google's guidance, and they are set where the user experience starts to feel poor. They are not a pass mark and not a guarantee of anything. They are a measuring stick.

IF YOUR WEBSITE HAS FEW VISITORS, PAGESPEED INSIGHTS MAY NOT SHOW REAL-USER MEASUREMENTS AT ALL, ONLY A LABORATORY TEST. INP SHOWS MOST RELIABLY IN REAL TRAFFIC DATA, SO ON A SMALL SITE IT IS NOT ALWAYS AVAILABLE.

Why the score alone is not enough

A single number is easy to understand, and that is exactly why it is trusted too much. Remember three things.

First, the measurement applies to one address at a time. A score of 95 for the home page tells you nothing about the service page that has a map, videos and a photo gallery embedded in it. Measure the pages where customers actually end up.

Second, the laboratory test is run on a simulated device and connection. The result varies from run to run, and real users' experience can be better or worse. If real-traffic data is available, it weighs more than the score.

Third, a good score does not mean the site has no problems. A fast page can still be hard to use on a phone, its form can be broken or its text can be wrong. Ninety points only says that the loading went well.

And the other way round: anything under ninety is not bad. Many well-working sites settle around seventy, because they have some necessary heavy part. So look at the metrics, not just the score: if LCP is under 2.5 seconds and the content does not jump, the visitor hardly experiences the page as slow, whatever the score.

How speed relates to Google

Google uses page experience, Core Web Vitals included, as one signal in ordering search results. It is, however, clearly a weaker factor than whether the page answers the searcher's question. Speeding up a page does not automatically raise rankings, and slowness alone does not explain why a business cannot be found.

In practice speed has two different benefits, and it is worth keeping them apart. Towards the search engine, it removes one possible disadvantage. Towards the visitor, it affects whether they stay on the page or go back to the search results, and that is the side that shows in the number of enquiries.

If a business cannot be found in a search engine at all, the cause is almost always elsewhere. The seven most common reasons are covered in a separate article.

What slows small websites down in practice

Photos at their original size. A photo of several megabytes straight from the camera is shown on a phone a few hundred pixels wide. Properly resized, the same photo is a fraction of the original. Because the largest visible element on a page is most often an image, an oversized image shows directly in the LCP value.
Too many scripts. Every chat window, social feed, carousel, tracking code and cookie tool brings its own code. They do not just load, they also run in the browser, and that is exactly what shows in INP when the page does not respond to a tap straight away.
Poorly implemented integrations. An embedded map, booking system or review carousel widget fetches its content from another server. If that server is slow or hiccups, your page waits for it, and you cannot see the cause in your own site.
Elements with no space reserved for them. An image or ad slot whose size is only known during loading pushes other content downwards. This is CLS, and it is exactly the jumping that makes you press the wrong button.
External fonts. The typeface is often fetched from another server, and it takes its own connection before any text shows. The device's own fonts look good these days and cost no loading time.

What can be done about speed without a new website

If your website is otherwise in good shape, speed can often be fixed without a redesign. Resize the images to the right size, remove plugins that nobody uses, move non-essential scripts to load only after the page, and ask your maintainer to switch on caching and compressed transfer. These solve the majority of cases, and they belong to normal maintenance.

If, on the other hand, the website is built on a heavy theme and ten plugins, fixing it can cost more than rebuilding. You can see it in the measurement result: when all three metrics are clearly in the red and no single culprit can be found, the problem is in the foundations.

Frequently asked questions

What is fast enough?

The practical yardstick is whether the page's main content is visible on a phone in about two and a half seconds and whether the page responds to a tap immediately. If both hold, the visitor does not experience the page as slow. There is no point chasing a more precise target for the sake of a score.

My website scored 100. Is everything fine?

Loading is fine on the one page that was measured. It does not tell you whether the form works, whether the phone number can be found with one tap or whether the content matches what the customer is looking for. Measure the service pages too, not just the home page.

Should I remove the cookie banner or the chat for speed?

A cookie banner cannot be removed if it is needed, but how it is implemented makes a big difference. With a chat, ask yourself whether it actually produces enquiries. If it does not, it is just dead weight. The same goes for social feeds and carousels.

Does the server location affect speed?

It has some effect, but rarely a decisive one on a small Finnish website. Oversized images and extra scripts almost always explain more of the slowness than where the server is located.