SEOMoz’s Matt Peters Research: Well Worth a LookPosted 8 October 2012 by Phil Smulian
Rather than attempting to explain this absurdly complex topic to you all by myself, I’d rather recommend that you first read through Matt Peters post on SEOMoz about correlation analysis between Mozscape data and the big G’s rankings.
You might want to take a stroll down to the local beach with your laptop/tablet and an umbrella, or at least prop matchsticks between your eyelids and a broom handle under your chin. Matt whips out his guns and delves into the deep waters of recent algo activity. He did some highly interesting analysis into recent Google algo changes, looking at ranking correlation in the face of the 7-SERP and the ‘domain diversity increase’ phenomena. More than reaching many definitive conclusions, the analysis seems to have raised many questions.
Will Critchlow asked, has anyone seen 7-serps around the world? gfiorelli1 reported “Scala di Milano” in Google.it (Italy). This SERP showed me 10 results, perhaps one of the SERPS that switches arbitrarily between 7 and 10 results.
carl joel reported a brand and a knowledge graph object :
Both showed me a 7-SERP.
I picked up a 7-SERP result right here in my home country, via Google South Africa.
http://www.google.co.za/search?q=telkom&rlz=1C1LENN_enZA474ZA474&oq=telkom&sugexp=chrome,mod=0&sourceid=chrome&ie=UTF-8 – Telkom is our local, and very dominant, and widely scrutinised, telecommunications / internet provider – a big brand.
From this it seems fairly clear that the change has been rolled out internationally (South Africa always gets things late, like the awesome Dredd movie that was out in the US and UK ages before here), and affects certain queries and then, well, doesn’t. what is covered in Matts post is that 7-SERP results made up 144 of the 1000 keywords used in the test set, 808 were 10-SERPS, and 48 were those SERPS that change their mind every short while or so.
donthe mused that “since 7-SERPS are usually for branded terms, it sort of makes sense that results should be domain level and not page level”. Matt responds that a significant portion of the 7-SERPS are generic terms rather than branded, and that 7-SERPS correlated well with page level metrics, and almost not at all with domain or subdomain level metrics.
One question I’d ask is if Google deliberately reduced emphasis on domain level metrics for 7-SERPS, simply to counter balance the emphasis that other “brand friendly” parts of the algo already possess, such as from the Vince update.
Google seems to be changing the search experience more frequently than the colour of a chameleon on a roulette table. Perhaps they are cranking up page metrics as a factor, after having turned the industry on its head and giving ridiculously bias favour to (good or bad) pages on powerful domains. Maybe the algo can safely return to its most effective state, now that the spammers and gamers have been crushed and have abandoned all their less than spotless projects.
With Google changing their core algo like they do their underwear, and the rapid introduction of new products and ideas, it’s difficult to tell what’s coming next, or how to prepare. They have the industry in thrilling suspense at all times, like a Thomas Harris story.
Anybody else noticed more 7-SERPS, or interesting facts relating to them, or examples of strong pages that have beaten lesser pages that are on hefty domains? Please share with us! Unlike a B-movie buff, we love a good story.
Weightlifter image courtesy of the Official U.S. Navy on Flickr.