Some of the largest and most successful WordPress sites worldwide, such as the Wall Street Journal or People Magazine, perform enormously poorly in Google's PageSpeed Insights tremendously poorly. This is despite the fact that their business model depends on good performance. Using the New York Times as an example, I'll explain why you can neglect the PageSpeed Optimization Score and what concrete advantage your WP business can gain from this knowledge.

Update: Google changed its PageSpeed Insights tool in November 2018. Since then, the analysis data is based on the open source tool Lighthouse. The new PageSpeed Insights includes even more factors in the evaluation, which is why many websites perform worse in the new PSI score than before. This also applies to our case study - the WordPress website of the NYTimes: its desktop PSI score is now 46 and the mobile score is 21. More information about the new PageSpeed Insights can be found in the video of #SEODRIVEN, which you can also find at the end of this article.

What do the websites of Forbes, Time Magazine, New Yorker, Wall Street Journal, People Magazine and Harvard Business Review have in common? They are all major publications with million-dollar reach and corresponding online revenues. And they all run on WordPress !

You can imagine that performance is a hot topic especially for such big publications. The better the site performs, the better the user signals and the more people read it. This benefits the publication twice over:

• The better the user signals, the more advertising sales.
• The more readers, the better the subscriber count.

The example of the Financial Times also shows that performance pays off. In 2016, they testedhow a delay in the loading time of one to five seconds affects the behavior of readers. The result: the slower the site , the fewer articles visitors read. The result: reduced ad revenue and fewer completed subscriptions. Not surprisingly, load time optimization was a top priority when the Financial Times website was revamped about six months later.

However, if you look at the results that Google PageSpeed Insights spits out for the publications mentioned above, it doesn't look like performance optimization plays a big role at first glance.

All but two of the publications tested achieved a mobile optimization score in the good range (80-100). The desktop score, however, is quite different: The PageSpeed scores of NY Times, HBR and People Magazine are "low", the scores of WSJ, Forbes and Time Magazine are only "medium" and only the New Yorker just makes it into the good range.

What is it about these "bad results"?

The PageSpeed score has nothing to do with loading speed

Many believe that the score given in PageSpeed Insight (e.g. 60/100) indicates the loading speed of site . The name of the tool suggests this. Only: "PageSpeed" and "Page Speed" are not the same thing in this case. The optimization score that the tool ultimately outputs has no correlation to the page load time.

Read correctly: The Google PageSpeed Insights score not the loading time.

Instead, it checks whether the site operator has implemented certain measures that are considered "best practice" in performance optimization. The implementation of these measures is then rated on a scale from 0 to 100. 

A second myth that persists: A good PageSpeed score improves your Google ranking. However, that is just as not the case. Yes, the speed of a site affects the ranking. But the score that the tool gives out is not taken into account by Google (especially since it doesn't correlate with speed anyway). Therefore, you can largely disregard the Google PSI score when it comes to SEO.

In addition, the page load time is not relevant for the ranking, i.e. the time that a site needs to be completely to load. Instead, Google includes the Time To First Byte (TTFB) value as a factor. This is the time that elapses until the browser receives the first response from the server after an HTTP request. As a rule, this is a matter of milliseconds.

The correlation between TTFB and the ranking could already be proven in 2013 (you can find corresponding articles by MOZ here and here). On the other hand, Gary Illyes - a highly respected web trend analyst at Google in the community - publicly announced via Twitter that one should not not to worry too much about page load time. need to.

Case Study: The New York Times PageSpeed Insights Score

Let's take a closer look at the New York Times as an example. It achieves a PageSpeed Insights score of 84 ("good") on mobile and 52 ("low") on desktop. So what does PageSpeed Insights suggest to improve load time? According to Google, the desktop version could benefit from the following measures, among others:

Eliminate Java Script and CSS resources that block rendering in "above the fold" content (visible without scrolling).

The background of this suggestion is that JavaScript should normally be loaded in the footer or at least asynchronously. If it is loaded in the header, this can lead to delays because other resources are put on the back burner. In some cases, however, it significantly affects the functionality of site if all JavaScriptsites is located in the footer. It must therefore be checked on a case-by-case basis how this measure affects site .

And if the CSS resources are only loaded at the end, the whole site is completely without design - not exactly a nice user experience. Of course, it would be theoretically possible to filter out the CSS that is needed for the content "above the fold" and insert it at the top and then load the rest of the stylesheet at the bottom. However, this is almost impossible to do afterwards, this trick would have to be taken into account during development. In addition, it means considerable effort for the developer and ultimately only improves the PageSpeed score, but not the actual page load time. The effort is therefore probably better invested elsewhere.

Use browser caching

Doesn't sound wrong at first. But if you look at the suggestions of what else could be cached, there are elements that are not hosted on the site of the NY Times itself. These include, for example, files that are hosted by Google Analytics or Facebook and are included for monitoring purposes at the NY Times. The site operator of the NY Times has no influence at all on the cache configuration of these elements - so the suggestion is for nothing.

Google also criticises the use of a content delivery network (CDN) - a network of servers distributed around the world but connected to each other. However, international users in particular benefit from this. In principle, a CDN is advantageous for performance, since the response time of the server is greatly reduced and the content can be delivered much faster. And with a publication as relevant as the New York Times, you can assume that readers all over the world will access the content and won't want to wait long.

Optimize images

A large part of the images that PageSpeed Insights suggests to optimize would only be a few kilobytes smaller due to compression, in some cases only bytes. Of course the compressing images is an important factor for performance optimization. However, with such small savings, it's doubtful that it will significantly improve your load time.

Reduce JavaScript

A total of just under 72 kilobytes could be saved here. Whether this makes a fundamental difference for such a huge site as the New York Times remains to be seen.

Reduce HTML

As with the images and JavaScript files, the proposed savings potential is minimal and thus unconvincing.

Some of the measures suggested by the tool are probably simply uneconomical, and others would only entail such marginal changes that their use is not worthwhile. The sobering conclusion: PageSpeed Insights throws up all kinds of suggestions for improvement. Not all of them, however, result in a significant improvement in the performance of the NY Times. Otherwise, we could assume that they would have already been implemented - after all, performance directly influences the success of the business model here.

PageSpeed Insights-score remains a customer issue

In the professional world, the complete discrepancy between PageSpeed score and loading speed has led to a heated debate. After all, the tool is available to laymen who may not be aware of this discrepancy. Posts from respected online marketing gurus like this onewhich talk about a PageSpeed score of 100 being equivalent to a fast loading time, cause additional confusion.

Again and again, developers report calls from highly unsettled customers who complain that everything is red and orange at PageSpeed Insights and order an implementation of all proposed measures. The bottom line is that the tool often leads to time being wasted twice: during optimization, when nonsensical suggestions are implemented, and during communication, when it is explained to the customer why they are nonsensical.

The PageSpeed Insights score may improve if you reduce image sizes and HTML by a few KB. The performance, however, mainly benefits from measures that the PageSpeed Insights tool does not even suggest. In the end, professional performance optimization is more than just orienting on a single key figure. This is also shown by the relaunch of the Financial Times: A comprehensive redesign of site is usually part of larger optimization efforts.

I'm particularly interested in high-traffic areas. Smaller websites should of course first see to it that they adhere to basic "best practices". After a certain threshold, however, massive changes have to be made to the pages in order to increase performance at all, for example, switching to a good host or a fundamental revision of the page architecture. This should be clear to you and your customers.

Arguments for real performance optimization

The uncertainty about the PageSpeed Insights score offers a good opportunity, especially for design agencies: because if you recognize the connection between loading speed and business and know how to use it, you stand out from the competition. Concrete numbers and case studies like the example of the NY Times help you to convince existing and potential customers:

• In 2006, Amazon conducted A/B testswhich showed that a 100 millisecond delay in loading speed meant about 1 percent lost revenue per year - or in other words, $1.6 billion.
• Studies show that in recent years the average attention span of users has decreased from 12 to 8 seconds. So once a website loads for five seconds, there are only three left to convince the user of the content. (The validity of this data is debatedbut you're on the safe side site if you assume that users are spending less time on your content rather than more).
• Especially with mobile sites , the loading speed is highly relevant for business. In e-commerce, the loading time has a fundamental impact on sales: If the site is too slow, more than half of the customers prefer to leave their money elsewhere. 53 percent of users bail out if a site takes longer than three seconds to load on a mobile phone. And for every second that a mobile site takes longer to load, the site operator loses 20 percent conversions. And mobile traffic is not to be ignored: The average time spent on the Internet via mobile devices devices is already at around 87 minutes, and the smartphone has smartphone has overtaken the laptop as the most common Internet device..

How to get your customers to ignore the PageSpeed Insights score

So how do you help your clients properly rank Google PageSpeed Insights and give the tool less importance? Here is a summary of the most important arguments:

• The PageSpeed Score has nothing to do with loading speed, but assesses whether certain measures have been implemented that are commonly recommended. Not all of these measures make sense. You can offer your customers to check them in detail and implement those that you consider useful.
• The PageSpeed Score is not relevant for SEO. For the ranking, the Time to First Byte (TTFB) is included, not the complete loading time. You can measure this value for example with the tool Webpagetest tool. How to correctly analyze the real page load time with Webpagetest, we explain in our e-book.
• The PageSpeed Insights tool only checks "publicly" accessible factors. For example, the tool cannot see how the database is doing (and that's a good thing for security reasons). With a tidy database, a lean Theme, which does not send too many HTTP requests to the server, and as little Plugins as possible, your load time will gain considerably. However, these factors are not taken into account by PageSpeed Insights . So actually performant WordPress sites still get bad scores.
• PageSpeed Insights does not include all performance optimization measures. Make your customers aware of the importance of a good hoster that works with HTTP/2 and the latest PHP version. If the hosting is no good, you can optimize the site as much as you want, the loading time will not change fundamentally.

Focusing only on PageSpeed score is like taking a horse to a Formula 1 race. Even if you dye your horse's coat red and shave a Ferrari logo into its flank, you're not going to outrun the motorized race cars.

Conclusion

Forbes, Time Magazine and the New York Times may not have the most visually appealing websites, but they are among the most successful in the world. WordPress sites worldwide. And that's because design, function and speed work together to promote a coherent overall experience.

However, the PageSpeed Insights score does not reflect this. It regularly challenges developers to explain to concerned customers that their website is not disappear into the depths of the search results if the tool's verdict is "poor". The loading time of a site depends on many factors, and many of them are not reflected by simplistic tools like Google PageSpeed Insights .

Whoever ultimately decides to implement the to implement the proposed measuresshould in any case measure the actual loading time before and after, in order to put the effectiveness of the optimizations to the test (for example with our performance framework).

The bottom line is that the tool points out some standard measures (compress images, use SSL and/or HTTP/2, set up caching, etc.). For a good user experience, however, the design of the sites , the load-time optimized display (which the PageSpeed tool does not measure) and the UX optimization are the most important.

Have you ever had the experience of a bad PageSpeed score? Or do you know the concerned inquiries from customers about this topic? Feel free to leave me a comment with your experiences and tips.

The new PageSpeed Insights