Everybody has run a speed test on their own site, seen a coloured score, and been none the wiser. The problem is not the tool: it is that nobody explains what each number measures, so people end up trying to raise the score instead of making the site fast. Those are not the same thing.
This article is about reading the numbers. What to do afterwards is in what to measure before you change plan and in what actually makes a difference.
The tools, and what each is for
| Tool |
What it is for, and its weakness |
| Your own browser’s tools |
The most honest of the lot, because it measures from where you are. It shows every file, its time and its weight. It is where you find the culprit. Weakness: it measures only your connection and your computer. |
| Online tests that give a score |
Good for comparing before and after a change, and for a list of suggestions. Weakness: they run from somewhere in the world that may not be where your visitors are. |
| Tests that film the page loading |
They show, in order, what appeared to the visitor and when. It is the clearest way to understand why a page feels slow even when the numbers look fine. |
| Real visitor data |
Some tools show measurements taken from actual visitors over weeks. That is the truth, but it arrives late and only exists if the site has enough traffic. |
| A request from the command line |
It does one thing and does it well: how long the server took to start answering, with no browser in the way. |
The numbers that matter, in words
| Number |
What it really measures |
| Time to first byte |
How long the server took to start answering. If this is big, the problem is what runs on the server: PHP, the database, missing cache. If it is small and the page is slow, the server is fine and the page is heavy. |
| When the largest element appears |
How long the visitor waits for the main part of the page to show up. On most sites the largest element is a photograph, which makes this, nearly always, a number about images. |
| How much the page jumps while loading |
It measures the page moving under the reader’s finger. Nothing to do with speed, everything to do with annoyance. Usually images with no declared size, adverts, and fonts that load late. |
| How long until the page answers a click |
If the page appears quickly but stays stiff for a few seconds, that is too much JavaScript running at once. |
| Number of requests and total weight |
The two simplest numbers and among the most useful. A page asking for hundreds of things is loading too much, and that is nearly always plugins. |
The measurement worth doing, in five minutes
| 1 |
Open the page that matters, not the home page. The home page is almost always cached and will lie to you.
|
|
| 2 |
Open the browser tools on the network tab, switch on the option to bypass the cache, and reload.
|
|
| 3 |
Load twice. The first visit to a cached page is always the slowest, because it is the one that builds the page. Measuring once, changing something and measuring once again leads to wrong conclusions with remarkable ease.
|
|
| 4 |
Look at the first row, which is the page document. Its time is effectively the server’s time.
|
|
| 5 |
Sort the list by size. Whatever is at the top is what weighs. Nine times out of ten it is images.
|
|
| 6 |
Write the numbers down. It feels silly and it is what separates people who improve their site from people who go in circles: with no before, the after means nothing.
|
|
|
Test on a mobile connection too, on a phone. Most of your visitors are in that situation, not on your office line. A page that is pleasant on a computer can be unbearable on a phone, and you only find out by looking.
|
Three traps in reading the results
|
A score is not a time. A score is a sum of several measurements with weights chosen by whoever built the tool. You can raise the score without the site getting faster, and make the site faster without the score moving. Optimise the time, and let the score follow.
|
|
Where the test runs from changes everything. A test run on another continent mostly measures the distance to there, not your site. If your visitors are near you, pick the nearest test location in the tool, or give more weight to what you see in your own browser. On where our machines are: where our servers are.
|
|
One measurement is a sample, not a fact. Networks vary, machines have peaks, and the same page gives different numbers minute by minute. Take three measurements and use the middle one. A single measurement is quite enough to make somebody change plan who did not need to.
|
And before you look at the site at all, rule out the simplest possibility: open it on another network, on a phone using mobile data. If it is fast there, the problem is between you and the site, not in the site. The other tests of that kind are in what to measure before you change plan.
|
Measured it, wrote the numbers down, and still not sure what to do with them? Send us the numbers and the page.
Open a support ticket
|
RECOMMENDED PRODUCT Web hosting with cPanel Domain and SSL included, daily backups and the panel you already know. from $10.00/mo See plans |
Português
English
