|Ranking based on page load speed seems to be in full effect|
| 1:59 pm on Jan 8, 2010 (gmt 0)|
Here's an experiment I did from August/2009 to the first week of 2010(yes, concluded days ago). My observations got me excited so I decided to share something done in months with all of you.
I installed Analytics on pages which had AdSense in them. I made pages that would sleep(N) where N is a number of seconds, between AdSense loading and Analytics(at the bottom). autoflush on so content was sent to the client as soon as it was produced.
Slow pages ranked terribly, fast pages shot straight up. This is a good reputation site so I was able to get pages spidered quickly for my experiments, though the keywords used were not very important, so as to be able to test without spamming their index or anyone seeing the tests on relevant SERPs.
The observation which brought me here: Difference between fast and slow pages was about 30 positions.
I took analytics out in September and both pages ranked closely. In October I decided to test the contrary, once more, to be sure. Added analytics and the slow pages fell 30 positions. Remove analytics once more and pages shot up. Each boost/drop took about 10 to 11 days.
So that's the closest I've gotten to know that Analytics is sending back your page load speed in comparison to AdSense or some other Google widget on that same page, such as Maps.
Not only is it sending speed back but Google is also taking data collected on Maps, AdSense and Analytics in consideration for ranking purposes. Of course I can't be 100% sure, so this is only an observation which may be right or entirely wrong.
| 7:46 pm on Jan 8, 2010 (gmt 0)|
Thanks for sharing your test results, kjennings2. Interesting.
How fast were the fast pages, and how slow were the slow pages?
| 8:12 pm on Jan 8, 2010 (gmt 0)|
When you say Analytics do you mean Google's product? I've recently been removing Google Analytics from sites that I don't really need detailed info on and it does seem to speed things up (I still use awstats on the server). I haven't noticed any difference in traffic/ranking due to that though.
| 8:29 pm on Jan 8, 2010 (gmt 0)|
Some people think that Google uses visitor behavior as a ranking factor. If you made the pages extremely slow, visitors could have tired of waiting for them to load and left the page early. Google's algo could have interpreted that as a negative vote, and dropped the rankings of those pages for that reason.
| 10:35 am on Jan 9, 2010 (gmt 0)|
I'm very curious what the load time was for both, the slow and the fast pages in the above test.
While load speed may indeed be a factor, it is going to be just one of many factors. On my site I have a product review page with numerous images. The page is around 750 kb in size! Huge! Yet, even though the size makes the page load time slow, the page does perform well. In searches for the product name (from various computers, and personalized search off), the review comes up #1 in Goog** (both .com and .uk), out ranking the product developer, Amaz**, and Wikip****.
(I will work on compressing the images today!)
| 12:36 pm on Jan 9, 2010 (gmt 0)|
@kjennings2 Could you explain you are talking about client side pageload or server side? "autoflush on" shows that you are slowing down main page on the server?
I have experienced, direct relation between server side generation speed and traffic, but havent tested it in depth.
My real total client side page load maybe high... but the page is viewable before everything loads up... hope Google is taking that into account.