How fast does a website have to be before it costs you customers?

Somebody searches for what you do. Your result is right there. They tap it. Then they wait.

You will never hear about the ones who did not wait. There is no missed call, no voicemail, no form that came in half-filled. They went back to the results and tapped the next business down, and as far as your phone records are concerned that customer never existed. This is the quietest way a small business loses work.

The number, and who set it

You do not have to take my word for this one. Google publishes the threshold: 2.5 seconds for the main content to appear, measured on real visitors. Under that is “good”. Past four seconds is “poor”. Those are Google’s words, not mine, and they are the numbers it uses when deciding what to show.

Two things about that number matter more than the number itself.

It is measured on your actual visitors, not in a test. Google collects it from real Chrome users over a rolling twenty-eight days, and grades you on the slowest quarter of them — the 75th percentile. Not your average customer. The one having a worse-than-usual time.

It is not linear in how it feels. One second reads as instant. Two and a half reads as normal. Four reads as broken, and people do not think “this site is slow” — they think “this business is a bit rubbish” and leave. The judgement transfers to you, which is unfair and completely real.

I build well inside 2.5 seconds, because leaving yourself no headroom means the first heavy photograph somebody uploads takes you over the line.

Why the test says four seconds and Google says you are fine

This one catches nearly everybody, and it is worth understanding before you panic at a bad score.

Run any speed test and you get two different kinds of number.

Lab numbers are a simulation. The tool loads your page on a pretend mid-range phone over a deliberately throttled connection, roughly a bad 4G signal. It is harsh on purpose, so that problems show up. Lab numbers of four, six, eight seconds are completely ordinary — some of the best-built sites on the internet look poor under it.

Field numbers are what really happened to real people on real phones.

Google’s 2.5-second threshold is a field number. Applying it to a lab result is comparing two different things, and it is why a beautifully built site can “fail” a test it is actually passing. Google says as much itself: lab measurement is not a substitute for field measurement.

So: if the lab says six seconds and the field says two, you are fine. Believe the field. Use the lab number for what it is good at, which is showing you what is slow rather than how slow you are.

The catch is that field data only exists once enough people have visited. A new or quiet site has none, and then the lab number is all anyone has — which is a reason to read it gently, not a reason to read it as gospel.

What is usually making it slow

In small business sites, it is nearly always the same short list, in roughly this order.

Photographs straight off a phone or a camera. This is the single most common cause and the easiest to fix. A modern phone photo is enormous. Uploaded without being resized, your visitor downloads a picture large enough to print on a poster in order to look at it in a box two inches wide. Ten of those on a page and there is nothing else to discuss.

Page builders. The drag-and-drop tools that make a site easy to assemble ship a great deal of machinery to every visitor to make that possible — much of it for features your page does not use. Convenience for whoever built it, paid for by everyone who visits.

Fonts loaded from somewhere else. Every custom font is another file, and if it comes from another company’s servers it is another connection to open before any text can appear.

Third-party scripts. Chat widgets, popups, review carousels, tracking, an analytics tool, a second analytics tool nobody remembers installing. Each one is code from another company that has to arrive and run. A chat widget alone can weigh more than an entire well-built site.

Cheap shared hosting. Less common than people assume, but real. If the server takes a second to think before sending anything at all, you have lost a second before the race starts.

How to measure your own site, today, for free

Do not take my word for it or anyone else’s. There are three ways, and they take minutes.

The one that matters most: your own phone. Turn off wifi. Open your site as a stranger would — search for your business rather than typing the address. Count. That is the experience you are actually selling. Do it standing outside if you can, because that is where people are.

Google’s PageSpeed Insights. Free, at pagespeed.web.dev. Put your address in and read the mobile tab. The section at the top headed “Discover what your real users are experiencing” is the field data — that is the part that counts. Everything below it is the simulation. Ignore the big score in the circle; it is a summary of the simulation, not a verdict on your business.

The check on this site. I built a free site check that runs fifteen measurements on any address and tells you plainly what came back. It does not ask for your email and it does not give you a score, because a score is an opinion with a number stapled to it.

What is worth fixing, in order

If your site is slow, the order matters more than the effort:

1. Resize the images. Often this alone gets a site most of the way there, and it does not require rebuilding anything.
2. Remove the widgets you do not use. The popup, the second tracker, the chat you never answer.
3. Check what your host does before it sends a single byte. If that is slow, nothing downstream can save it.
4. Then, and only then, consider a rebuild. If the platform itself is the weight, no amount of tuning gets you inside the threshold, and at that point starting again is cheaper than fighting it.

Plenty of sites need step one and nothing else. If yours is one of them I will tell you that rather than sell you step four.

The part people skip

Speed is not only about the people who leave. It also affects whether they arrive.

Google uses how quickly a page becomes usable as one input among many when deciding what to show. It is not the biggest factor — being genuinely relevant matters far more, and anyone who tells you speed alone will move you up the results is overselling it. But when two businesses are otherwise similar, the one that does not make people wait has an edge, and it is one of the very few ranking factors that is entirely within your control.

The short version

Google’s line is 2.5 seconds, measured on real visitors rather than in a test. If a lab score looks alarming, check the field number before you spend anything — they measure different things and only one of them is what you are graded on. Most real slowness is oversized photographs and third-party widgets, both fixable without a rebuild.

If you want a second opinion on what your own site is doing, run the free check — or send me the address and I will look at it and tell you which of the four steps above you actually need.