How To Optimize Website Speed
Contents
Running a speed test is the first part of your website optimization journey.
Once you have your metrics, you need to know how to interpret them and what to do to improve them.
In the Metrics Overview area of your website speed report, you’ll see the key metrics we’ll focus on to help speed up your site:
In addition, you can use the request drop rate to see how long requests take and how it affects those metrics.
How To Speed Up First Contentful Paint (FCP)
Let’s start by making your website appear faster to your visitors; we’ll cover the First Satisfied Paint, first.
What Is First Contentful Paint?
First Page Satisfaction measures how a page’s content first appears when your visitor lands on that page.
It is important that your keyword content shows up quickly in order to keep your visitor from leaving your website. The faster a user leaves your website, the faster Google learns that the information on the page may be bad.
But how do you really know what is causing your website to load so slowly?
How do you uncover which server issues are slowing down your website? Let’s see.
Why Is My First Contentful Paint Taking So Long?
Your FCP can be affected by server connection speed, server requests, provisioning-blocking resources, etc.
It sounds like a lot, but there’s an easy way to see exactly what’s slowing down your FCP – the waterfall application.
This useful tool shows what requests your website is making and when each request starts and ends.
For example, in this screenshot, we first see a request for an HTML document and then two requests to load the styles displayed in the document.
Screenshot showing debug data for the First Contentful Paint metric in DebugBear, October 2022
Why does the First Paint Satisfy happen after 0.6 seconds? We can pull down what is happening on the page to understand this.
Understanding What Happens Before A First Contentful Paint
Before the first clips can be loaded on your web page, your user’s browser must first connect to your server and retrieve the content.
If this takes a long time, then it will take a long time for your user to see your website.
Your goal is to learn what’s going on before your website even starts to load so you can spot issues and speed up the process.
Page Loads Part 1: The Browser Makes a Connection to the Server
Before you start requesting a website from the server, your browser needs to establish a network connection to the server.
This usually takes three steps:
These three steps are performed by the browser, one after the other. Each step requires a round trip from your visitor’s browser to your website server.
In this case, it takes about 251 milliseconds to establish a server connection.
DebugBear screenshot showing network round trips used to establish server connections, October 2022
Page Load Part 2: The Browser Requests the HTML Document (The First Byte Happens Here)
Once the server connection is established, your visitor’s browser can request the HTML code containing the content of your website. This is called an HTTP request.
In this case, the HTTP request takes 102 milliseconds. This time includes both the time spent on the network round trip and the time spent waiting for the server to issue a response.
After 251 milliseconds to create the link and 102 milliseconds to make the HTTP request, your visitor’s browser can finally start downloading the HTML response.
This special event is called Time to First Byte (TTFB). In this case, that happens after a total of 353 milliseconds.
After the server’s response is ready, your visitor’s browser spends some additional time loading the HTML code. In this case, the response is small and the download only takes about 10 milliseconds.
DebugBear screenshot showing the details of an HTTP request, October 2022
Page Load Part 3: Your Website Loads Some Render-Blocking Tools
Browsers do not render, or display, pages immediately after loading a document. Instead, there are usually some extra-blocking resources.
Most pages look bad without any visual style, so CSS styles are loaded before the page is rendered.
Installing the two additional styles in this website speed test example takes 137 milliseconds.
Note that these requests do not require a new server connection. The CSS files are loaded from the same domain as before and can reuse existing connections.
DebugBear screenshot shows extra-blocking resources being loaded after HTML document, October 2022.
Page Loading Part 4: The Browser Converts the Page
Finally, when all the necessary resources have been loaded, your visitor’s browser can start rendering the page. However, performing this task also takes some processing time – in this case, 66 milliseconds. This is indicated by an orange CPU task marker in the waterfall view.
DebugBear screenshot shows the steps leading from loading the HTML document to rendering the web page, October 2022.
We now understand why the FCP occurs after 632 milliseconds:
Some editing work includes small tasks like running inline text or moving the HTML and CSS code when downloaded. You can see this phenomenon as small gray lines at the bottom of the filmstrip.
How To Optimize First Contentful Paint (FCP)
Now that you understand what leads to your website offer, you can think about how to optimize it.
Now that the first parts of your website are loading fast, it’s time to focus on making the entire site load fast.
How Speed Up Largest Contentful Paint (LCP) With DebugBear’s Recommendations
There are many ways to run your LCP.
To make it easier, DebugBear gives us the main steps to follow in the Recommendations section.
Let’s look at some examples of recommendations and learn how to run this website LCP.
Recommendation 1: Initiate LCP Image Requests From The HTML Document
If the main content item on your page is an image, the best practice is to ensure that the URL is contained directly in the first HTML document. This will help it start downloading faster.
However, this best practice is rarely used, and sometimes it takes a long time before the browser realizes that it needs to download a large image.
In the example below, the main content, which is an image, is added to the page using JavaScript. As a result, the browser needs to download and run a 200-kilobyte script before it can find the image and start downloading.
DebugBear screenshot showing the sequential request chain leading to the image request, October 2022.
How to Fix: Depending on the website there are two possible solutions.
Solution 1: If you are using JavaScript to load a large image, then increase the size of the image and remove the lazy loading script or replace it with the current loading=”lazy” attribute, which does not require JavaScript.
Solution 2: In some cases, server-side rendering would prevent downloading JavaScript before the app page could render. However, this can sometimes be difficult to do.
Recommendation 2: Ensure LCP Images Are Loaded With High Priority
After loading the HTML code of the page, your visitors’ browsers can see that, in addition to your main image, a large number of resources such as styles may need to be loaded.
The goal here is to make sure that your large, large image is loaded to meet the Satisfying Paint dialog that Google wants.
Other resources, such as third-party analytics sources, are not as important as your big picture.
Additionally, most images displayed in your site’s HTML will be below the fold when the page is rendered. Some may be completely hidden in the nested header navigation.
Because of this, early browsers set the value of all image requests to Low. When the page is loaded, the browser determines which images are important and changes the first one. You can see an example of that in the screenshot below, as indicated by the asterisk in the first column.
DebugBear screenshot shows LCP image loaded with low initialization, October 2022.
Waterfall shows that while the browser was aware of the image initially, it did not start downloading, as indicated by the gray bar.
How to Fix: To fix this you can use a new browser called special tips. If you add the fetchpriority=”high” attribute to the img element, the browser will start loading the image from the beginning.
Recommendation 3: Don’t Hide Page Content Using CSS
Sometimes you can check the request for water drop and all the offer-blocking resources are loaded but still, no content page shows. What’s going on?
A/B testing tools often hide the content of pages until test variations are applied to the content on the page. In those cases, the browser rendered the page but all of its content is transparent.
What can you do if you can’t get rid of the A/B testing tool?
How to Customize: Check if you can customize the tool just to hide the content affected by the A/B test. Alternatively, you can check if there is a way to make the A/B testing tool load faster.
DebugBear screenshot shows render filmstrip where content is hidden by A/B testing tool, October 2022.
Monitor Your Site Speed With DebugBear
Want to keep testing your website? Try our paid monitoring tool with a free 14-day trial.
That way, you can check if your optimization process is working and be notified of any activity on your site.
Screenshot showing the website’s running speed in DebugBear, October 2022
The views expressed in this article are those of the contributor.
Which tool is best for performance testing?
About Performance Testing Tools
- WebLOAD.
- Vadivelu Comedy LoadNinja.
- HeadSpin.
- ReadyAPI Performance.
- LoadView.
- Keysight’s Eggplant.
- Apache JMeter.
- LoadRunner.
Which tool handles performance testing? JMeter is an open source tool that can be used to perform and measure diagnostic tests and measure the performance of various services. This tool is mainly used for web and web service applications.
Which is the best open-source tool for performance testing?
JMeter. JMeter is the most popular open source testing tool with almost as many features as commercial tools. You can record test scripts, run large tests, and view various performance metrics in easy-to-understand charts.
Which tool is better than JMeter?
Loadrunner from Micro Focus is a very useful tool. It is one of the best alternatives to JMeter that provides a user-friendly monitoring and analysis interface with colorful charts and graphs. Features: Supports continuous testing.
Which is better LoadRunner vs JMeter?
The performance of JMeter is less compared to LoadRunner. Small details and threats can be identified with this tool, but it is not enough to detect threats in the main application. The LoadRunner is more efficient and faster to detect the current bus in use compared to the JMeter software.
What are the two types of tools used for performance testing?
We have various types of performance testing tools available in the market; some of the most commonly used performance (load) testing tools are the following: Apache JMeter. LoadRunner[HP]
What are the two key ways of testing a website?
Performance testing – Performed to show server response time and performance under different load conditions. Load testing – It is a simple method of testing that is done to understand the behavior of the system under real load.
What are the two methods of testing? These testing methods are usually performed in order and include: Unit testing. Integration test. System test.
Who uses JMeter?
| Company | Dailymotion SA |
|---|---|
| Revenue | >1000M |
| The size of the company | >10000 |
Which is the best open-source tool for performance testing?
JMeter. JMeter is the most popular open source testing tool with almost as many features as commercial tools. You can record test scripts, run large tests, and view various performance metrics in easy-to-understand charts.
Which is better LoadRunner vs JMeter? The performance of JMeter is less compared to LoadRunner. Small details and threats can be identified with this tool, but it is not enough to detect threats in the main application. The LoadRunner is more efficient and faster to detect the current bus in use compared to the JMeter software.
Which tool is better than JMeter?
Loadrunner from Micro Focus is a widely used Load Testing tool. It is one of the best alternatives to JMeter that provides a user-friendly monitoring and analysis interface with colorful charts and graphs. Features: Supports continuous testing.