Your own connection, device and cache hide the problem. Test on a phone, on mobile data, with the cache cleared, and the result is usually several times slower than what you have been looking at.

Why your own experience misleads you

You visit your own site frequently, on a fast connection, on a good computer, with everything already cached from last time.

All four of those make it faster than a stranger will experience it, and together they make it much faster.

A first-time visitor on mobile data, on a three-year-old phone, holds nothing in cache and is downloading everything for the first time.

That can easily be five or ten times slower than what you see, which is why owners are frequently baffled by complaints about speed.

The four differences

The second is underrated. Phones vary enormously in processing power, and a page heavy with scripts is far slower on a mid-range phone than on any computer, regardless of connection.

Testing it honestly

  1. Use an actual phone, not a resized browser window.
  2. Turn wifi off so you are on mobile data.
  3. Use a private browsing window so nothing is cached.
  4. Time it, from tapping to being able to read.
  5. Do it somewhere with average signal, not beside the router.
  6. Try the pages people actually land on, not only the homepage.

The third is the one that changes the result most. A cached repeat visit is a different measurement entirely and it is what most people accidentally test.

A worked example

A restaurant whose owner insisted the site was fast, having checked it that morning.

On his phone, in the restaurant, on wifi, it appeared in about a second.

Tested properly, in a car park on mobile data with the cache cleared, the same page took over nine seconds before anything readable appeared.

The cause was a background video on the homepage, several megabytes, loading before the menu.

Removing it and replacing it with a single photograph took the page to under three seconds. Nothing else changed.

The important part is that nobody had ever complained. People had simply left, and there is no record of somebody who waited four seconds and gave up.

What slow actually costs

Difficult to measure directly and consistently real.

People leave. The longer a page takes, the more of them leave, and the ones who leave never appear as anything in your records.

It is worse on mobile because the person is frequently standing somewhere, in a hurry, with an alternative one tap away.

And it compounds with everything else: a slow page that then asks for nine form fields loses at both stages.

Where the weight usually is

In roughly this order for most small business sites.

Images, uploaded at camera resolution and displayed small. Frequently the majority of the page weight on its own.

Video, particularly anything playing automatically in a header.

Scripts, from plugins, analytics, chat widgets and tracking pixels.

Fonts, several weights loaded when two are used.

The theme itself, where a general-purpose template loads capability you do not use.

The first accounts for most of the problem on most small sites, and it is also the easiest to fix.

The counter-case

Where speed is not the thing to work on.

A site that is already reasonable, where further optimisation is measured in fractions of a second nobody perceives.

A business whose visitors arrive almost entirely on desktop from an office, where mobile conditions are not what they experience.

And any site where the real problem is that the page does not answer the question. A fast page nobody finds useful is not better than a slow one.

Speed behaves as a threshold rather than a scale: below a few seconds it stops mattering to anybody, and further effort is better spent on the page itself.

What a visitor sees while waiting

Perceived speed is not the same as measured speed, and it is worth attention separately.

A page that shows text quickly and fills in images afterwards feels fast even if the total load takes longer.

A page that shows a blank screen for four seconds and then everything at once feels slow even if the total is shorter.

Which means the order things load in matters as much as how much there is. Text and the main heading first, images after, anything decorative last.

Most themes handle this reasonably. What breaks it is a large header image or a font that blocks the text from appearing until it has downloaded, both of which are common and both fixable.

What to do first

In order of return for effort.

Resize the images, which is usually the whole problem and takes an afternoon.

Remove anything loading that nobody uses: an old chat widget, a tracking script for a service you left, a plugin from a campaign two years ago.

Turn on caching, which most hosts offer and many sites have never enabled.

Then measure again, because the next thing to fix depends on what is left rather than on a general list.

Speed and the phone version being different

A related problem worth checking at the same time.

Some sites load a different, lighter page on mobile, and some load the same page and simply rearrange it. The second is more common now and it means the phone downloads everything a desktop would.

Which is why a page that looks simpler on a phone is frequently not lighter. The large header image is still being downloaded; it is just displayed smaller or hidden.

Hidden is the worst case: the file is fetched, costs the visitor time and data, and is never shown to them.

Checking for that is worth an hour, because anything downloaded and not displayed is pure waste and usually easy to remove.

Judging the result

By experience rather than by score.

The question is how long before somebody can read the thing they came for, on a phone, on mobile data. Under three seconds is good, under five is acceptable, above eight is losing people.

Scores from testing tools are useful for finding causes and poor for judging whether it is fast enough. A page can score badly and feel fine, or score well and still take six seconds on a real connection.

Trust the stopwatch and the phone over the number every single time. The tool is a diagnostic instrument, not a verdict.

And test again a few months later. Sites get slower on their own, as images are added, plugins accumulate and somebody enables a feature that loads another script. Speed is not a job that stays done.

The version of this that Google looks at is covered in Google is reading the phone version.


Frequently asked questions

Why does my own experience mislead me?

Fast connection, good computer, and everything cached from last time. A first-time visitor on mobile data has none of those and can be five to ten times slower.

How should I test properly?

On an actual phone, wifi off, in a private window so nothing is cached, somewhere with average signal, timing until you can read the page.

Which step changes the result most?

Using a private window. A cached repeat visit is a completely different measurement and is what most people accidentally test.

Where is the weight usually?

Images uploaded at camera resolution, then video, then scripts from plugins and widgets, then fonts, then the theme itself.

What does slow actually cost?

People leave, and they never appear in your records. It is worse on mobile because they are in a hurry with an alternative one tap away.

How fast is fast enough?

Under three seconds to readable is good, under five acceptable, above eight is losing people. Judge by stopwatch rather than by a tool's score.

West Coast Media Solutions Inc. provides web design, web development, hosting, digital marketing, and business consulting to organisations across Canada, drawing on more than twenty-five years in the field.

Convinced your site is fast?

Test it in a car park on mobile data with the cache cleared. The number is usually a surprise.

Start a Conversation