Steph I'm reading so much conflicting stuff online about Google PageSpeed Insights. I'm curious to hear how much you guys use / rely on them for judging website speed. Thanks so much x 6 👍🏽67 💬🗨📰👈
Daniel Let me clear some things up for you then. I'm not a God or anything but believe I have a good grasp on this topic. First point PSI is not the same thing as Google's CWV, although they did just release a new version and from what I read it might finally address the issues that make it very different. I'm not sure yet, but I'm talking about CWV as we know it. First PSI – PSI servers are owned by a brand and they bounce your website off of their server to measure its speed. Now your first test is not always accurate. Normal protocol for PSI testing is three back-to-back tests. Why? Because your first test is going to load everything as though it's the first time it's being seen. PSI servers have a very short TTL so they're caching of your website disappears in minutes that's why you need to go back to back. First test loads all of the tech and is there for longer second test shows you the result of all that tech and will therefore be a shorter time. All of this is especially true if you use a CDN. So PSI is the best way to determine user experience because on that third test you are seeing the results of how your users are going to actually experience your site and that's what counts most. Google's CWV for the most part is only going to measure the origin server. So a CDN will pretty much make no difference in your score. So it's perfectly possible to fail Google cwv while having an instaclick fast website. Another way to think about CWV is it's just measuring your site without all of the tech. This is why unless you're a WordPress developer able to hard code websites you need to get on amazing servers. Or just build really boring sites that don't have a whole lot of imagery or videos or animation. There's so much more to it and many holes can be poked and things I have said but that's because I can't sit here and write an entire book on this. Lol
Daniel And I understand you said Google's page speed insights but that's Google. The relativity is between Google's way of measuring sites and normal PSI servers. Daniel And I understand you said Google's page speed insights but that's Google. The relativity is between Google's way of measuring sites and normal PSI servers. Also when it comes to PSI make sure you understand throttling. Also know that certain things like carousels can wreak havoc on some PSI tests. This is not necessarily a Bad thing but this is why I said I'm not going to be writing a book on all the factors. I thought I'd just give you a bunch of information and whoever else wants to join in or whatever other questions you have then I would answer. Or learn something more. Sorry if there's a duplicate reply it seemed to fail and I can't delete it either. But this should be the 4th entry. Thang » Daniel I am not an expert myself, and sorry for the rant. 1. first test is not always accurate. -> Test will never be accurate because other factors can change affects your scoring such as your Server Response time, 3rd parties domains, async function, location, Network connection, etc. 2. Normal protocol for PSI testing is three back-to-back tests where did you get this information. They not loading everything like GTmetrix does with fully loaded stuff. PSI focuses only on the above-the-fold content that is why you're seeing the CLS issue on the Google search console even though PSI tells you to have a 0 CLS score. 3. PSI servers have a very short TTL so their caching of your website disappears in minutes that's why you need to go back to back. – This is true and Google is trying to simulate first-time visitors without any cache. 4. So PSI is the best way to determine user experience because on that third test you are seeing the results of how your users are going to actually experience your site and that's what counts most. -Not really PSI will tell you to have a good FCP score but webpagetest will tell you otherwise. No matter how many tests you run on PSI it will always be different & you can't determine which is best & you can't use this for measuring CWV tools. 5. Google's CWV for the most part is only going to measure the origin server- Nope, it collects the data anonymously from the real users if they decided to sync their account to Google, not from your server. CDN helps a lot and what it does is cache your static assets to their edge network and serve it to your visitors rather than hitting your servers. If you're relying on lab data score, you will never pass CWV at all because in the field you have to consider 1. Server Response time 2. Audience network connection 3. Your Code too. 4. 3rd parties domains. 5. Cache condition 6. Cache hit ratio (if you use CDN) 7. What about network congestion 8. Identifying which resources to prioritize etc. 6. I am not an expert either but I pass my CWV because I monitor my real user performance and act accordingly. 7. You can use Oxygen builder if you want clean code and less bloated but a pain to set it up or use any page or theme builder if you want to build a dynamic website that is not boring and follow best practices it. 8. If you don't trust PSI you can use webpagetest it is a bit advanced and it will give you more information. 9. Throttling is applied on Mobile tests because we can't expect everyone will be using 4G connections (it does but not always) or using high-end devices. 10. Carousels can affect your LCP performance if you enable auto-looping or use margin to move the slider around. You can fix it by disabling the auto-loop feature and using the transform CSS property. I think I give you insights Cheers Daniel » Thang Thanks for writing the book I wasn't willing to write. And I agree with a lot of things you said but there's a lot of things that have even more caveats you have to point out and more circumstances and more factors so we could do this all day. Like for instance you seem to be talking about shared hosting. Which is yet another caveat because if you're on a VPS you don't have to worry about what the other websites are consuming resource-wise. The only reason why I was pointing throttling out is because for instance pingdom tests with a full throttle. GT metrics defaults to no throttle. You can set that up and change it on both if you want. I believe GT metrics only goes up to 25 MBS. I'm not sure how far Pingdom goes with the throttle though. Oh and by the way there's more than one way to fix that carousel. But yes without question that is one way to address that situation. But you really got to read my language carefully because even in that case I said carousels "can be" troublesome. Meaning I realize there are numerous ways to make it not be. And wrong the test is accurate and all you have to do to eliminate everything you say that makes it not accurate is be on a VPS. So you see this is what I mean there's just so much information to cover you and I both can't cover every single circumstance and situation here. Thang » Daniel 1. Like for instance, you seem to be talking about shared hosting. This is yet another caveat because if you're on a VPS you don't have to worry about what the other websites are consuming resource-wise. – No I am talking for all the instances. I am talking about website performance, not backend performance. 2. Stop relying on Pingdom guidelines like a) "Make fewer HTTP requests" – if your site runs on an HTTP/1.1 server then it makes sense because multiple connections cannot be made but if your site runs on HTTP/2 or 3, Pingdom should not recommend "Make fewer HTTP requests" and should remove it. b) "Compress components with gzip" – even though if you enabled "Brotli compression" on your server Pingdom will still flag you. Brotli > gzip PSI is using custom throttling (10MB) and tries to measure from real user experience metrics. We will never know what network connection our visitors will use. GTMetrix uses outdated Lighthouse version and chrome and for some reason, you can't test Mobile when Mobile-First indexing is the most priority. C'mon 3. I don't know much about carousels because I never use them and no one should use them either. 4. I didn't get this one "wrong the test is accurate and all you have to do to eliminate everything you say that makes it not accurate is be on a VPS. So you see this is what I mean there's just so much information to cover you and I both can't cover every single circumstance and situation here" To which point are you referring. Seems like you are talking about Shared hosting and VPS the most so let me tell you (from my experience). Site speed doesn't matter if you use shared hosting. if you run a portfolio website with proper caching set up and use CDN because most of the time your static assets will be served from their CDN network and most likely it will never hit your server. If you're talking about WooCommerce and other dynamic stuff then VPS will easily out-perform and have better UX in the backend.
