Using Page Speed Insights is worth it - some hints
7 views
Skip to first unread message
Thomas Hey'l
Aug 26, 2013, 4:13:31 PM8/26/13
to page-spee...@googlegroups.com
Hello, folks!
For those who don't know me: I'm a Top Contributor at the German Webmaster Central (Google Help forum).
Here's a personal story, might be TL;DR (warning, a longer one ;-) !).
Back in June, 2012, it was already known Google would soon drop the speed information from the Google Webmaster Tools. This was a wise decision as the tool evaluated only requests from users having the Google Toolbar installed and using google.com instead of local properties, so there were only few and unreliable data points available, specially for me targeting users who speak German. It was then I decided to create my own tool and to apply it to about 15% of the requests at my private site. The reason was rather to gain some insights and rather a hobby than a professional interest.
As you all might know, Page Speed Insights does not say anything about the real load times, it simply evaluates the technical and measurable conditions. You might also want to check whether your changes improve your score, but this does not give you too much information about the real impact, only about preconditions.
Once my tool had collected enough data points (I can adjust how many), I also started to evaluate them so I can really see the impact of changes. One very important thing: You should find some means to exclude third party tools from the PSI tests, i.e. third party scripts or AdSense. Those are definitively not under you control, so there's little or nothing you can do to improve them (unless the providers provide a real ansynchronous solution).
In my case, as I write many things out using PHP (including scripts and ads), I use a noads=1 or noads=true GET-Parameter which exludes all superfluous requests.
Back to the basics. I'm not a savvy programmer, but able enough to check from my log files which popular pages cause issues (or whether the average and median times are not too good). It would be far beyond such a post to explain the details. That said, I've been able to get me the desired data as those who use Analytics might be able to (I don't use it).
Long cut short, I compared the real world numbers to the PSI suggestions, tried to implement what seemed to be reasonable and all I can say is "wow"!
Let's put this into numbers, ok? October, 2012 was a very poor month regarding speed. The average of overall measured loading times was 1,309 ms with daily peaks of 1,600+ ms. The median was only around 1 s. You think that's fast? No, it is not. 50% of the users has to wait more than 1s until the pages had fully loaded.
Following Pagespeed suggestions and using my own brain I brought the average overall loading time down to less than 900ms by now, the median is at 718ms. But what's really impressive is the shares. The share of pages fully loaded within 1.5s was 70%, those with <= 1.0s 46% and that with <= 0.5s 15% back in October.
As for today in August, 2013, they are 86%, 68% and 28% (!). Now this is what I call a real gain.
22 percent more of the pages requested by real users load in less than 1s and more than two thirds overall in 1.5s. With "load", I mean the fully loaded page, not only the content above the fold.
Next, I looked at the popular entry pages (i.e., first load coming from Google search results) and tried to track down their issues (i.e. by spriting images or removing other requests). This measure has a high impact as you might not know what the user's entry page will be - probably not your highly optimized homepage.
To conclude, as Dave Mankoff correctly stated,
It should rather be to understand the suggestions, to evaluate whether the required effort is worth it and to achieve as much as possible with the least amount of work after you checked the basics. If you i.e. think that you need several dozens of images on you home page, there's probably rather something wrong with your concept. Don't judge one thing without looking at the other. If your users are fine with the way it is, that should be fine for you, too.
Regards, Thomas
For those who don't know me: I'm a Top Contributor at the German Webmaster Central (Google Help forum).
Here's a personal story, might be TL;DR (warning, a longer one ;-) !).
Back in June, 2012, it was already known Google would soon drop the speed information from the Google Webmaster Tools. This was a wise decision as the tool evaluated only requests from users having the Google Toolbar installed and using google.com instead of local properties, so there were only few and unreliable data points available, specially for me targeting users who speak German. It was then I decided to create my own tool and to apply it to about 15% of the requests at my private site. The reason was rather to gain some insights and rather a hobby than a professional interest.
As you all might know, Page Speed Insights does not say anything about the real load times, it simply evaluates the technical and measurable conditions. You might also want to check whether your changes improve your score, but this does not give you too much information about the real impact, only about preconditions.
Once my tool had collected enough data points (I can adjust how many), I also started to evaluate them so I can really see the impact of changes. One very important thing: You should find some means to exclude third party tools from the PSI tests, i.e. third party scripts or AdSense. Those are definitively not under you control, so there's little or nothing you can do to improve them (unless the providers provide a real ansynchronous solution).
In my case, as I write many things out using PHP (including scripts and ads), I use a noads=1 or noads=true GET-Parameter which exludes all superfluous requests.
Back to the basics. I'm not a savvy programmer, but able enough to check from my log files which popular pages cause issues (or whether the average and median times are not too good). It would be far beyond such a post to explain the details. That said, I've been able to get me the desired data as those who use Analytics might be able to (I don't use it).
Long cut short, I compared the real world numbers to the PSI suggestions, tried to implement what seemed to be reasonable and all I can say is "wow"!
Let's put this into numbers, ok? October, 2012 was a very poor month regarding speed. The average of overall measured loading times was 1,309 ms with daily peaks of 1,600+ ms. The median was only around 1 s. You think that's fast? No, it is not. 50% of the users has to wait more than 1s until the pages had fully loaded.
Following Pagespeed suggestions and using my own brain I brought the average overall loading time down to less than 900ms by now, the median is at 718ms. But what's really impressive is the shares. The share of pages fully loaded within 1.5s was 70%, those with <= 1.0s 46% and that with <= 0.5s 15% back in October.
As for today in August, 2013, they are 86%, 68% and 28% (!). Now this is what I call a real gain.
22 percent more of the pages requested by real users load in less than 1s and more than two thirds overall in 1.5s. With "load", I mean the fully loaded page, not only the content above the fold.
Next, I looked at the popular entry pages (i.e., first load coming from Google search results) and tried to track down their issues (i.e. by spriting images or removing other requests). This measure has a high impact as you might not know what the user's entry page will be - probably not your highly optimized homepage.
To conclude, as Dave Mankoff correctly stated,
The goal for PageSpeed Insights should not be to get 100.
It should rather be to understand the suggestions, to evaluate whether the required effort is worth it and to achieve as much as possible with the least amount of work after you checked the basics. If you i.e. think that you need several dozens of images on you home page, there's probably rather something wrong with your concept. Don't judge one thing without looking at the other. If your users are fine with the way it is, that should be fine for you, too.
Regards, Thomas
Reply all
Reply to author
Forward
0 new messages