GTmetrix tests a page and hands back a grade, metrics and, above all, the waterfall: the list of everything the page downloaded, in order, with the time of each item. The grade is what everyone looks at. The waterfall is what shows what is actually holding things up. This article teaches how to read the second. For the tools in general, see measuring your site speed.
Before reading: the test settings
| 1 |
Pick the test location nearest your visitors. A test run from the other side of the world mostly measures distance. The options available depend on the type of account on the tool.
|
|
| 2 |
Keep the same settings for every test you want to compare: same location, same device, same connection.
|
|
| 3 |
Test two or three times and keep the middle value. The first test of a cached page is the slowest.
|
|
Reading the waterfall, top to bottom
| What you see |
What it means |
What to do |
| The first row, with a long “waiting” bar |
The HTML document was slow to start answering: that is the server’s time. |
Page cache, database, plugins. See high TTFB. |
| Long bars on image files |
Heavy images. |
Shrink, convert. See image optimisation and WebP. |
| A “staircase”: a file only starts when the previous one ends |
A dependency chain: one script calls another, which calls another. |
Remove what is not needed, combine or defer. |
| Hundreds of small rows |
Too many requests, nearly always from plugins. |
Switch off plugins you do not use. See updating themes and plugins without breaking anything. |
| Rows from other addresses (ads, chat, maps, fonts) |
Third-party code. |
Decide whether it pays for itself, or remove it. These do not depend on your server. |
| Rows with errors (404, 500) |
Something the page asks for that does not exist. |
Fix the address, or remove the request. Every error is lost time. |
Click or hover over a row to see the request details: times per phase (connect, wait, download) and the response headers. That is where you confirm compression and browser caching. The coloured vertical lines mark when the page began to paint, when the largest element appeared and when loading finished.
What to ignore, or read with care
| 1 |
The letter grade. It is a combination of checks, not a time. You can have an “A” and a page that is slow on a mobile network, and a “C” on a page that feels fast.
|
|
| 2 |
The fully loaded time. It includes scripts that keep running after the page is ready (chat, statistics). A page can feel ready long before.
|
|
| 3 |
Recommendations that do not apply. Some concern third-party things or files you cannot touch. Skip them.
|
|
| 4 |
Comparison with another site. Sites are different. Compare yours with itself: before and after.
|
|
|
One test is not a fact. Networks, test servers and traffic peaks make numbers swing. If you changed plan or configuration because of a single test, repeat it first. See what to measure before you change plan.
|
|
Keep the reports. A before and an after report are the proof that a change worked. With no “before”, the “after” proves nothing.
|
|
Understood the waterfall but there is a row you cannot explain? Send us the test address and the row in question.
Open a support ticket
|
RECOMMENDED PRODUCT Web hosting with cPanel Domain and SSL included, daily backups and the panel you already know. from $6.59/mo (3-year plan, with coupon) See plans |
Português
English
