page speed

[chaumi]

This might be better as a poll, and maybe has already been covered as one and I missed it, but even if it was I guess we might get some other viewpoints.....

Managed to get desktop to 98 in Google Page Speed

But mobile (2 weeks after new hosting, CriticalCSS, autooptimize, CDN, head scratching and other miserable attempts - technically I have no real idea what I'm doing and trying to learn on the fly) best managed is around 50

Which is clearly rubbish (or is it??)

So I'm curious.

Is anyone else bothered yet about mobile speed?

Have you already cracked it and it's no longer a problem? (in which case are you prepared to share how you did it?)

Did you never have a problem anyway?

Or do you know it's an issue, and have a plan to tackle it?

And any views on whether 50 is reasonable or completely disastrous?

[Michael Martinez]

"Mobile speed" and "desktop speed" are misrepresentative.

The speed of a Website is affected by many things including server load, the device trying to load the site, the Internet connection, and more.

Google's page speed estimating tools have been harmful, in my opinion, because they create the illusion that if you fix the problems for the tools you've fixed the problems for everyone.

Some of the tool recommendations are ridiculous, such as advising you to fix resources hosted on Google's own services and compressing already compressed images that are less than 25Kb in size.

When Core Web Vitals is integrated into the ranking system in a few months, it will only be 1 of many signals - and most likely all sites will be weighed against each other. Until I see otherwise (and 2 screen captures from someone who claims they've found "proof" isn't sufficient), I'm assuming these page speed estimates won't be much of a problem for most people.

Some sites really do need to speed things up, both for search and for users.

[chaumi]

Thx Michael

Do you consider it a possibility this is really all just semi-scaremongering tactics by G? In other words 'trickery' to get everyone taking (more) notice of their mobile speed (which clearly will be a good thing for many sites and hence good for Google/searchers) and/or trying to drive AMP?

Clearly they have intent, and we all know the intent (or at least the base concept of it and the likely result) is a valuable one.

I know you'd be guessing (or at least guess you'd be guessing)...but a highly educated guess is normally as good as it gets where Google is concerned (other than very specific/clever testing in some circumstances of course)

[Michael Martinez]

I don't really have an opinion either way.

Google occasionally tries to frighten everyone into doing what they want. The push to get all sites onto HTTPS is a clear example.

I have no idea of what possible benefit their engineers think this push on Core Web Vitals will provide. Websites only control 1 side of the equation so it doesn't guarantee a better user experience.

Their page speed tools strike me as reflecting an Ivory Tower syndrome. But they might be thinking about users in Africa, India, Pakistan, and other emerging Internet economies. Maybe in order to monetize all that traffic they need to put ads on superfast sites that compensate (as much as possible) for really poor user connections. Most people in those countries access the Internet over cellular networks.

That's the best guess I can offer.

[chaumi]

''they might be thinking about users in Africa, India, Pakistan, and other emerging Internet economies. Maybe in order to monetize all that traffic they need to put ads on superfast sites that compensate (as much as possible) for really poor user connections. Most people in those countries access the Internet over cellular networks''

Suspect you may be spot on.

[Olessia]

We have developed our site without templates on CMS WordPress. Our reports were also red and terrible. BUT. If you go to gtmetrix and google page speed and look at the Askgeblers website - all optimization statistics are RED. Draw your own conclusions

[SmartThomas]

What kind of tools do you recommend to measure the site / page speed ?

[xecutable]

Most tools use now Lighthouse 7.1 which can be check via your Chrome developer tools.

Depending on how the server handles the request, I get mid 80s to high 90s on mobile and high 90s to 100 on desktop.

I'm only interested in mobile these days and more and more traffic is coming from mobile devices, for me at least.

P.S If I lazyload everything even above the fold, I can essentially boost the scores, but why mess up the UE for a bunch of %.

[chaumi]

Personally I'd say the only one you're interested in (from a results perspective) is Google Page Speed.

https://developers.google.com/speed/pagespeed/insights/

The others (GTMetrix etc) are ok for analysis of what's wrong...but in my view (and I could be wrong) there's only one measurement that counts to Google, and that's theirs. So irrelevant of what any others say, if you're no good there then you're no good anywhere (even if they tell you you are).

[xecutable]

Personally I'd say the only one you're interested in (from a results perspective) is Google Page Speed.

https://developers.google.com/speed/pagespeed/insights/

The others (GTMetrix etc) are ok for analysis of what's wrong...but in my view (and I could be wrong) there's only one measurement that counts to Google, and that's theirs. So irrelevant of what any others say, if you're no good there then you're no good anywhere (even if they tell you you are).

That too uses lighthouse, the same as your Chrome Dev tools if you right click on your site, click on Lighthouse and benchmark. However gtMetrix are still using version 6.3 which is why probably the mismatches occur.

[Michael Martinez]

All of these page speed tools are simulating mobile devices (and wireless connections). Google's algorithms use aggregated data from the Chrome User Experience (CrUX) report.

But it's important to remember that the capability of the user device is only one factor in the equation, which looks something like this:


User-specific Page Speed = Page complexity + server responsiveness + user device + user connection + Internet congestion

Server responsiveness = Hardware + server load + data center hardware + data center load


You really only have control over your page design and your hosting. But 1 rogue bot can bring your server down.

On shared hosting plans, if you use AddOn domains and eat up all your resources, your account will be throttled.


While it's important to improve user experience, it will always be a compromise.

[xecutable]

Actually the Lighthouse calculator allows us to set the weight of each stat from which the score is derived. At the moment they are:

First Contentful Paint - 15%

Speed Index - 15%

Largest Contentful Paint - 25%

Time to Interactive - 15%

Total Blocking Time - 25%

Cumulative Layout Shift - 5%

In the previous version 5, the value were totally different. Some where not even there. So I'm guessing these will keep on changing and changing.

[Michael Martinez]

According to their GitHub page, "Lighthouse applies network throttling to emulate the ~85th percentile mobile connection speed even when run on much faster fiber connections."

lighthouse/throttling.md at master · GoogleChrome/lighthouse · GitHub

I don't know how well the Chromium team syncs their defaults to any Web crawl statistics. So it could be that their default values are based on a recent analysis of what they're seeing in their crawls. But I suspect that is kind of a random situation.

That aside, it's my understanding (and I could be wrong) that Lighthouse allows you to change the weights so you can model different configurations to see where you need to do the most work.

The GitHub page says the DevTools throttling may differ from their version. It's also going to depend on what version of the DevTools someone is using. Some people don't update their browsers very often.

They also announced last year that they had switched their simulation model from a Nexus 5X device to a Moto G4 device. I don't know how much that transition affected people's perceived performances.

What's New in Lighthouse 6.0 (web.dev)

[xecutable]

That aside, it's my understanding (and I could be wrong) that Lighthouse allows you to change the weights so you can model different configurations to see where you need to do the most work.

They do. If you do the test you get a message below the score "Values are estimated and may vary. The performance score is calculated directly from these metrics.See calculator."

And when you click on Calculator you can adjust these values.

[gm2891]

SmartThomas, don't use any third-party speed testers. What the point if it can be tested directly with Google?

Simply, use Web.dev for checking your speed and site issues and try to fix those things out.

Everything in detail on how to do it is explained right there as well.

If you don't know how to fix it, install Nitropac. There are no better plugins for speed than this.

I tried WPRocket, but had to ask for a refund, still haven't got it yet as their office burned down or something...

[xecutable]

Ive tried WPRocket as well and I wasn't happy either. Litespeed cache was better, good options, nice image conversions and resizing, but was still not enough. I'd say Cloudflare proves to be the fastest service out there for me at least, but it comes with a price, like everything in life I guess.

[PROFRBcom]

Working on a new project.

It is, as my good friend likes to call it, part of the #confetticlub now