Hi, first post here but have been following most of the related threads for more than a year, so take it easy. Firstly, sorry for the long and disjointed first post, but I feel some of my thoughts may be useful within this thread.
@claaarky I tend to agree with you in part, that this could be a helpful metric in identifying weak pages but, in a similar manner to 'Rockzer', I very much doubt this is the be-all-and-end-all of the Panda layer to the algorithm.
I've been paying attention and reading between the lines to what various Googlers say or imply and Matt Cutts has said that 'serp blocking' has been used in Panda. I believe this is probably in a confirmation capacity to Panda's findings.
Recently at SMX Advanced when asked the question about 'bounce back to serp rate' I found it very telling that MC skirted around the issue, tried to change the question around to imply 'GA bounce rate' and then stated they don't used that 'noisy' metric. He was further pressed and told that wasn't the question and still gave an inconclusive politician's answer. A lot of the transcripts at the popular SEO news sites seem to omit the meat of this exchange.
Early this month (June) a question was posed on twitter to Matt Cutts:
if you got hit with both penguin and panda should you just give up?
to which the reply was
I think the site that set the record most recently had nine completely different things that we flagged on it.
Whether the reply was regarding Panda, Penguin or both is open to interpretation. I feel those 9 things can't cover all Panda browser metrics and Penguin linking or site metrics. I may be off the mark here I take it to read Panda (can) equal on-site.
Reading the thoughts and findings of certain patent bloggers gives an idea of how complex (or simple) Google could really be.
My view of Panda, in really simplistic terms, is like so.
Metrics were used in the initial data seeding of Panda, with the help of Quality Raters, and sites were put into taxonomies/categories. Sites with only enough of certain metrics were used and were put into good and bad piles. They then looked for differences with these groups, that they could get a machine to identify, that were highly likely to apply to one group. Then the algorithm was run across all caches on the relevant data centres. As the machine learns, new factors are possibly being found and identified that correlate strongly to certain groups within certain category of site. Metrics could then possibly be used to try keep to Panda on-track with the different groups in different site categories. Also, as Panda's evolved, groups have been split down into less distinct groups with very fine differences between them - hence turning more into a ranking layer which could move a site up or down a little (should things be 'not to bad'), as-opposed to showing more of penalty effect many initially saw.
So maybe not so much about the metrics, which explains why smaller sites with inconclusive metrics can get hit, but it's about 'Perceived Metrics'. So I think that claaarky's right but also think those that disagree with him/her are also right.
The thought on keeping users going/navigating in the right areas is an interesting one and quality sites, in certain classifications, would be able to do this.
I have a site that was hit in April 2011 that I believe was showing improvement month-on-month from August last year until the Penguin recently started to peck at it. I tinkered too much with it but everything seemed to lead to perceived user satisfaction, mainly stemming from the fact that a site that updates nearly all of it's pages in this sector on an almost daily basis is likely to be a quality one for the user, plus there were a few content/grammar issues.
Another site is an old blog which isn't updated, some of the posts could make an semi-interesting read for the search terms they're found for. This site seemed to be fine until the 'Page Layout' algorithm, and has a large Adsense block floated left of the content with a large banner ad at the top of the right side bar with another large Adsense block below. That's not the issue and I've recently removed the left floated block to see if this was the case and time-scales etc...
On this second site there's a subdomain which fits the same sector as the first site I mentioned. It's really just for testing, is pretty low traffic and has only 7 pages, all have sailed through Panda and ranked at an acceptable level until April 27th update. These pages have a basic layout, OKish text and the information that users would visit for is populated via xml from the database of the first site mentioned. Wording pulled from the xml feed is unique as I have a double entry system in place.
Mid-April I added 2 links right in the centre of the eye-line on each of those 7 pages. These pointed to related sites offering the same/similar info and I did this because no sites in this sector do. It may sound mad but this was a little test that would get users to a better site quicker, showing lower satisfaction and worse metrics with this site and not looking around as they'd found something better.
April 27th saw Panda 3.6 drop all rankings of those pages like a stone, to well outside the first 100 results. So I removed the links from these pages and left it.
June 8th then saw Panda 3.7 reinstate all rankings of these pages.
Rankings were altered so abruptly on those dates and the links were the only changes. It's a possibility it wasn't the links and that the site sits on the cusp of being in or out, but that seems slim to me. I don't believe Google could have collected enough 'real' user metrics to make this decision at all, they sent <10 visitors during this Panda'd phase - which makes me firmly go along the lines of 'on-page factors that fit in with poor metrics/satisfaction'.
[edited by: danwuk at 10:52 am (utc) on Jun 27, 2012]