January 08, 2018

Why Google PageSpeed scores don’t matter

It seems lately that there's more and more concern over the Google PageSpeed score and how it affects your SEO. 

Last week we had a client who emailed us slightly worried saying that some company had offered a free analysis of their site and found it to be in "poor condition."  Not even a day later, we received a similar email about the Echo 5 site.

Here's why the Google PageSpeed score is not nearly as important as everyone's making it to be and why you can mark those "free audits" as spam.

PageSpeed doesn't measure… speed

Ironically enough, Google's PageSpeed doesn't ever actually measure your site's speed.  What it does is scan for common issues with most site setups and assign rather arbitrary points based on whether you pass that test or not.

Now don't get me wrong, I am a HUGE advocate for site speed.  I've even written about it extensively.  But this service does not measure site speed in any way.

Here's our site measured by PageSpeed.  If you take it at face value, it looks… embarassing.

Now here's another site many are familiar with (Pingdom).  Note that while they also have a somewhat arbitrary point system, the load time and "faster than" are the important items here:

This demonstrates the main flaw in relying too heavily on a PageSpeed score.  Even though we score a 59/100 on PageSpeed, we have a load time of 341 ms and are faster than 98% of sites tested!

So why aren't we scoring at least 98/100?

The PageSpeed suggestions may offer little improvement

Again, the score is based on a ruled point system and not speed itself.  And in some respects it may be more fair since comparing a site speed from say Craigslist (small assets and few images) and a site speed from Youtube (way more assets) would always result in the less media-rich site with the higher score.

The goal is to compare a site to itself so that scores are relative to the site's needs, however, this may be misleading if you don't read carefully.

Problems with the recommendations

Many of the recommendations are flat out wrong.  And this is not entirely PageSpeed's fault as it's just running your site through a quick set of tests, but you need to know how to avoid the items that are out of your control or have little to no influence on actual site speed.

From our recommendations currently, here are the remaining "optimizations" and the issues with each:

Eliminate render-blocking JavaScript and CSS in above-the-fold content

The remaining file for us is our stylesheet (CSS).  While making this non-render blocking is not a difficult task, it will produce FOUT (Flash of Unstyled Content).

To prevent this FOUT, we can extract the styles needed for content above the fold on every page and add them as inline CSS.  But this is a huge pain. 

Is this a valid recommendation?  Sure.  But in the end, is lazy loading our stylesheet at 20.7 kb really going to make a significant difference in page load time?  Probably not.

Optimize images

Images are a great place to start for many sites as they tend to be the largest assets and a great way to significantly cut down on load times.  However, PageSpeed is unable to determine the correct sizes needed, which is where we often see the largest waste (e.g., using a 5000 pixel wide image in a 400 pixel wide container).

The main issue with ours is that many of the image optimization recommendations result in a difference of under 1kb.  

Leverage browser caching

We already do.  Can you guess the file that's throwing this recommendation?  https://www.google-analytics.com/analytics.js

It's possible to "trick" the analyzer into thinking these are cached to get a perfect score, but this again won't affect your real page speed.

Enable compression & Minify CSS

Both great things to do and I highly recommend.  Unfortunately in our case, these were both recommendations shown for third-party assets (Typekit and Youtube) which means we have little control over them.

Should you skip Google's PageSpeed tool? 

Not exactly.  It's still a valuable tool and may give you insight to some low-hanging fruit in terms of speed optimization.

But the real question is "Will making these changes be a good use of my time?"

In our case the answer is a pretty easy "no."  We've already handled the big items that are going to make a large difference and the remaining items may take hours to implement to shave off several milliseconds if we're lucky.

And despite the low scores, we still have a fantastic page load time.

Do PageSpeed scores affect SEO?

From what we know right now the answer is no, PageSpeed scores are not correlated to search engine rankings. 

Plug in some of your favorite top ranking sites and see how they fare.  You'll find that the scores look pretty inconsistent if the PageSpeed score was an underlying requirement.

  • Reddit (63 / 100)
  • Stackoverflow (61 / 100)
  • Amazon (54 / 100)
  • Facebook (55 / 100)
  • Ebay (47 / 100)

Page speed (2 words referring to your site's speed), on the other hand, is a definite factor in your site's rankings.  In fact, Google has been very explicit that this will be a factor, especially in mobile rankings.

Bottom line (TL;DR)

PageSpeed Score ≠ Site Speed

If you haven't already, you should highly consider optimizing your website's speed, especially if the site is taking over 4 seconds to load.  Not only will this improve SEO, but its been shown time and time again to improve customer conversion on websites.

If you're spinning your wheels trying to get that oh-so-sweet perfect 100 score on Google PageSpeed, you might want to revisit how you're spending your time.