At SMX London 2011 I presented the results from a few experiments I’ve been running over the past few months on the impact of Facebook Likes on search rankings. For those of you who didn’t manage to make it, here is a write-up of the experiments and results that I presented.
Following on from a great talk by Bas Van Den Beld on the changes that Google have made to become more social, I began my presentation by asking the audience whether they thought Facebook likes were an indexation factor – the response was mixed.
We know that there is correlation between likes and rankings, but is there causation too?
There are currently multiple studies that have suggested that the correlation between pages that are publicly liked or shared on Facebook and pages that rank very well on Google. In fact, Jim Yu who also presented on the social signals panel backed this statement up with his own set of data. However, there is still a lot of ambiguity around whether Facebook likes cause good rankings and indexation, which is exactly what I wanted to prove correct.
In search for some authoritative statements I found two of particular interest, one directly from Google in a Search Engine Land interview, which said:
“We treat links shared on Facebook fan pages the same as we treat tweeted links. We have no personal wall data from Facebook.”
And another from that an interview Barry Schwarz didwith a Google Webmaster Tools engineer on Search Engine Roundtable:
“Right now (March 2011) Google use Twitter, Google Reader, and other sites but currently do not use Facebook. So Facebook has no real direct impact on rankings as of yet”
My First Experiment (February 2011)
For my first experiment, I took two unindexed domains; one that was two years old and one that was two weeks old. Both had 0 back links and no ability to ping Google. I began progressively adding links into both of these domains and the result was that both domains became indexed within several hours of first adding Facebook Likes. There is more information on one of these tests over on the blog post I published on the SEOptimise blog.
My Second Experiment for SMX (April 2011)
Following on naturally from my previous experiment I wanted to scale this up and see whether you could hypothetically add likes to 1000 unindexed pages on an ecommerce site and get them all indexed. I started by taking 100 unindexed URLs across 11 different domains with varying domain and link history and tracked the raw log files to identify when Googlebot first visited the URL after being liked. I wanted to see whether the quantity of likes and type of like (share, fan page like, personal like) affected the duration it took for the Googlebot to first visit the URL.
The Result: Dislike
0 URLs got indexed, and there were 0 visits from Googlebot in the raw log files.
Did Something Change? Given that the first experiment was ran in February and the second was mid April, I thought perhaps this might have something to do with the Panda update or the launch of Google +1, but in many ways this is irrelevant.
In my opinion, the value of social signals for search is not in the direct signal itself, but in the traffic it drives. Facebook can drive significant volumes of traffic when your pages get liked, which in turn can generate back links and provide behavioral data, which is supposedly becoming more of an important ranking factor.