I was recently chatting with my friend Matthew Brown of AudienceWise about the distribution of the web’s traffic, and we both wondered – do referrals from external domains follow a “long tail” distribution pattern?
I surmised that only ~20% of the referrals that the average website receives comes from the tail of the distribution curve, whilst Matt felt that number should be considerably larger. As a first step, I figured I’d check SEOmoz.org’s own referral traffic (via Google Analytics) to see what our distribution curve looks like. It’s visualized below:
Whoa. Less than 10% of the referring domains to SEOmoz send more than 79% of the referring visits. That’s a shocking distribution.
But, it could be that we’re an outlier, so I looked at the traffic distribution for a few other sites.
Here’s the distribution for Inbound.org:
And for this blog, Moz.com/Rand:
It would seem that the distribution is strongly biased in every case (at least in our tiny sample set) to the fat head of the referral demand curve. Yet, the web is truly massive. At last count, Moz’s big data engineering reported in excess of 80 million domains we discovered in our web crawl that appeared to be live and functional (out of ~200mm total domains seen).
This data’s only served to pique my curiosity, and now I’m hungry for more. Is the web’s traffic distribution from referral sources really this biased? How much of the web’s traffic is collectively driven by the top thousand, hundred thousand, and million domains? Are the top traffic referrers very similar across industries, segmented by niche, or relatively unique from one site to the next?
I would love to see someone with time and access to more data dig into this project. If you do, I promise to lend all my sharing power to distributing your work and giving credit. Here’s to hoping someone takes on the challenge!
p.s. I removed referrals from search, but opted to include social referrals, so Facebook, Twitter, Google+, LinkedIn, etc. appeared in the top 20 for all of these sites.