Correlation, Causation and SEO



Much speculation abounds when it comes to reverse engineering which criteria are prioritized in search engine ranking algorithms. In the SEOmoz ranking factors summary from this year, the correlation of higher page ranks to greater Facebook "like" numbers was noted.

As my team lead (co-author of Audience, Relevance and Search - here's its companion blog) and I discussed, it was clear that there was an implication that people were likely to interpret this correlation as causation - that is, cum hoc ergo propter hoc. It's a very easy trap to fall into: thinking that we can rely on a quantifiable factor such as SNS-driven endorsement counts to predict how high up a SERP the page would be likely to be found.

However, if one examines the use case scenario of a "popular" page, here's what I would easily envision happening:

Something that is useful, very entertaining or plain memorable is published. News of its existence begins to virally spread - this used to be mostly via email, but now it's typically via SNS. The backlinks also increase in number from pages with related content, as people wish to share it for various reasons: this used to be copy/pastes in newsgroups. Nowadays, most web pages use designs that will track comments, tweets, diggs and Facebook likes: some fraction of visitors will leave their approval of such content on the page.

So, since SNS metrics are proof of high traffic to a page, it seems likely that pages with many Facebook "like"s would have higher backlink counts and endorsements of the quality of content it provides. It would be those other factors, rather than the "like", tweet or similar sorts of numbers, that are probably resulting in elevated search engine rankings.

This is why using giveaway contests and other artificial means to raise SNS-driven endorsements aren't likely to bump up your SERP ranking.

Comments

All time popular posts

Is larger (PPC) better? Size matters, but... the #G+ strategy

「ほとんど日本人と異ならないですね」

Google+ increasing its reach