3 Levels of Page Speed Optimization for SEO

Performance Score and Core Web Vitals Myths

How do you recognize someone who has little or no knowledge of SEO? They proudly show you their high score in Google's PageSpeed Insights tool. I don't mean only the SEO score, but especially the performance score. In 24 years of doing SEO, I have met a few people who spend months trying to achieve a perfect performance score. After they finally got it, they posted it across all their social media profiles waiting for clients to line up because of their incredible skills. Even more people faked that score and also posted it on social media waiting for clients to line up because of their impressive Photoshop or inspect element skills.

The only people they attracted were others similar to them, tech enthusiasts who know very little about SEO. Don't get me wrong, everyone is allowed to start at the beginning. But before you put months of effort into something, please make sure it is worth your time. The same goes for Core Web Vitals. It was understandable that most SEOs were obsessed with the metrics when they first launched. I was one of them. I wanted to see how big the impact of a nearly perfect core web vital implementation would be. It was zero. Nada, nothing. A few months after they were launched, Google confirmed that Core Web Vitals contribute significantly less than 1% to the algorithm.

But that didn't stop some SEOs from caring about them and investing a disproportionate amount of hours and developer time into them. Core Web Vitals might be the main reason why many developers dislike SEOs. SEOs chase developers to implement Core Web Vitals down to the tiniest detail without any meaningful result. Their disappointment is understandable. So, please take page speed seriously, but focus on the basics if you don't want to make enemies instead of friends and if you want to keep your clients instead of getting fired every few months. Page speed is one of many page experience factors. Other important factors are website responsiveness, security, ad load, and intrusive interstitials, better known as annoying pop-ups.

The Main Issue You Should Be Concerned About

They all matter, and page speed is also a major user experience factor. Studies show that loading times above 5 seconds are not tolerated by most users anymore. They close the website when that happens. In recent years, that tolerance has shrunk even more, and it's now around 3 seconds. If your website doesn't load in 3 seconds, you can expect the potential customer to be gone, often for good. Even returning visitors do not tolerate much more. Here is the main issue you should be concerned about. The loading time that matters is a user's loading time, not yours, and not the one Google tools like PageSpeed Insights, Lighthouse, or even the Googlebot measure.

Why? Because they measure the current speed of a single URL you provided from their point of view. PageSpeed insights runs on powerful servers with great network connections. They throttle bandwidth to simulate a user experience, but they aren't the user. Google's Lighthouse plugin does the same from your device. So depending on how slow or fast your connection and hardware are, the results will vary. Even the Google Search Console is only an indicator of average loading times as experienced by the Googlebot, not by actual users.

So, the first level of page speed optimization is measuring with Google tools and taking their recommendations seriously. Don't get me wrong, if your website has awful loading times and you implement recommendations from those tools, your page speed will improve, so will the rankings, but that's usually not enough. Those tools mainly focus on image sizes, the number of JavaScript and CSS files, and opportunities to minify them. There are many ways to optimize these areas. We will get to them later. The second level of page speed optimization is measuring with tools like webpagetest.org where you can choose the test location, connection bandwidth, browser type, and browser dimensions. That sounds more like a real user, does it not?

The Hidden Bottlenecks Slowing Your Site

So, let me ask you - what is the average connection and display resolution of your users? You can learn that from analytics tools. But the current global average is a DSL connection and a 1366 by 768 pixel resolution for desktop devices and a 4G connection and a 360 by 800 pixels resolution for mobile devices. You should always choose Google Chrome as the browser when using webpagetest.org because it is the most used browser and the Google bot runs on it. If you run the test with those settings, you will get much worse results than with the Google tools because the test is closer to reality though still not the same as true user experience.

The report you get is great. You see a waterfall charts of your website loads starting with the time to first byte (TTFB) and then every single file that loads. You can see which images are too large, how many JavaScript and CSS files are being loaded and that blocks the page from loading fast. If you look carefully at the report, you will see the number of requests needed to load your page. That metric is often forgotten. Although it is very valuable once you know what it represents and how to use it for optimization. You will notice that websites with fewer requests load faster. For example, a basic WordPress installation with a popular theme can have over 200 or even 300 requests while an optimized theme can have 30, 20, or even fewer requests.

That number is the count of files retrieved during the loading process. Far too many websites load files they don't use or haven't bundled multiple JavaScript and CSS files into one. You should go through them one by one and decide what you need. A common mistake is loading fonts you don't even use. They are heavy. Remove them. JavaScript and CSS files should not only be bundled when possible, they should also be minified. Images are often the heaviest and slowest element. You should load them only in the sizes needed for specific breakpoints and compress them with tools like TinyPNG. Use the right format. JPEG for photos, PNG for graphics with sharp edges and transparency, and WEBP for both cases if you decide to use it.

If you want to do it perfectly, generate each image in multiple sizes and deliver the perfect size for each resolution. You can achieve that with Cloudflare or with a more expensive solution like Akamai if you are an enterprise company. Even then, it's not enough. First, you're still measuring the wrong page speed. And second, you have not optimized the most important part - database queries.

Importance Of Optimizing Database Queries

So the third level of page speed optimization is primarily based on measuring server-side loading times which is the closest you can get to real user loading time. You can measure them using an application performance monitoring tool APM by using your server logs from NGINX or Apache or by adding application level code also called middleware. APM is the easiest and most powerful method. APM tools are specialized services that automatically instrument your application to collect and visualize performance data. How does it work? You install a small agent in your application. It monitors requests, database queries, external API calls and other operations and sends the data to a dashboard. It is usually a paid service. Some well-known providers are New Relic, Datadog, Dynatrace, and Sentry.

Enterprise companies often use modules for NGINX, HAProxy, or Akamai for similar purposes. With this setup, you get real measurements of how long it takes users to load your site. You can even look up load times by page type, for example, homepage, category pages, product pages, filter pages, brand pages, and blog articles. You should do this to find speed problems by page type. Focus on the number of database queries and how long specific database queries take to execute. That is the real gold mine. Optimizing database queries is not easy, but sometimes it is necessary. If you're lucky, the problem will be obvious for the developers. You just need a strategic view from an SEO perspective on these issues.

Meaning, you should be able to identify the problem, but not to provide a solution. Sometimes you should decide not to load something initially and instead use lazy loading. It can even happen that you decide not to load something at all because it's too heavy. This level of optimization is suited for experienced senior SEO specialists who have a deep understanding of technical SEO and a basic understanding of programming and servers. Page speed optimization gets even more complicated with JavaScript frameworks where you need to choose between server-side rendering and client-side rendering. That is not covered in this lesson. If you optimize database queries after completing the basic page speed steps, you have done a really good job.

Page Speed: The Final Goal

Remember page speed optimization can be a never ending process. Your primary goal should be to measure the loading time properly and get it under 3 seconds for every page type, at least above the fold on mobile device. That is what most users want to see first when they open an URL from your site. After you achieve that, keep optimizing and try to reach a 1 second loading time. If you come close, that's fine. There is no need to be faster than one second. If you hit 1 second, let it be and give yourself and your development and devops team a pat on the back.

Results depend primarily on your team's knowledge, but also on budget and very often on the server. If you have a slow server and even loading robots.txt has a TTFB longer than 300 milliseconds, you need to upgrade your servers or change your hosting provider. You see, this is a never ending topic. Therefore, I will stop here. Hopefully, you now understand how to measure page speed right and what the major steps are to make a website faster. Remember that you shouldn't waste time on Core Web Vitals and instead focus on page speed, which Google will gladly reward with much better rankings and the users will appreciate it a lot. So your conversion rates will in most cases be much better and have an great impact on organic revenue too. Watch this video next to learn how to impress Google even more with a much easier fix. See you next time!