SanDiegoFreelance - 1:31 am on Apr 9, 2011 (gmt 0)
Are we comparing apples to apples and oranges to oranges? A page bounce being when somebody hits the back button vs leaving the page or turning off their computer.
For google they now show an option to "Block all example.com results" when they back bounce. This is very different than when somebody goes to an authority hub, then to a retailer for to buy the product. Although most stats programs combine the number into the "bounce" count (They are very different).
I am in agreement with tedster that these bounce back to google do not change the position in the search engines. As well as rlange: " ... they would see me click on the result and not return. Even if I clicked on a second result, that can't necessarily be taken as a negative for the first result. ...". I do research the same way, I never rely on a single source on the internet for information; Hence norms need to be considered. Bounce data is noisy.
When I was running a vertical directory and measured CTR and Bounce. People had all kinds of patterns - some would literally go through the whole list - and often sign up with a affiliate program or two. I did see some sites with low quality have high bounce rates - I removed them - some site closed and had high bounce rates - I removed them. I saw high quality sites with low CTR based on title and description (I adjusted the title and description to properly show their content and improve quality of the directory).
Bottom line is this the data is very noisy, and the hit miss ratio very low, but also very useful. It can not be used in any kind of raw form - but the spam would in the ad-norm group. I can see uses for it when it comes to improving algo filters; that is to say a site fails one test but other factors are normal treat it like a false position.