Search Metrics SEO Ranking Factors are WRONG

Search Metrics has been the source of a lot of confusion, starting with their 2012 report.

Another year has passed, and they released their 2013 Correlated Ranking Report for SEO ranking factors today.

Let me start by saying the 2013 report strikes me as both inaccurate and damaging to the overall health of the web.

Check it out:

SEO Ranking Factors - SearchMetrics2013

So what has me so convinced their report is flawed?

  1. Keyword in Title was given a correlative coefficient of 0. Really?
  2. Social metrics are shown as massively correlated. For example, +1’s are shown with a coefficient of 0.04.  How about some beachfront property in Montana?
  3. The general impression their conclusions are in many ways opposite of what is true. Methinks Search Metrics is living on Pluto, not Earth.

There are many reasons for why their conclusions are flawed. I’m no mathematician, but my assumption is they’re showing disproportionately high relevance for factors that experience significant variance. I suspect they are experiencing issues with collinearity, where the correlations among the independent variables are strong. Do you remember “correlation is not causation” from philosophy class? In a nutshell it means when two things happen at the same time, one is not necessarily the cause of the other.  Collinearity also causes some variables to be statistically insignificant when they should be otherwise significant. It’s easy enough to land 180 degrees off if you have no clue how to look at something.

So why does their report show high correlation between social metrics and rankings?

Look at a set of results from Google for a random search query. You will find the biggest variance in data for the following factors:

  • +1’s,
  • Likes
  • Tweets
  • Pins
  • Length of URL
  • Number of backlinks

Some sites don’t add social sharing to their sites. Others do. Those who focus on social media may see thousands of likes while others who don’t won’t have any. The number of backlinks vary significantly, too. These same factors are all ranked highly on the correlation coefficient graph from Does this mean you should be trying to get more likes in order to improve your ranking?

No, absolutely not.

Social metrics don’t matter.

At all.

Not even a little bit.

Search Metrics is drinking some strong potion. I’d like a glass of whatever they’re having. It sounds like a lot of fun!

Don’t take my word for it.

Here’s Eric Schmidt, telling it like it is.

The video above clearly shows some frustration regarding the supposed level of data Google is able to collect from Facebook. Twitter is no different. Google has no better grip on Twitter’s data than it does on Facebook’s data. The Twitter firehose is locked down for everyone but Twitter’s favorite customer-developers. Imagine a world where Google is one of Twitter’s best customers. Maybe Search Metrics is operating in a completely different universe. That’s one explanation for the conclusions they’ve drawn. Twitter stopped supplying Google with firehose data almost 2 years ago, back in July 2011. 

More about Twitter and Google (links courtesy of Marketing Land):

We don’t need to have a debate about whether or not social metrics would be good for search. Of course they would be. We don’t need to have a debate about whether or not Search Metrics’ correlative coefficients are wrong. We already have all of the answers we need. They are wrong. Google doesn’t have access to either Facebook data or Twitter data. It’s impossible for these data sets to be part of Google’s algorithm. It’s just as impossible as eating a cake you don’t have.

What About That “ZERO” Coefficient for Title Keywords?


Search Metrics jumped the shark, that’s what.

Your page isn’t even selected to rank for the query if you don’t have the keywords, or some closely related keywords, prominently on the page. Google’s transparency page uses this image to describe the process of indexing and returning queries from the web. Google uses “barrels” to quickly identify relevant pages. Those relevant pages are then sorted by what most SEO’s call the “algorithm.” The “sorter” is where the “ranking factors” work their magic. It’s where you’ll find the popular Panda and Penguin algorithms hanging out.

What Does That Mean?

Content relevance ranking factors will always look less important than they really are. Why? Because only relevant pages show up in the results that are analyzed. However, the relevance of these ranking factors would appear to be extremely high if Google sorted the web without using the “barrel.” Without content relevance ranking factors, the page won’t even show up in the “barrel,” which is its first stop before it’s sorted again by the algorithm.

Imagine if a relevant keyword in the title really had a coefficient of 0.

You might search for “puppies” and get “Viagra” in your results (Because Viagra sites have way more links than puppy sites).

The Search Metrics report misleads new SEO’s. Ultimately, the web ecosystem will be hurt by people who don’t have the first clue about how to interpret Search Metrics’ 2013 ranking report.