While continental Europe is on the edge of its seats, waiting for the Panda Update to be implemented, both the US and British indexes are slowly getting a breather. There, the changes are already made and the dust has settled somewhat. This is surely way too early for a final verdict on the Panda-Update but I feel like putting up some thoughts and conclusions for discussion.

The Panda-Update has its roots in Google Caffeine. This is the codeword that Google used last year when they implemented a reworking of their search-infrastructure from the ground up. Since then, new pages as well as ranking-signals can be added to the index much more quickly and the total size of Google's index has also grown massively. This has also become a problem for Google: pre-Caffeine, only a limited number of URLs from domains with questionable content were indexed, which worked as sort of a natural filter. It is this filter that Caffeine got rid of, which means that quite a lot of low-quality content has found its way into the index and can now be shown. Panda is Google's way of getting a grip on this problem by cleaning the rankings of content that may not clearly violate the spam-guidelines, but which is still not up to par with Google's (and the users) quality expectations.

Over the past few weeks, I have already written about domains that were affected by Panda. The first Panda-Update in the US surely caused the largest commotion but both the UK-version as well as the update for the US gave us access to some interesting data-sets. Since then, a discussion has been going on about the possible causes: some think that Google upped the amount of brand-searches, many are of the impression that Google rates the quality of text on a page, while others believe that a reduction in indexed pages could help. It is too early to go and work on solutions, seeing how we have yet to see a domain that manages to substantially bounce back after getting hit by Panda. Personally, I think that Google is assessing a multitude of signals at once. When a large enough sample of these signals for a domain show a large discrepancy from the norm, a filter gets activated. This is where it gets interesting to start speculating about possible signals that are used in evaluating a domain. The comment of a Google-employee on one of the support-forums let us know that, at the moment, it seems that Google is still content on using domain-wide signals:

„In addition, it’s important for webmasters to know that low quality content on part of a site can impact a site’s ranking as a whole.“

For a closer look, it is interesting to delve into both qype.co.uk as well as yelp.co.uk: both domains have the same business model and while Qype got hit, Yelp managed to be spared by Panda. Many of the metrics for both domains are quite similar, though once you look at the 'time on site', the pageview/user as well as the bounce-rate, you start to notice some differences. The following numbers are Alexa-numbers and while I am aware that these numbers are to be taken with a grain of salt, we are looking at two very similar sites and the differences between them, which should be enough to make this work:

	Qype.co.uk	Yelp.co.uk
PV/User	3,15	4,39
Bounce %	61,7%	52,0%
Time on Site	2,6 min	4,8 min

You notice off the bat that Yelp has the better numbers, across the board. Similar “domain-pairs” (one hit by Panda, the other spared) show striking differences in these metrics as well. Here for example, we have Pricerunner.co.uk as well as Ciao.co.uk:

	Ciao.co.uk	Pricerunner.co.uk
PV/User	2,11	3,10
Bounce %	63,2%	47,6%
Time on Site	1,7 min	2,5 min

For all these metrics, I would advise not looking at the absolute numbers but at the deviations a certain domain shows in comparison to their competition. And while we certainly cannot be sure whether Google is using such metrics in their decision-process to use a Panda-filter, I feel confident that this theory has a lot going for it. It is certain though, that numerous other signals will also affect the outcome.