Msg#: 4215221 posted 5:36 pm on Oct 8, 2010 (gmt 0)
Around September 20th, all of our site performance data improved substantially. This is the chart that is at Webmaster Tools-> Labs-> Site Performance.
We made no changes at all on any of the sites yet they all showed big improvements of speed. One went down from an average 2.8 seconds to 1.7 seconds. This moved the site from "faster than 57% of sites" to "faster than 76% of sites".
The site also states that "these estimates are of high accuracy (more than 1000 data points)." Previously, they all said "these estimates are of low accuracy (fewer than 100 data points).
The sites are all on different servers located in different parts of the U.S.
Msg#: 4215221 posted 6:49 pm on Oct 11, 2010 (gmt 0)
What I have noticed is Google is only reporting our site speeds when there is downtime of maybe 5-10 minutes a month. Matches up exactly with our reporting. Since I build any site to be super fast it is still a bogus figure to me.
Msg#: 4215221 posted 2:13 am on Oct 12, 2010 (gmt 0)
Have you had certain pages/scripts accessed more than before 20th (or some other pages less accessed than they were visited before 20th)?
What I noticed is that the time performance from Labs depends on which pages are visited by users with G. toolbar. If you have a fast loading page that for whatever reason suddenly becomes popular and is visited more (by users with G. toolbar), then your overall site performance will show as faster.
Or if some slow loading page went out of visiting favour - you will most likely also see improved speed.
It seems that opposite is true too - if you have a slower loading page that becames more popular with visitors (or your fast loading pages lose the popularity) then your overall site performance is going to be shown as slower.
So I would suspect that since 20th Sept. your visits came in higher proportion to either your fast loading pages (or in lower proportion to your slow loading pages) - which will both show speed improvement in Google Labs Site Speed - perhaps worth checking?
Msg#: 4215221 posted 2:23 pm on Oct 12, 2010 (gmt 0)
aakk9999, interesting thought. We did add a couple of fast loading pages (no graphics) during that time on one site, but that site has over 6,000 pages, it's hard to believe that this would make that big a difference, but maybe it does?
Also, some of the other sites had no changes at all made to them, but one other site had many new pages added and it showed the most significant drop.
Msg#: 4215221 posted 1:31 am on Oct 13, 2010 (gmt 0)
I don't think it is necessary influenced by the size of the site. The way I see it, if you have 6000 pages and only one page is really fast, but this page draws a large proportion of visitors, then this page will contribute to your site speed calculation proportionally to visits it receives versus the total visits and not proportionally to number of pages on the site.
So, 6000 pages, 6000 visitors, each visits one page - you get true average site speed. Or 6000 visitors, half go to very fast loading page - and the rest spread over your other 5999 pages - your average site speed shows as faster. At least this is my (crude) observation.