SIte speed in the webmaster tools

[Jeztah]

A point rather than a question.

I have a really strong feeling that Google is using analytics data for
people with page speed installed Along with thier own bots speed when
determining site speed for a site and putting it back into webmaster
tools.

I came to the conclusion when checking my site speed on a development
site that only about 10 people have the URL for (and google
obviously). I noticed that since December (around the time that I went
on holiday and have really bad slow internet in my holiday location)
that my average site speed has gone from 0.5 seconds (faster than 96%
of sites) to 3 seconds (slower than 51% of sites)...

The only thing that has changed on the site is me developing a page
that is not even linked in any navigation but does have analytics
running on it and appears in analytics as accessed....

Just to confirm.

Nothing on the site / server changed and I can load it remotely on a
faster connection in it's old fast time.
The traffic of the site hasn't changed
The database hasn't changed enough to reflect 2.5 seconds of lag


Can anyone confirm that this is the case with site speed in webmaster
tools - I understand it's very hard to verify on higher traffic sites
and I know if I am correct that once I get back to my super fast
leased line that I can bring the average down again.

Regards

[Jonathan Klein]

The data in webmaster tools comes from users with the Google Toolbar that have the PageRank feature activated:

http://googlewebmastercentral.blogspot.com/2009/12/your-sites-performance-in-webmaster.html

So if your site has low traffic numbers and you have been browsing it from somewhere with a slow internet connection (and you have the toolbar installed) you could be skewing the numbers by yourself.  Does that make sense?

-Jonathan

--
You received this message because you are subscribed to the Google Groups "page-speed-discuss" group.
To post to this group, send email to page-spee...@googlegroups.com.
To unsubscribe from this group, send email to page-speed-disc...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/page-speed-discuss?hl=en.

[Alex McAuley]

Thanks Jonothan - that's exactly what I thought was happening - I just wanted to make sure before I went through every page on my site trying to work out what was choking it!!.

 

Kind regards

 

 

Alex Mcauley

[Paul]

Hi Jonathan

This is interesting and opens up a new topic for debate.

You are relying on people with a decent internet connection and using the google tool bar in order to get a decent performance record in webmaster tools which some people use as a benchmark.

So a handful of dial up users from a shared internet connection at a cyber cafe in the middle of Africa that uses healf dead pigeons for their carrier tones are going to drive my sitespeed down.

For someone like myself that works for a boss I have certain standards that I need to meet and maintain and our site speed is one of them.

Basically what your answer says is that webmaster tools is not a correct indication of the page speed of your website.

Then what is?

[Alex McAuley]

Paul.

 

A couple of quick points!.

 

1. I don't have the google toolbar so I am assuming it's done on analytics data + page speed plugin.

2. It is the most accurate data because it's the average speed that is accross your site from your userbase - adminitly people can manipulate this in varying ways to up or downgrade a website in the SERP when it comes into play properly. From the words of a google employee - site speed is only one signal or metric so it should not be the be all and end all for making sure sites appear high in search engines.

 

I think the point of page speed is for the developer to do as much as they can to ensure best coding practices so that where applicable the site loads in an optimal fashion for the client accessing it ... For example someone on dialup wold see a slower load of a 50/100 page score than they would if the same site had 100/100!.

 

I think once the metric carries more weight these things will change.

 

Regards

 

 

 

Alex Mcauley

----- Original Message -----

From: Paul

To: page-spee...@googlegroups.com

Sent: Friday, January 14, 2011 10:36 AM

Subject: Re: [page-speed-discuss] SIte speed in the webmaster tools

--

[Jonathan Klein]

Webmaster tools is still a decent indicator of your site speed.  As Matt Cutts said it shows "information very close to the information that we’re actually using in our ranking" (http://www.mattcutts.com/blog/site-speed/). 

Your example of a group of users who experience extremely slow load times is relevant, but not as important as you deem, I think.  I doubt that Google is blindly averaging all of the load times together to give your site a speed ranking.  Outliers (like a 10 minute load time when every other request is sub 2 seconds) are probably not going to yank up your overall load time in Webmaster Tools.  I don't have a whole lot of evidence to support this, but in all performance analytics the average can be deceiving exactly because of issues like this - big outliers. 

With that said, if the majority of your users are 6000 miles away from where you host your site and you aren't using a CDN you are going to see poor performance in Webmaster Tools, and your search engine ranking will suffer.  This makes sense, your core user base IS experiencing slow load times, even if the site is fast for you, who is siting a couple miles from your datacenter/colo/basement server room/etc. 

So I think the answer is to trust that real user data is going to provide a decent perspective on how fast your site is.  Webmaster Tools also indicates how accurate the data is based on how many data points Google has.  Presumably if the data is of low accuracy then Google won't weight it as heavily in your ranking. 

If you are responsible for hitting certain performance metrics then I suggest you use a more specific measurement.  As Sean Power suggested in his talk at Velocity last year, you need really well fleshed out SLAs if you are responsible for keeping your site fast.  Something like "the homepage of our site will load in under 2 seconds at the 80th percentile measured with 100 synthetic tests from Chicago during peak business hours".  This is his talk:  http://velocityconf.com/velocity2010/public/schedule/detail/14327.  The numbers you use don't have to be synthetic, you can establish SLAs with any kind of data, but I think you need to establish exactly what numbers you are responsible for, or else you could get into a slippery slope situation where your boss says things like "I was browsing the site over dial-up while I was on vacation in China and it took 30 seconds to load our homepage!!!  What's going on?!". 

There are a number of tools you can use to get performance data, some free and some paid.  Boomerang is a good free way to get real user data that you can then compare with Webmaster Tools to see if things tie out, so you can have a better idea of what you should be responsible for:

http://developer.yahoo.com/blogs/ydn/posts/2010/06/performance_testing_with_boomerang/

I hope this helps.

-Jonathan

[Sreeram Ramachandran]

Wonderful reply, Jonathan. I agree with everything you said.

[Jonathan Klein]

Thanks!  It's definitely nice to hear that from the source :-). 

[beussery]

Hey Guys,

I think there may be some confusion here about the difference between site performance and PageSpeed.  Google Webmaster Tools provides data about site performance.  Site performance data is from toolbar users in the same region the site targets and includes the total time from when a user clicks a link until the time the page onload event fires.  

In other words:

PageSpeed + redirects + latency + distance time + other stuff :) = site performance

Because site performance data includes things like redirects it seems impossible for this data to be derived from analytics.

-Brian

[Brian Ussery]

Hey Guys,

I think there may be some confusion here about the difference between Site Performance in Google Webmaster Tools and PageSpeed.  Google Webmaster Tools provides data about Site Performance.  Site performance data is from toolbar users in the same region the site targets and includes the total time from when a user clicks a link until the time the page onload event fires.  

In other words:

PageSpeed + redirects + latency + distance time + other stuff :) = site performance

Because site performance data includes things like redirects it seems impossible for this data to be derived from analytics.

-Brian

On Fri, Jan 14, 2011 at 2:15 PM, Sreeram Ramachandran <sre...@google.com> wrote:

--
Brian Ussery
google.com/profiles/beussery

[Alex McAuley]

Hi Brian.

 

I don't think your statement is entirely true.

 

The site (which made me start this thread) is a .co.uk site with a UK target audience I went to a European country (namely France) for a little under a month and watched the site performance in Webmaster Tools go down and down  (up in the graph!) - since France is not in the UK perhaps the statement should be changed to reflect this as other users may get confused by this.

 

Kind regards

 

Alex

[beussery]

Hey Alex,

You said, "I have a really strong feeling that Google is using analytics data for people with page speed installed Along with thier own bots speed when determining site speed for  a site and putting it back into webmaster tools."  You asked "Can anyone confirm that this is the case with site speed in webmaster
tools?"

I replied that Google Webmaster Tools reports "site performance" not PageSpeed and this data comes from Google Toolbar users and not analytics.

It's possible that you influenced site performance in Google Webmaster Tools but only if you use Google's Toolbar.  Either way, thank you for your feedback hopefully this makes more sense...

-Brian