deadsea - 1:41 pm on Nov 14, 2011 (gmt 0)
Google internally refers to a user that doesn't bounce back as a "long click". A long click is a click that leaves google and doesn't come back for a long time: until that person wants to use Google for an unrelated search. The user doesn't refine their query, nor does the user use the back button and click another result in the SERPs instead.
I agree with goodroi, I have a site that I am trying to satisfy the user on the landing page itself. That site should have a 100% bounce rate (closer to 80% right now) but it has an amazing user experience. So far it is doing phenomenally well in Google search.
On the other hand, bounce rate is easy to measure and can be a great proxy for long click on a site with deep content. I worked with a large product website. We noticed that some products we ranked really well for and some we couldn't rank for no matter how many links we had for them. We noticed that the products that ranked well had lots of content (reviews, photos, editorials, places to buy, etc) and the ones that didn't rank had very little content. We started looking at what content did for our bounce rate. With no content, a product page would have a 70% bounce rate. With lot of reviews, lots of photos, and some of every other type of content, the product could have as little as 10% bounce rate. (We also didn't count clicking on ads as a bounce.) In that case, I believe that our bounce rate metric very closely corresponded with the long click metric that google was actually using and we did very well to work on improving our bounce rate instead of trying to measure long click more directly.
But it means that bounce rate "improvements" like splitting articles into multiple pages are just gimmicks. They won't help your rankings. They will probably hurt your rankings as they hurt user experience and the long click.