suggy - 8:36 am on Jun 22, 2011 (gmt 0)
Improved in May Panda, only to be slapped down again in June. Boo hoo... No time for self pity though....
I have three possible theories on what's going on:-
1)Google is doing some kind of document classification that works on a search phrase basis (notice how when you 'twist' the phrase a little you shoot back to the top?).
In short, Google compares everything in the results set with everything else, computes means or standard deviations (or similar), maybe using n-grams and assigns multiple scores that filter pages down according to what's being searched. It then builds an overall site-wide score based on means. So, your pandalized pages screw themselves and drag down your site.
The fact that it has to build these comparitive matrices of search words and scores and might explain why only periodically calculated.
This doesn't really explain how come pages with very limited content win out. Although they may benefit from the absence of content to critique and a good site wide score.
2) Google is using the greatest resource at it's disposal to rate quality; it's searchers. Again it has to build complex matrices based on search terms/ keywords comparing everything with everything (an 80% bounce is not bad if everyone else's is 99%!). Individual pages are pandalized and then a site wide mean score is applied, affecting domain trust/ authority and pulling everything down (from a bit to a lot).
This kind of fits with google's original objective of tackling content farms. After all, they didn't hate content farms per se, they hated the fact that users hated content farms (and were therefore not getting optimum satisfaction from their search at Google).
Explains why update is infrequent (lots of data to collect and mash-up).
Explains why good content on ugly/ non-user-friendly sites still get's pandalized (like that book review site mentioned here - sorry no offence to owner).
Explains why big brand sites do well; people give them more of the benefit of the doubt and spend longer trying to satisfy their search. Unbrand have a harder time convincing folks there's any hope of satisfying their needs there unless it's bloomin' obvious.
Explains why e-how bombed less than ezinearticles; the latter was just way to ugly and the former superficially engaging (like one of those chat magazines!)
3) It's a combination of the two. Classify the worst statistically, then let the public decide.
Anyone support any of these ideas/ dispute them/ have an alternative theory/ have anything to add?