How Brian Dean's Backlinko acted on Core Web Vitals

How Brian Dean's Backlinko acted on Core Web Vitals

I wrote about the Backlinko results in November 2019 on LinkedIn, tagging Brian as well. But it looks like Brian already put some effort into the pagespeed of his website. As a result, he is now passing Core Web Vitals.

This was the LinkedIn post back then.

Backlinko lab data changes

Why I was surprised? Well, in November Backlinko had a Time to Interactive (TTI) of 44 seconds and a SpeedIndex of 25.9 seconds. These values improved to a TTI of 14.1 and a SpeedIndex of 6.0 seconds.

Both are lab data results, so not as insightful as field data. Unfortunately, most metrics together with Core Web Vitals were introduced later, in May 2020. So, we only have limited data to compare.

Backlinko field data

Besides passing Core Web Vitals, Backlinko's field data metrics, being First Contentful Paint (FCP) and First Input Delay (FID) also improved by a lot:

  • First Input Delay improved from 642ms to 41ms.
  • First Contentful Paint improved from 5.3 seconds to 2.2 seconds. It is the only one of four metrics which isn't in the green area, as it should be around 1 second.

However, two things are important to know:

  • PageSpeed Insights is now using the 75th percentile to display the field data metrics. This used to be the 95th percentile in November 2019. Google changed this to prevent that data will be overly impacted by outliers;
  • Despite not having a green FCP metric, Backlinko is still passing Core Web Vitals as the FCP metric isn't part of the three metrics that make up Core Web Vitals (being FID, LCP and CLS, but not FCP).

Backlinko obviously improved by a lot, but not as much as this screenshot might tell you as percentiles that are used to display a fixed number was lowered by Google.

When Backlinko improved their site

The above is data as provided by Google, or Chrome User Experience Report (CrUX for short) to be more precise. When using PageSpeed Insights or Google Search Console, you will get it for free. Just like when you would use different tools to analyze the Core Web Vitals of your online competitors.

Using the same data, we can see that Backlinko addressed the Core Web Vitals of their site quite recently. Chances are they started doing so in May when Core Web Vitals were announced, to push changes to production in August.

While only 45% of visitors had a good LCP in July, being below 2500ms, they changed the amount of visitors having a good experience to an impressive 79% in August.

Backlinko isn't safe yet

The origin summary LCP of Backlinko is now 2.3 seconds, while the threshold is 2.5 seconds. This means they are passing Core Web Vitals, but only slightly. Chances are they might not be passing the Core Web Vitals anymore from one day to another. For example when:

  • they would roll out new features;
  • more people would visit their website under poor condition;
  • more users would use a mobile device, as still 51% of users are using a desktop when visiting;
  • or if global warming is real and devices start to perform worse under warmer circumstances.

Moreover, there are no 2G users left, while the website did have 2G users before. When they would come back as well, averages may fall again. Obviously just by a little, as the amount of visitors on 2G will be low anyway.

And if more visitors from African countries would visit his site, numbers may decrease as well. At the moment, visitors from Algerio, Kenya, Morocco, South Afrika are way below averages, impacting the origin summary of backlinko. But once again only slightly, as most visitors might come from other countries.

What Backlinko did to improve

Backlink used to use Wordpress. They now switched to Next.JS, while still using Wordpress as their CMS. Although Next.JS is a JavaScript React framework, they are still doing server side rendering.

2020/12/16 Update: Brian Dean now wrote a bit about the upgrade in a new post

First Contentful Paint

Moreover, there are no script-files left in their head. On top of that, they inlined CSS. Fair enough, inlining all of their CSS might not have been the best idea, but they did succeed in getting rid of any render blocking resources. Instead of 40%, it is now around 60% that have a good First Contentful Paint experience. LCP improved from 45% having a good experience to 75% having a good experience. It is likely that they improved asset priority as well.

Time to First Byte

However, using the same CrUX data, one can see that the Time to First Byte almost changed in the same way, showing the same trend over time. This indicates that they did not only do frontend changes, but also addressed server side matters.

Most likely they improved hosting settings, server side rendering and/or implemented a better server side caching strategy.

Get your own CrUX report

Want your own report? This only works when your site has enough visitors. You can use my bot to get your own report. Would you rather have a quick insight into your online competition? Start using one of the tools as summed up in my previous article.