diberry - 6:47 pm on Jun 28, 2012 (gmt 0)

I'm really happy that claaarky shared his insight here. He noticed a very strong correlation between high Exit Rate pages and pages that were dinged by Panda. Then he took a hard-nosed look at those pages and they looked very weak to him and not just to his visitors - so weak that he started a project fixing them. That sounds really sane to me.

I feel what Claarky is really onto is a user metric that, in many cases, tells a webmaster a lot about how much visitors are liking the site. In fact, that was my only interest in this discussion. I ignore Google and build my sites for users, but I wanted to make sure the Exit Rate wasn't a valuable metric I had missed - not for beating Panda, but for knowing if I'm pleasing visitors.

But I reject the notion that it's the key to Panda. We've offered numerous examples of cases where it would actually be a really poor indicator of visitor satisfaction - quick answer pages, for example. Recipes. Sites where people check in to a community to see if there's any news, and upon seeing there isn't, back out and go somewhere else. Google is plenty smart enough to realize that exit rate, useful as it is, just cannot be the center of their calculations.

I think what Panda may be doing is parsing pages into niches, and then comparing various metrics within each niche so that your page is getting compared to similar pages. If that's true, engineers would be able to create metrics profiles for an imaginary ideal site in each niche, and compare your page to it. I don't know if that's what Panda is trying to do, but that's the feel I get from it. If so, there is no single metric - not even a single collection of metrics, or ranking of those metrics in importance - that applies to every site across the web.