Interesting read.
http://www.searchmetrics.com/media/documents/ranking-faktoren/download-ranking-factor-study-2013.pdf
I do like a nice correlation graph.
Got to take some time with that one!
Markus from SearchMetrics gave a great presentation at Pubcon on social media impacts. While everyone is getting hand wavy and making claims based on vague info, he had tons of data, was detailed in how the data was obtained so you could judge its value for yourself and then threw out his conclusions. I developed a lot of respect for the Search Metrics crew.
It's a great read but I do feel 2 things need to be remembered.
1. Correlation is not the same as causation.
2. If you cover of all correlated points you are likely to include all of the points that delivered causation!
Quote from: JasonD on June 26, 2013, 04:20:29 PM
2. If you cover of all correlated points you are likely to include all of the points that delivered causation!
But not necessarily in the same order of importance. Notice, for example, on page 14 that "existence of an h2" is correlated to high rankings more closely than "existence of an h1". That of course is not because an h2 is more important for high rankings, but due to the fact that almost all high ranking pages have an h1, so there is little correlation between h1 and ranking.
Obvious, I know... just picking that example because we all know that one is obvious and we of course need to apply the same filter to the ones that are not immediately obvious.
Still, very interesting report and everything I see from the Search Metrics guys makes me think they're pretty damn smart.
Definitely some food for thought in there.
Perhaps the only thing I immediately questioned when reading through it was that the EMD update has only rolled out in the US.
QuoteInterestingly,
Google only seems to have rolled out the EMD update in the US. No algorithm change has been recorded outside google.com to date.
The reason may be perfectly obvious: In other countries ‒ for example, in Germany and also in the UK or France ‒ the ranking of most of the suspicious domains has already been successfully degraded in the course of the Penguin update, and therefore, another algorithm change may not appear necessary in this case.
That's not what I see. I'm looking regularly at several European markets and saw an immediate drop for EMDs in my datasets around the time of the US release, and continued decline ever since.
My datasets are going to be far smaller than Search Metrics' set, so surprised they haven't seen the same thing, though they do give a clue as to why...
QuoteHowever, one question still remains: Why is the correlation for the factor "keyword in domain" still positive? Here, too, the answer is not far away: brands.
They talk a lot about the influence of brands in this report. It would be useful to be able to see the same analysis conducted over a keyword set that has all brand keywords removed.
Frankly
brand.com is almost always going to rank immovably in the top spot for "
brand" keyword. Major brands have huge search volume for their brand names, and this can skew the dataset considerably when analysing ranking factors
Are you guys seeing as much social influence as this report is suggesting?
>That's not what I see. I'm looking regularly at several European markets and saw an immediate drop for EMDs in my datasets around the time of the US release, and continued decline ever since.
100% agree. Twas a blood bath too.
>Are you guys seeing as much social influence as this report is suggesting?
No. None at all, yet.