New Google Insights - Huge Drop Using LightHouse - How can there be so many inconsistencies?

[Sara]

Hello, Please can you help. 

We create many WordPress websites and when we launch one we would make sure of the following:

1. Loading under 2 seconds on mobile and desktop on pingdom, keycdn, and GTmetrix

2. Every site scored high 90's in Google Insights

3. Reddis on the server and litespeed, no plugin over load

The sites haven't changed, they still load very often in under a second on mobile testing virtually and in real life, yet our results on Google insights have suddenly dropped from high 90's to 16...for mobile..

We have tested over 45 premium themes for WordPress and they all score terribly too.

How can we get these results yet score so low in your one test:

Pingdom: Load Time 600 ms

GTmetrix: Page Speed Score Grade A - YSlow Grade B - Fully loaded Time 2.9 seconds

Mobile test held in hand on 4g: less than a second,

yet your test shows a score of 16...for mobile - Node errors and Doms etc - Surely you need to look if its a word press site even if they have all that going on but still load under 2 seconds your results are slightly off..?

You cannot have all these other sites showing great results, your old system great results and now suddenly we score terribly... AND you email every analytic customer saying there website is now terrible..

that can put a company out of business?! We work tirelessly and hard to make sure we create fast loading websites and your results say its not fast.. when it actually is?

Is there anyone we can chat to? What can we do about this? Are others having the same issues?

Thank you

[Uziar]

Dear, 

You are right. even google didn't reach 90+ on page speed insight for mobile and we all know how lightweight google is.

[Seb Dinatale]

I don't know how to help I use notepad++ to optimize my website but I am unable to download the content file from pagespeed insights

--
You received this message because you are subscribed to the Google Groups "pagespeed-insights-discuss" group.
To unsubscribe from this group and stop receiving emails from it, send an email to pagespeed-insights-...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/pagespeed-insights-discuss/f1cce4dd-33ec-4c82-9e66-89761a505fe4%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Frank]

Hey Sara,

I'm also not happy with the Lighthouse update.

It's possible to get to 100/100 tough, but it's a

lot harder - and Lighthouse is buggy too.

Would you share, Sara, why it is so important

to you to have an amount of speed points in the

upper range? Is promoting speed part of your

marketing strategy? Or are customers blaming

you for bad score? What is your concern?

The best,

Frank

[Maddy]

I would love to know how it's possible to get at least a green score on mobile, my sites all scored 80 85 90 on previous pagespeed and since this update, mobile dropped to 14,20,25 and even getting desktop to go up to 75 or 80, Mobile score is not moving, not to mention every time you test there is a different score

GTMetrix and Pingdom is giving really good markers and yet this pagespeed update is messing everything up

[Frank]

Messing everything up could be by intent giving

some motivation to dive deeper into AMP :-)

Seriously, I have no idea why the Insights thing

that wasn't really popular before and was even often

blamed as too professional or worse ("irrelevant"),

now got even doubled in complexity by neerdy

and therefore mostly confusing Lighthouse.

That's a classic shoot in the knee - or by intent.

Maybe this speech helps u a bit with better scores:

https://www.youtube.com/watch?v=Mv-l3-tJgGk

The best,

Frank

[Shane Garland]

The new update is a nightmare. The results are wildly inconsistent I have a huge problem with the image compression in particular. I'm having to compress my images 4-5 times over in some cases. I get them down to the required kilobytes and then a re-scan is asking me to compress them again. The image re-size is the same. I resize them and the scan is clear, 5 mins later its asking me to resize the images for further savings or its now adding in images that hadn't shown up earlier.

Even the page speed result itself is erratic. One scan I get 70, 1 min later its 82. Then its like 60 and it goes on and on. It's like there is someone at the other end flipping a coin to decide what your result is.

Before the update my main site scored between 88-100 it dropped to 47 after the update. I have it back up to around 85 with a lot of hair pulling. 

But anyway a few tips on getting your pagespeed up. Images will massively improve your score. Ridiculously they advise you to use webp which is not fully supported but once you get potential savings down under 8kb that will do you. So if its says image file is 40kb and potential savings is 20kb  you will need to compress the image by 13 kb and get it to 27kb. But be prepared for some hair pulling because lighthouse seems to constantly move the goalposts and change its mind on what size the image should be. 

Also serving resources with an efficient cache policy might need to be looked at. Before the update my htaccess caching script was fine but after the update I had to totally redo my htaccess file. 

But yeah you defo are not alone I think this new update is a real headache for everyone. I'm not against changing things up and I know technology is always moving forward but this update seems so glitchy and buggy its adding hours on to my workload for every site. What used to take me an hour now takes an age 

caching

[Gints Grīnbergs]

Hi,

I think that the new PageSpeed is going the right direction (probably I'm 1 out of 100 who is thinking like that), but at this stage, it's a bit too strict. Previously the only thing what was included in the rating related with actual page speed was server response time and that's it. Sometimes it was pretty easy to play with the numbers and get 90+ but at the same time the actual page speed in seconds got 2-3x slower then it was with 40-60 old scores :) If you read all the new recommendations, they are way more useful and really can help to optimise the site speed even you pretty much won't get 90+ in reality in the tool, but in the real world that could help a lot.

In my understanding, the new PageSpeed contains a bit too hight weight on some new technologies what came up on mainstream more recently even was introduced ages ago like WEBP image format is in the market like 10 years or so, but only recently the most site speed optimisation start to talk about it and more browsers start to support that format. So the tool is weighing these new generate image formats very highly and what makes sense, you really can reduce the image size. The same with lazy load, in the new tool it's way important now. As well JS rendering and parsing times are hugely important. Just using jQuary, you can reduce your score drastically by new calculation quite easy.

Another thing, I don't know why you guys are so happy with GT Metrix scores? Just because they are higher? I think some of the things in GT Metrix are way too lightly checked and taking too easy. For example, the GT Metrix showed me 5-10% opportunity on image optimisation in one of the websites. I did manual image optimisation with GIMP with these images and I reduced the size from 850kb to 160kb for only one image. It's not even close to 5-10% reduction, it's like 400-500% reduction with the same quality image for human eyes :) And that even wasn't converted to WEBP but standard JPG. With WebP could go under 100kb easy, so real-world reduction potential is 600-700%. In this case, the new Page Speed was way closer to real image optimisation potential.

The main point on this. Why you are chasing the scores where the actual real-world speed is more valuable? I will repeat, that the old system and scoring system was too easy to manipulate when even the actual real-world site speed was pretty bad. So keep calm and don't' chase the score, but actual speed.

[Frank]

Cannot see any antagonism between "real-world site speed" and score.
And this applies to the former version of Insights and should be
also applicable to the Insights tool after the Lighthouse update.
"Score" means some kind of semi-professional metric beyond "load time".

This is because "load time" (which one, btw? a page has at least 2 of them)
can never be a constant and it depends heavily on user devices, net speed,
geo location etc. etc. What you call "real-world" site speed is simply an
accidental number out of a range of non-constant consumer perspectives.
Therefore RUM data!

Big problem with the Lighthouse update is for me, that weighting
of scores isn't transparent and doesn't even seem to be stable
for the same set of test conditions.
This motivates not really and this way negates the purpose of the
Insights at all.

And so it's much more like trying around instead of doing work!

People will give up earlier compared to the Insights tool before
the Lighthouse update, I guess.

Cheers.

[AnselTk1]

I think that:

A) the feedback of Pagespeed Insights is important from a developer standpoint but..

B) The results are wildly inconsistent and..

C) SEO consultants and Product Managers are using a (very) high score as a goal and assessment of a site which is frustrating to say the least.

In my opinion more needs to be done by Google to outline exactly to SEO professionals and managers what these scores mean and they are a metric not an overall value of the site itself.

GTMetrix etc - the above folks don’t care. They care what Pagespeed Insights says as Google is their primary target.

At least this what I’ve experienced.

-- 
You received this message because you are subscribed to the Google Groups "pagespeed-insights-discuss" group.
To unsubscribe from this group and stop receiving emails from it, send an email to pagespeed-insights-...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/pagespeed-insights-discuss/7c082154-0366-4a70-98a9-757434754a57%40googlegroups.com.

[Gints Grīnbergs]

As I understand, then the scores are based on the times only - that is based on all these 6 Lab Data time scores. As you can see here: https://docs.google.com/spreadsheets/d/1Cxzhy5ecqJCucdf1M0iOzM8mIxNc7mmx107o5nj38Eo/edit#gid=0 

And here you can see how the weighted average is calculated: https://www.wikihow.com/Calculate-Weighted-Average#Weighted_Averages_without_Percentages_sub 

As I understand previously the PageSpeed score was more based on all these issues and wasn't based on times. So the new PageSpeed is based on times and everything that you see under Opportunities and Diagnostics are just an idea what you can try to optimise to get better Lab Data. So, based on this information I would believe that the new scoring system is going in the right direction as previously the score was build based on all these diagnostic points no loading speeds and I have seen 90+ old score sites which load 5-10secs just because the caching is working pretty bad and all these score points was gained during website rendering each time and added huge weight on server load but in the same time score on PageSpeed was pretty high.

Why the scores are with so big swings? Probably, because there are very close to the weighting system on all these milliseconds and depending on server load and activities on the website, these some milliseconds can play a big rule on the final score.

So, the primary optimisation goal should be in priorities like this:

1. Time to Interactive
   
2. Speed Index
   
3. First Contentful Paint
   
4. First CPU Idle
   
5. First Meaningful Paint
   

And don't case about Estimated Input Latency

Another thing what is not clear is about what network limits they are using, but I'm guessing that for Mobile is 4x slowed down CPU and applied some network limits what you can see here: https://github.com/GoogleChrome/lighthouse/blob/master/docs/throttling.md and that could explain why desktop vs mobile is so big difference.

• Latency: 150ms
  
• Throughput: 1.6Mbps down / 750 Kbps up.
  
• Packet loss: none.
  

As there are these CPU and network limits, it's logical that these milliseconds could swing way more as on good speed and good performing laptops.

I'm not pretending that I'm 100% right here, but that's how I understand this based on currently available information.

Bonus. Here you can play with the scoring calculator: https://docs.google.com/spreadsheets/d/1dXH-bXX3gxqqpD1f7rp6ImSOhobsT1gn_GQ2fGZp8UU/edit?ts=59fb61d2#gid=283330180

[Gints Grīnbergs]

Sorry, the calculator is wrong. You need to check the score calculator on these google doc: https://docs.google.com/spreadsheets/d/1Cxzhy5ecqJCucdf1M0iOzM8mIxNc7mmx107o5nj38Eo/edit#gid=283330180

I just manually added the number from the Lab Data and got ~98% the same score in this calculator what I see in the Page Speed.

Cheers,

[JJ]

+1 for this. It is really frustrating - the new speed test is next to useless to help developers.