Level 6, 122 Walker Street, North Sydney 2060  |  (02) 9993 3300  |
Posted 01-02-2017 by Mitch Brown

Google Analytics

No one can deny the ubiquity and importance of Search Engine Optimisation (SEO) in the modern web. Having the world’s most beautiful, functional and interesting site doesn’t mean much when no one can find you!

In the rush to optimise every keyword, craft the perfect META tags, manage your redirects and generally out-do your competition, it’s tempting to turn to online analysis tools as the be-all and end-all of your SE, treating their automated scoring as gospel. Today we’re going to have a look specifically at Google’s PageSpeed Insights tool (https://developers.google.com/speed/pagespeed/insights/) and answer the question of how useful hitting that coveted “green score” is in technically optimising your website SEO.

As developers, we are often engaged with SEO companies who consult on client projects, and often brand manager, marketing staff and other stakeholders become heavily involved in the development process. Quite often, the topic of PageSpeed Insights rankings comes up, with a strong push to hit the optimum green ranking levels on Google’s tool. PageSpeed Insights will analyse your website and give it a grade based on a number of rules and recommendations (most of them eminently sensible) for improving your website’s performance e.g., compressing images, ‘minifying’ code etc.

This can create an extremely difficult balancing act between matching aesthetic design, site functionality and hitting these PageSpeed Insights targets. And as a set of guidelines and recommendations, PageSpeed Insights is an important part of your SEO analysis but, in our view, “hitting the green” is often not as important as people have been led to believe.

Obviously, page load speed is incredibly important to your visitors. A webpage that takes too long to load may leave them frustrated and quick to close their browser tab in favour of an alternative site.  Google too, uses page load times as an important metric in ranking your site. While content is most certainly king, out of two identical sites with similar levels of optimisation, external links, the faster loading one will likely have an edge in the rankings.

However, it is important to note that when crawling your site, Google does not use your PageSpeed Insights score to determine if your page is fast to load. The GoogleBot will simply record the actual load time of your site as it crawls your site.

It is actually quite possible to somewhat game the PageSpeed Insights score by hitting all of their recommendations, while still having an incredibly slow site. Why is this?

Well, because, while in general they will boost the load time of a site under Google’s assumptions of how your website looks, acts and is coded, a lot of the metrics they use to determine your score are extremely flawed under various scenarios.

Our first example comes from the complication of the rising practices of using CDNs (Content Delivery Networks), to deliver commonly used components or “libraries” like jQuery. These CDN services can greatly improve performance of your website by taking the load off your own hosting and leveraging the powerful bandwidth and server distribution system. 

Sounds good right?  Well, the thing is that because these libraries exist off-site, you don’t have any control over a number of the metrics that PageSpeed Insights uses. A classic, and rather amusing example, is the embedding of Google Analytics from Google’s own CDN, will cause warnings and score reductions like the below.

Save Browser caching

That’s right, Google PageSpeed Insights is complaining about HTTP headers for its own code for Analytics that is served (by their recommendation) from their own servers! A developer cannot fix this but it will negatively affect your PageSpeed Insights score. This sort of warning flag can occur on all manner of third party CDNs and libraries, including common components like jQuery, Angular and Bootstrap from Google’s own servers!

The other recommendation that causes massive headaches for us as developers is the dreaded “Remove Render-Blocking JavaScript and CSS” warning. In general, to decrease the load time of your page, it’s best to make all your JavaScript and CSS load last so all the content of your page loads before the code that formats the content and makes the page look pretty. This prioritises the “meat” of your site to hit the browser first. Sensible right? 

Well it is, until you realise that there a lots of common site components like image sliders that may either break or cause unwanted rendering of hidden content if the code for them doesn’t load first. An example – we have an image slider. The code automatically forwards a list section on the page of 5 photos into an image carousel. If the code loads first, everything looks as it should. If the code to format the carousel loads last? Well, we may end up with those five images just loading in a long list on the page, potentially blocking the page content and just making things look ugly until the JavaScript kicks in at the bottom of the page. This becomes more of an issue the more media and content-heavy your page is. In this case, it makes no sense for the user experience to load the slider code last, but that’s just what PageSpeed Insights suggests you do!

Obviously, smart developers will be able to find ways around the above example but, at the end of the day, it’s often not worth spending the amount of additional coding time necessary as, it’s actually the real-world page load time that matters to Google – not the arbitrary score it gives you in PageSpeed Insights.

So, how do we check how the website actually loads? There’s plenty of tools for that, our favourites being Pingdom - https://tools.pingdom.com/ and GTMetrix - https://gtmetrix.com/. These will actually give a more accurate reflection of the speed of your site.

So, does that mean PageSpeed Insights is useless? Absolutely not. It’s a fantastically useful tool for flagging potential problem areas such as un-minified code, uncompressed images and web server settings. But, chasing the spectre of green or perfect rankings in the tool is an often wasted exercise. Our recommendation? Use it as a guide, not a hard measure.  Carefully review each warning and recommendation and determine if actioning them is firstly, possible and secondly, going to actually improve the page load times and check your real-world scores.

If you'd like to discuss your software development options with The Bridge at no cost or obligation, feel free to call Lawrence on 02 9993 3300 or email lawrence@thebridgedigital.com.au