Fuck PageSpeed

Posted on

Note: Content might be outdated!

Profanity warning: I’m fucking fuming.

As customer support rep for a popular WordPress caching plugin (not linked in order to not harm my employer by language used here), I see people losing their shit about not scoring high enough in Google’s PageSpeed shit show on a daily basis.

It’s excruciating.

If tools like PageSpeed Insights or Test my Site with Google required a developer login, so they got used by people who actually know how to read those darn arbitrary, entirely context-lacking results, and still make good use of them.

But not Google, no. Google has its tool scan your real life problems and generate replies from a pre-defined catalogue of a dozen items you have either succeeded, or failed against.

And because Google is the same Google whose search algorithm separates winners from losers, people think a low PageSpeed score will make their lovely website rank bad.

IT!!! DOES!!! NOT!!! FOR!!! FUCKING!!! FUCK’S!!! SAKE!!!

At Google they’re actually smart enough to not have their own random commentary on your opus optimized-by-blood-sweat-and-tears have any relevance for your ranking.

Yes, your website should be fully responsive, ideally “mobile-first”. But that’s it. Render-blocking shit won’t do any harm to your ranking, nor will missing browser caching on Google’s own servers. (Have you noticed they actually ask you to turn on browser caching on their own servers? It’s hilarious!)

Yet, Google has the nerve to keep calling their monster “PageSpeed Insights”—as if actual speed (i.e. load time) had anything to do with it.

PageSpeed doesn’t even measure fucking load times!
(And it doesn’t for good reason, but that’s an entirely different story that won’t help you come to grips with performance.)

PageSpeed Insights update January 2018

After its recent update, PSI now displays “speed data for popular URLs that are known by Google’s web crawlers”. This is not to be confused with load times as displayed by tools like Pingdom, or GTmetrix.

Obviously, performance optimisation (as in: the kind of shit developers do, because they’re developers) usually should focus on different metrics than load time.

However … (sips coffee)

The context of this post is not general standards of performance optimisation.
The context of this post is the amount of fucks you should give about a PSI score when you’re not a developer, and you don’t know shit about all the technical wizardry happening when a browser loads a page from your self-configured WordPress site.

In which case I’d argue you’re still better off focussing on testing load times as your relevant metric. Because WordPress caching plugins tend to improve that metric by default, whereas optimising your JavaScript and CSS can quickly lead you down a rabbit hole you shouldn’t be entering without a clear idea of what you’re after and sufficient technical skills to achieve it.

And now on with the post …

If you don’t fully understand how to read and hack code in order to please Google, here’s a quick way for you to check how your website is doing for your visitors:

Testing your WordPress website’s performance

  • Go to Pingdom Tools.
    (Developers, shut it now! You may know better, but not everyone can make sense of WebPagetest.)
  • Enter the URL of your front page.
    If it’s HTTPS, enter https://.
    If it is www, enter www.
    Best you copy-paste it from the browser’s address line.
  • Pick—this is important!—a location nearest to your most likely audience for “Test from”.
    If your site speaks any other European language than English, Pingdom currently offers you a testing location from Sweden. Pick that. UK, you too.
    If you’re in the Americas, pick an appropriate location near you from there.
    If you’re elsewhere in the world, you’re in bad luck at Pingdom, but just pick anything approximately near (like, for Indonesia try Melbourne).
  • Run the test.
    Then run it again, from the same location.
    And again.
  • Look at the load time.
    Don’t look at the “Performance grade”, it’s generated via PageSpeed’s API, screw it.
    Look at load time only for now.

✅  If your website is a self-configured, self-hosted WordPress site with a purchased theme and anything up from 5 active plugins, and load time is below 2 seconds, rejoice. 👍
If it is below 1.5 seconds, cheer loudly, grab a cold beverage, and celebrate! 🍻
If it is below 1 second, songs will be sung about you. 👑

🚫  If load time is above 2 seconds, scroll down the Pingdom page, sort that “File requests” chart by “Load time”, and get ready to deactivate or replace some plugins (or a theme) that may send costly external requests to slow-performing services. (Hello, Finstabook!)

There you go. Free performance optimization advice: Unless you’re a developer who actually knows how to “eliminate render-blocking JavaScript and CSS in above-the-fold content”—fuck PageSpeed, test load times.

Update August 2017

Interested in more facts, less rant? Read this one from MH Themes:
PageSpeed Insights – How important is a good performance score for your website?

Also, this commenter bothered to actually phone Google and get them to clarify on whether or not your PageSpeed Insights score is relevant for your search ranking. Thanks, MickyLaPaz!

Update July 2018

Here’s a great post that illustrates the limitations of PageSpeed Insights:

Derek McBurney, Prioritizing visible content for Google PageSpeed Insights made my responsive Jekyll website slower

42 reactions on “Fuck PageSpeed”

  2. Well then I´m more than happy to listen to the many songs that will be sung ´cuz I´m at 600ms even with a purchased theme an 5+ plugins. Freakin´awesome. Thanks to your employer 😉

  3. Morning Caspar,

    I just pray that Joe Bloggs WordPress user is reading this post too. There is so much confusion out there regarding what data PageSpeed does and does not provide.

    However, even though I’m not a developer, I must say that in our experience at RAIDBOXES Webpagetest.org is the better choice for testing a website’s speed. Especially due to the “Visual Comparison” functionality => https://www.webpagetest.org/video/

    If you want to determine how fast a website loads, you will need a point of orientation. That can be a fix value like 2.0 or 1.5 secs. Or it can be a before-optimization-version of a website, a competitor’s website or an internal benchmark. In such cases Webpagetest can provide very neat and vivid data.

    Of course, I must admit that if you are using Webpagetest for the first time, it will be a little bit of pain learning the vocabulary (i.e. visually complete, document complete etc). But once you did this Webpagetest will deliver better data than pingdom.

    Anyhow, great post and I really hope, that the average user will become more aware of this issue 🙂

    Cheers from Münster and have a great week!

    1. Hey Jan, appreciate the comment! Once a person is ready to invest the time to learn how to use any of these tools effectively, the problem actually vaporizes. When used appropriately, a tool like PageSpeed Insights can provide value even when its results aren’t as detailed as WebPagetest.

      The people I’m worried about, though, aren’t those who’ll take the time learning, but those who—for whatever possible reason—don’t understand there actually is something to learn in the first place. That type of person (DIY site owners mostly) will make assumptions like the ones roughly outlined in the post above (“Google = ranking” and “less than 100% = bad”), often to wind up wasting hours or days with chasing a “performance grade”, while their website already performs as decent as a self-configured WordPress site will get.

      Google themselves fuel those false assumptions, in my opinion, by referring to their product’s rating system as “speed”.

  4. Great article! There is only one disadvantage with using Pingdom though, and that’s the limited number of locations to test from. My website, hosted in Amsterdam, needs to be tested from Stockholm., It is nearest, but it costs some extra ms delay.

    Try webpagetest.org instead.

  7. Thank you Caspar. This post is bookmarked and will be sent in reply to every client who forwards an email from their “SEO agency” complaining about this exact issue.

  9. I know which cache plugin you work for. It is THE best. Thank you for this post, I will share!

  10. If you get really mad get some laughter therapy by plugging the You Tube URL into Pagespeed and seeing the results 😉

  13. I absolutely hate that tool, and people just don’t get it. Worse yet, there are SEO’s out there sending clients to it like it has some value beyond the pretty colors and grades. Thanks for getting the word out!

    1. @Richard Yep, you’re outta luck there, but you can still compare results against results from the same location – even if the closest one available is half around the globe. Like LazyLoad versus no LazyLoad, page caching versus memcached, Varnish versus page caching etc. Load time results may not be as realistic for your local audience, but you can still get a basic idea of relative speed gains.

  15. Thanks Caspar!
    Bookmarked and already shared a couple of times. Lovely you found exactly the right words to describe my feelings about this 😉

  16. Seriously?
    Check this test. It captures screenshots of how site loads in browser.
    webpagetest org/video/compare.php?tests=170212_P6_GPZ-r:1-c:0
    According to that test, your site starts rendering page only at 1.3 sec, which is not even close to 500ms. And that is what Google unhappy about.

    If you would have listened to Google PageSpeed advices and removed those blocking css and js, you could pump up your site to start page render at 500-600 ms and your google PageSpeed score could jump to 99-100.

    If to run test with bandwidth shaped to 3G connection, page render will start only at 3.5 seconds. If you’ve followed Google PageSpeed advice, it could start about 1.5-1.6 sec.

    For such a light site as yours this is not what you need to worry about. It is quite fast, WP Rocket plugin does good job here.
    But it can happen that some other user would read your article, and will treat Google PageSpeed advices as something not worth following.
    So, to my mind, telling that Google PageSpeed is total crap is not quite correct.

  17. That render blocking java script stuff has been giving me a headache, to me java is a type of coffee and I’ve no idea what render blocking is. So thank you! I’ll stop banging my head against the wall and just work on getting my load speed below 2 seconds.

  18. Caspar, this is brilliant! Have you had any feedback from Google on this?

    Last year I spent several months implementing the suggestions from the PSI tool on a Magento EE site, trying to prove the fact that the Page Speed Insights tool is total nonsense in regards to actual performance. It was pretty much proved as after making the changes, we didn’t see any noticeable difference (as expected). At the time, I asked our Google account managers to explain the score in relation to performance, as the stats didn’t match up. Guess what? I heard absolutely nothing back.

    I’ve since tried to contact another 2 account managers (new one’s I’ve come into contact with) and got nothing back. One example I always include is this…

    https://developers.google.com/speed/pagespeed/insights/?url=https%3A%2F%2Fwww.missguided.co.uk%2F&tab=mobile

    This site is ‘Poor’ according to the PSI tool. However, I know for a fact that this website processed +5 orders per second on cyber weekend 2016. How on earth can this site be poor? They’re definitely not worried about site speed; their hardware is absolutely insane and the frontend is rapid.

    Page Speed Insights is becoming more of a pain point as it seems to be getting more momentum and weighting for it’s recommendations. I had a client just last week put a ticket in with the following…

    ‘Our sales have dropped because of page speed’ – with a link to a Page Speed Insights report of their homepage. Obviously this is insane.

    It’s getting more difficult to push back on this and explain to clients not to listen to it, solely because it’s Google.

    The main thing I’ve noticed is that wherever I’ve looked, whoever I’ve contacted, Google don’t seem to be answering any questions or defending themselves. I’m sure you’ll agree they need to start answering these sorts of questions and responding to posts like this one, in order to give us some sort of motivation to use this tool.

    Where are you Google, don’t be so friggin ignorant!!

    1. Thanks for this comment, great context! I have not seen Google themselves clarifying on score and ranking. Other SEOs seem not very keen on keeping things clear either.

      It took this comment from my teammate to have a YOAST author finally explain what they meant by “page speed” (i.e. load time) while we had seen multiple support tickets where people actually had thought that very post had stated “PageSpeed” was a ranking factor.

      The YOAST post links to this article which is pretty unclear as well, so I left a comment there. The author replied once, however, their reply was so generic that I tried leaving a second comment trying to get them to clarify on their own terminology re: “PageSpeed” versus “page speed”. That second comment didn’t even get moderated:

      https://glueckpress.com/wp-content/uploads/2017/05/Screen-Shot-2017-05-23-at-18.57.39-821×1024.png

  19. Thanks for getting back Caspar!

    That’s crazy! The reason SEOs aren’t clear with it either is because they don’t know. I’m not having a go at them for not knowing (how could they ever know if Google haven’t told them), but it’s very frustrating when people say you must get a good score on PageSpeedInsights to have a performant website.

    Google have created this tool, which is actually amazing at what it does. It’s great at suggesting improvements to your site; very similar to pingdom, webpagetest etc.

    However, the big problem is the weighting it’s given. It’s not just used for recommendations like Pingdom / webpagetest. I’ve just read back through the comments and you sum it all up perfectly in your response to Jan Hornung (sorry, I’m just echoing your frustrations).

    This is another perfect example…

    If you Google ‘the importance of page speed insights’, the top hit is a post which says…

    “You always want to score green in PageSpeed – at any cost”

    I’ve ripped it apart and they’ve surprisingly published my comments (respect).

    I think we’ll always be fighting an uphill battle with PageSpeedInsights because Google won’t tell us anything. I’ll keep trying to get something from them and if for some crazy reason they do get back to me, I’ll definitely let you know!

    If there’s anything I can do to help in any way to emphasise how truly shit it is working against PageSpeedInsights, please let me know. Thanks again!

  21. How do you see the ranking factor of the click back rate? Imagine your user is traveling on train and is waiting for your site to load with a bad internet connection. On a page with render blocking css and javascript the user will have a long period of time seeing just a blank page. The user will give up waiting and click the back button. Then he will probably click on the next search result which has followed Google’s advices. I think with more than 50% of mobile users, not optimizing is not acceptable, if you want to be competitive.

    1. I agree, but as mentioned further above, the context of this post is DIY WordPress site owners. I provide support for that type of person on a daily basis, and from my abundant experience they neither have the expertise to validate and implement “tips” from PageSpeed Insights, nor the budget to get their site optimised professionally. Instead they install a WordPress plugin (or a dozen of them) and hope for magic. For that type of person, a speed gain of 100-200% (and by speed I mean load time) achieved by a plugin is great, and far more important: it is the best they will ever achieve, because [see above].

      Your perspective seems typical for the technical/developer kind of person to me, which is perfectly fine, but naturally you are not going to make sense of anything said here unless you put yourself in the shoes of the DIY type of WordPress site owner I had in mind when writing the post.

      1. I think the problems for the DIY users are beginning at the theme. Marketplaces like Themeforest are using the number of features as an acceptance criteria for the listing of a theme. This made the themes overbloated and slow. That’s why I started a new initiative. A marketplace with speed as acceptance criteria http://www.getspeed100.com. This would solve the problem at the roots.

  24. Hi Caspar! Hope you’re well.

    I posted a couple of comments previous to this one and I wanted to let you know I’ve finally managed to get something out of Google…

    I managed to get through on a call and explain the problem I mentioned in my first comment. After going on a polite rant, I trimmed it down to a few questions:

    1 – Does Google Analytics’ speed reports take into account the Page Speed Insights score of that site?
    2 – How can one site with a terrible score load faster than a site with a great score?
    3 – How can the PSI tool score your page speed when it doesn’t actually test the speed of your site?
    4 – Does Google use the page speed insights score when ranking sites organically?
    5 – What would happen to a site organically if they completely ignored the PSI score?

    Q1 answer: “We can’t see any correlation between the Page Speed Insights score and the speed reports in Google Analytics”.
    To be honest, this is quite self explanatory considering a site may not have a GA account, but can still get a score from PSI.

    Q2, 3, 4 & 5 couldn’t be answered. The person I spoke to said that it’s out of their remit to answer these questions so I was passed onto another department and started an email chain where I posed the same questions. This was the response…

    “Although your questions raise a fair point unfortunately it is out of our remit to answer them properly”

    In the same email I was given information on how the site speed reports in GA work (not PSI tool). I explained it’s not information on GA i’m looking for, it’s specifically PSI and asked if I could be passed to another department who could answer my questions. This was their response…

    “Unfortunately the answer is no, I did look into this while gathering more information for you.”

    It seems I’d reached the bottom of an empty well, but they then they came back with some answers. Woohoo!!…

    2 – How can one site with a terrible score load faster than a site with a great score?

    “This could be for a number of reasons from the amount of content on the site to server speeds ect ect (scores aren’t the be all and end all of site speed, but are a great indication of where to improve)”

    3 – How can the PSI tool score your page speed when it doesn’t actually test the speed of your site?

    “PageSpeed Insights checks to see if a page has applied common site performance best practices and provides a score, which ranges from 0 to 100 points”

    4 – Does Google use the page speed insights score when ranking sites organically?

    “No not the score itself but site speed is a factor in the ranking of sites by google’s algorithms”

    5 – What would happen to a site organically if they completely ignored the PSI score?

    “The score itself wouldn’t have an effect but site speed is taken into account with google but has a reasonably low effect on ranking in comparison to other things. As PSI is a strong indicator of site speed it would be unwise to not use it to optimise”

    ———

    So the next time you hear something resembling…

    “Holy shit our sales have dropped and it’s because I’ve just tested my site on Page Speed Insights and we have a score of 20/100! It’s also going to affect us organically and we really need to improve the score right now. Drop everything and spend copious amounts of time and money getting a ‘Good’ score”

    …you can politely tell them their wrong.

    Any data reported in GA, isn’t connected in anyway to Page Speed Insights. It’s a stand alone tool that people can use to see where to make improvements; it’s not the be all and end all of speed (as Google put it).
    Due to the fact that it’s a stand alone tool not linked in anyway whatsoever to GA, I think it’s safe to say it doesn’t affect any part of your sites’ actual performance.

    The most important point for me is…
    What would happen to a site organically if they completely ignored the PSI score?
    Google – “The score itself wouldn’t have an effect…”

    They then go on to say “site speed is a factor in the ranking of sites by google’s algorithms”
    Yeah of course it is and it should be, but PSI has nothing to do with it. People should definitely take site speed seriously because as everyone knows users hate waiting, but they should only look at PSI as a tool for suggestions. It’s a good tool and there’s a lot of others like it, but that’s it! Nothing more nothing less.

    I really hope this helps anyone reading this and thanks again Caspar for this amazing post!

    2. Hi Micky, thanks – this is great! It’s awesome that you actually reached out to Google and insisted on some answers. These statements from Google certainly come handy if people are dealing with customers who are getting stubborn on their chase for a perfect PSI score. 🙂

  25. Pingback: How To Speed Up Your WordPress Website
  26. Pingback: Hotel Website Speed: Why It Matters & How to Improve It | Otium Boutique

Reactions are closed.

Mentions