getcooking - 6:18 pm on Jun 26, 2012 (gmt 0)
Ok @claaarky, I think you might be my new hero.
When I first read about your theory I applied it to my site as a whole and it didn't really seem to show any clear indicators of being applicable, at least not to me.
But when I pulled out only our category pages (the type of page on my site that was demoted in Panda) and looked at things it became a bit clearer.
I have 3500 categories I'm working with so it's been hard for me to get a really good (easy) way to look at Google Analytics compared to some other internal data that I use to judge my own page quality. I exported just the categories from GA into my own database and cross referenced a few things that look at internally.
Sure enough, the high exit rates are a pretty good indicator of what my own internal data considers "shallow" content. It's not a 100% foolproof indicator, however. There are definitely some exceptions to the rule but the high correlation between the pages I already considered shallow and the exit rate are a pretty good indicator that I wasn't the only one considering those pages shallow!
Interestingly enough, a huge majority of the pages I had already noindexed had the highest exit rates. I still have more to noindex it looks like but now I'm thinking it might be why we had a 15% recovery in Panda 3.7. I noindexed a few more since 3.7 and we might have had up to a 5% recovery from yesterday's update (still waiting for the dust to settle).
I don't recommend to anyone just removing (noindexing, deleting, merging, etc) pages based on this data alone, however. I'm looking many, many factors on my site but the exit rate is very helpful now that I'm looking at it from a different standpoint. Users shouldn't be exiting from a category page. They should be exiting from a content page. If large amounts are exiting instead of clicking through to a content page then they aren't finding what they want. I'm also looking at what internal pages link to those underperforming pages (maybe they aren't linked internally in the best way), the number of external entrances to that page, % visits from search vs total traffic to page, etc.