Why Page Speed Scores can be Dangerous

All of the popular speed testing tools typically provide a page speed score along with their objective results. Google PageSpeed Insights has a their “Speed Score.” Pingdom has a “Performance Grade.” WebPageTest has their five A-F grades. While these do have a purpose, most people use them incorrectly, in a way that can be dangerous to your real site speed.

The reality is that these scores are estimates of how well your site implements performance best practices. These best practices can make your site faster, but they don’t always. And they usually come at a cost of higher maintenance for your development team if not done properly.

What I’m going to show you is that optimizing towards a perfect score on these rating systems can be dangerous to your site’s speed (and your sanity).

Chasing the ‘100’ from any tool

A large majority of clients that come to us for help with page speed optimization start off with the same question:

“Can you help me get my PageSpeed Insights score higher?”

Generally, this is the wrong way to look at the problem. Who is your website’s users – Google, or your real customers?

Don’t get me wrong, we do want Google to look favorably upon our site’s site speed because it is a major component of Search Engine Optimization, but Google takes only the actual speed of the site into account. Many people assume that Google takes the score from their Insights into account, but that’s simply untrue. You can see then why chasing a specific score is dangerous.

Below is an example from Pingdom’s page speed optimization score. You can see that the owner of this site did a great job optimizing their site – cutting the size in half, lowered the number of requests from 90 to 44 – and as a result, their load time went from 1.52 seconds to .27 seconds! But their “performance grade” somehow went down. This is why it’s dangerous to chase any specific score. The information under the score tells us that this site now loads faster than 99% of websites, but for someone who is chasing the elusive grade of 100, they are frustrated for no reason.

Wasted Effort

Another consequence of “optimizing for score” is a serious waste of effort. Take this example of improving a pagespeed insights score, from 78 to 99, and the effect it had on real page load time:

An increase in PSI of over 20 points, and what to show for it? TTFB, Speed Index, Page Interactive, Document Load, and Page Complete are all unchanged.

Hours of work getting that score up to 99, yet no difference will be made from the perspective of our website users.

PSI Score Reality

Another reason no one should buy into thinking you must have a score of 100, is that often it’s rarely possible in the real world. If you use analytics software from a third party, like Google Analytics, these will show up as marks on PageSpeed Insights. Yes, Google is marking us down for using their own analytics tools.

If your website was built using WordPress or another CMS, like 35% of all websites are, this is even more apparent, as every third-party plugin or widget you use may throw a red flag to Google because of their caching settings. This is not something you have any control over, and in the vast majority of cases, is not something that your website’s visitors will ever notice. Sometimes it may be a good idea to abandon these third party tools that have poor performance practices, but you have to weigh the cost with the benefit they provide.

In addition, Google Insights will often mark users down for images or Javascript that can only be optimized down by minuscule amounts (1 or 2%). Is that really the best use of your time and effort? Perhaps if you’re the front-page of the internet, but most likely we’re not.

Since the grading algorithm is unknown to prevent developers from cheating the system, it’s unknown how much your grade goes down for something that may have no noticeable effect on how fast your site loads.

PageSpeed scores change over time

Google’s own John Mueller has gone on record saying your PageSpeed score can change without any changes being done to your site.

Sometimes the algorithm changes, but often there are just general inconsistencies in how a site loads on a case by case basis. Even if you were to achieve a “perfect” score, next week it may very well be gone. This is a frustrating and unnecessary goal to chase.

So What Is PageSpeed Good For?

Google’s PageSpeed tool can be very helpful, as long as you know how to use it, and what is actually important. PageSpeed Insights does help with identifying performance issues on any given website, provides suggestions on a webpage’s optimizations, and suggests overall ideas of how to make a website faster.  It can alert you to many issues that do need to be fixed for your overall site health. Sometimes, it will point you towards images that are way too big and are hindering your site’s performance. In other cases, it can be a good way to see if a new plug-in slows your site down an excessive amount so that you can look for a faster, leaner alternative.

So it’s important to look at PageSpeed Insights as one of many tools that you can utilize to get ideas on improving your overall page speed. Furthermore, your goal should always be to improve the page speed that your users experience, NOT to chase any “PageSpeed Score”.

Being a Hypocrite

What’s silly about all this is that even here at MachMetrics we provide you with a grade on your urls:

Why do we do this if speed scores are the devil?

I don’t want you to miss the point here – they’re not the devil. They are useful. Chasing and optimizing around only them is the devil.

Our speed score is still useful because it:

• allows you to see changes at a glance
• gives you a summary that’s digestible for non-technical folks
• is based on real loading times, and is objectively calculated based on percentiles

How do I utilize site speed scores properly?

• Use PageSpeed in conjunction with other tools, such as Pingdom, WebPageTest, and MachMetrics to get a complete picture of the performance of your site.
• Thoroughly read the suggestions that these tools offer you. If you aren’t sure what they mean, research the suggestion. If it’s something you can fix, great! If it’s not something you have any control over, then move on.
• Always test the recommendations. Never blindly implement a best practice because it is supposed to work. Perhaps your site is different and it isn’t worthwhile for your situation. Perhaps it gets implemented incorrectly. Test before and after, or better yet – monitor your site speed at all times.
• Always focus on the speed the end user experiences, rather than worry about what any website’s “grade” is for your site.

Going forward with your site

Hopefully I’ve convinced you not to chase a perfect score, but instead to look at scores as indicators and helpers towards the real goal: optimizing towards actual browser load times.

I think most of the webperf community would agree:

Lowering your load time from 2.0s down to 1.5s is way more impressive than increasing any score from 91 -> 94.

Go make your site actually faster.

In John Mueller’s words “Use these tools to find ways to improve your site for users, don’t see them as the final goal.”