Reviewing Rand: Correlation, Causation, and SEO

If SEO is part of your life and you missed Rand Fishkin’s session at SMX Advanced a few weeks ago, you now have their must read Updated SEO Policies Based on Correlation Data (where Rand says the comments section is now also filled with important info) and a video where Rand essentially interviews one of the numbers guys behind the study, Ben Hendrickson (whose bio page is worth skimming if you’re into geeky deadpan humor).

The real story is the study. AFAIU SEOMoz has the most link data of any site besides the search engines themselves. And they’ve been working hard to see which factors correlate best with SEO rankings. The post is well written.

Some thoughts on the video:

  • Funny how just showing their numbers guy made me believe their numbers more. Like those old commercials with the actor in a doctor’s coat, except for real. Believe the messenger, you’ll believe the message.
  • I love watching people who are passionate about numbers and statistics. Too many people (generally the people that aren’t good at stats) think that you can either have a heart and soul or understand (and even enjoy) stats.
  • The best way to tell a real stats guy from a fraud? The real thing can’t help repeating the limitations of the numbers. Correlation doesn’t prove causation. They use cautious words like “presumably” that drive many salespeople batty, but one of the marks of real statisticians and scientists is how they acknowledge and even embrace the limits of their knowledge.

Ben was funniest when he said “Google doesn’t actually use MOZRank (SEOMoz’s formulaic approximation of Google’s Page Rank). If we changed the MOZRank for SEOMoz, which we can do since we produce the number, it won’t actually help our site show up higher in Google.”

Sorry to pick on Ben, but I also laughed when he referred to antidotal evidence, a nice new Bushism or verbal typo that I guess means a piece of evidence that fixes a poisonous problem. I think he meant anecdotal.

Anyway, great study, good video.