[Case Study] SearchMetrics 2013 Most Important Ranking Factors

[Scritty]

A couple of points here.

Firstly this is a "Rank Spearman" correlation of factors that are common to a very specific and select sub group of sites that are ALREADY POPULAR. The top sites for the top 10,000 search terms. Might seem like a lot of breadth there, but there is not. "Bank" Insurance" "Diet" "Car" "travel" "money" "mortgage" "website" "school" ... I could go on and on - and that would just be the shortest of short tails (one word) search terms add "weight loss" "payday loans" "life insurance" "foreign travel" and you can quadruple the single word terms and the 10,000 figure is gone pretty quick while never even touching the sort of long tail your average affilaite would even consider going for.

Then they are only looking at the BEST sites in these terms.

Tis is a fraction of the number the guys at SEOmetrics look at (16 million sites completely randomly chosen) as a study it is a factor of about 100 x smaller in terms of dataset searchmetrics and seometrics are two different teams however.

Ok - this is very important as it means two things.

Firstly - It does not sort out cause from effect. These are popular sites to begin with. It begs the question..
Do social links create popularity and a rise in SERPs and traffic potential.. or do already popular sites attract more social links and therefore the addage that "correlation does not always equal causation" should be applied or at least considered?

Secondly - one way or another - the top 7 in the list are links. Be they social or otherwise. A retweet is a link, a facebook like is a link a Google +1 is a link. Yes they are social links - but links all the same

A take away from this might be that this is not a random selection of sites - this test was carried out on a preselected group of already popular sites. The rules for aged authority sites that attract a 7 figure number of visitors per month has always been VERY different to small sites with reaches a fraction of that.

Overdoing social signals can be the kiss of death. The BBC's coverage of the birth of the Royal Baby last week garnered under 1000 Google+1's despite being read by 17 MILLION people in the first 48 hours. (sdo about one in 20,000 visitors "+1d" it). Facebook likes were a lot higher but still under 15,000 (so under 1 in 1000 people "Liked" it)
Are google going to believe that some affiliate site about radiator covers from an Amazon affiliate is really going to get more +1's or "Likes" than that?
(Trick question - the answer is "No they fucking aren't")

Anyone with a scientific mindset knows that any conclusions extrapolated from a "massaed" or pre selected data set rather than a true random sample is at least something to be wary off - if not downright dangerous to draw conclusions from.
There was a very specific dataset used for this survey - hgh authority aged sites with extreme levels of pre-existing traffic.

If your site does not have that profile - then these results might well not apply to you.

But again - the fact that the first 7 lines on the chart are links of one type or another is pretty compelling even if regarding in purely macro analysis terms.

Great info. Thanks given.

Scritty

 

[QuantumBlue Bulgaria]

Social signals are important after all but we should not focus too much on it just like not focusing too much on other strategies. SEO is not a one-way course but nevertheless, this is excellent info and helpful.

 

[oozyluce]

Anyone with a scientific mindset knows that any conclusions extrapolated from a "massaed" or pre selected data set rather than a true random sample is at least something to be wary off - if not downright dangerous to draw conclusions from.There was a very specific dataset used for this survey - hgh authority aged sites with extreme levels of pre-existing traffic.

It is undeniable that this study holds scientifical hurdles, like you said the sampling is not random and the pool is rather limited. However i don't think that this means that it is always unpertinent to draw data from a pre-determined sample, all depending on what you're analyzing.

I see this study more as a statistical analysis than a scientific experiment. And where you read "statistical analysis" you read a strong interpretation value in most cases... And this implies by itself that you can't treat this as your new SEO bible, unless you want to keep chasing the proverbial carrot.

It is nontheless a worthwile study to evaluate the general algorithmic ranking factor of google in itself.

Ok - this is very important as it means two things.

Firstly - It does not sort out cause from effect. These are popular sites to begin with. It begs the question..
Do social links create popularity and a rise in SERPs and traffic potential.. or do already popular sites attract more social links and therefore the addage that "correlation does not always equal causation" should be applied or at least considered?

True it does not sort out cause from effect. HOWEVER. Some analysis points out more in a direction than the other, i wouldn't go as far as them by "declaring" anything, but please do carefully read page 60:

Several detailed analyses were conducted last year with regard to the causal
connection between social signals and good rankings.
For this purpose, two texts16 on two cognate subjects (four articles in total) that
were identical in content were written and published at the same time on the
same domain under as neutral conditions as possible (no sitemap, no comments
function, no internal / external links from or to the URL). Next, social signals
were specifically targeted to the article - under strict network-specific separation.
This means, that two articles only received Facebook signals, whereas the copies
of this text were exclusively shared on Google+.
While observing the strict separation and compliance with the neutral
environment, we evaluated the rankings for specific keywords, which have been
determined in advance and were identical for the respective articles.
The results of these social signal studies included: Indexing only through social
signals, without presence of any internal or external links, is possible; a
significantly smaller number of plus ones is in the position to produce faster
indexing or better ranking for an identical set of keywords in a direct
comparison; with a repeated increase of shares or plus ones after a period of
stagnation, the ranking increases again, even if this has previously also
stagnated or even declined.
Without having to reproduce the case studies here in greater detail, we can
conclude that it appears that social signals can influence the ranking of URLs
even in isolation.

Overdoing social signals can be the kiss of death. The BBC's coverage of the birth of the Royal Baby last week garnered under 1000 Google+1's despite being read by 17 MILLION people in the first 48 hours. (sdo about one in 20,000 visitors "+1d" it). Facebook likes were a lot higher but still under 15,000 (so under 1 in 1000 people "Liked" it)
Are google going to believe that some affiliate site about radiator covers from an Amazon affiliate is really going to get more +1's or "Likes" than that?
(Trick question - the answer is "No they fucking aren't")

True all the way. Overkills are only positive in Halo and Call of Duty



P.S: I really like the way you deliver your insight and knowledge about SEO around BHW, your posts are always so interesting... Thanks for that!

 

[baldonius]

Correlation does not imply causation.

 

[Scritty]

Correlation does not imply causation.

Yes - it does Correlation DOES imply causation.

Correlation doesn't PROVE causation

Solid conclusions infered from impilcations are the problem - that is the mistake people make.
The fact that you can draw in implication that makes sense logically (cause and effect) and yet has no proof to substantiate that the logical path you are taking is correct is where many people come unstuck.

Without the possibility of implication - people wouldn't make the mistake in the first place.

Scritty