Link love is a rotten proxy for attention

17 February 2006

The second part of Dave Sifry's State of the Blogosphere, for the start of 2006, contains an interesting but, to my mind, flawed assumption about what Technorati is able to measure. Sifry's analysis looks at "how attention has been shifting in the blogosphere". He then uses measures of link love to demonstrate that shift, but only to demonstrate that, even using blog-friendly metrics traditional media has captured much of the "attention".

The New York Times, CNN and the Washington Post are out in front still, according to his metrics. But the blog Boing Boing has overtaken BusinessWeek and Forbes among others, and is beginning to trouble the Guardian. Both Boing Boing and Engadget have apparently trumped Slashdot. The metric that Sifry is using is based purely on the number of unique links made to each site from blogs that Technorati tracks. In fact, looking at the data, it's not even clear that, by his definition, the attention is shifting away from the dreaded MSM. There are fewer blogs in the top 30 from January than from August 2005 or March 2005. January's graph is a sea of blue mainstream media sites with just four red blog bars in the list.

The problem with the analysis is that, using inbound links alone does not in any way capture attention. If you compare the Techorati results with the Alexa traffic stats, you come up with a completely different list of top sites. I know there are issues with any third-party stats, but it quickly becomes clear that Slashdot merrily trounces Boing Boing for actual attention - that is, people actually looking at the site's pages. I've seen plenty of anecdotal evidence recently that slashdotting is still a force to be reckoned with. Boing Boing traffic coming your way is nice, but nothing still melts servers like a popular post on the site no-one in the world of blogging wants to call a blog - even though it shares a lot of DNA. The BBC ranks much higher under Alexa than on Techorati's list, easily surpassing the New York Times for traffic.

What is clear is that the Technorati stats show something but not "a shift in attention". They simply show the sites that people are most keen to discuss online using one particular website format: that does not mean attention from a worldwide population but what remains a largely North American community. The New York Times and the Washington Post clearly collect a lot of links from political bloggers of both persuasions, alternating with praise and opprobrium. They effectively collect a two-for-one every time and benefit from early growth in North America.

Sifry has decided that the place we need to look at is in the Magic Middle, where vertical blogs can compete happily with traditional media. That makes a degree of sense, as blogs tend to have a narrow focus. But, again, using links to measure attention is going to get you nowhere. Concentrating on links makes blogs focus inwards rather than on communicating with an audience. People write stuff for what is likely to pick up a link from what remains a small proportion of the online population. The numbers that Technorati touts sound big, but any analysis of blog traffic stats reveals how much larger the potential audience for a blog really is.

Sifry's posting followed last week's about the apparent rampant growth of blogging around the world, which caused some to question whether blogging had indeed hit its peak already.

Even after taking out all the splogs it could find, the company came up with a total of 27 million blogs, up from 20 million blogs in the autumn. That 20 million from last autumn contained a number of obvious splogs, which prompted Matt Galloway to perform an analysis based on predictions made by Umbria. Because splogs looked to be multiplying faster than blogs written by people, Galloway wondered whether the real blog count was going to be on the way down real soon. Sifry stepped in to claim that the 27 million was indeed free of splogs (at least as far as Technorati could determine). Now, that indicates that blog growth over the last six months was actually faster than Sifry claimed, as the numbers from the autumn have not been adjusted - assuming that Umbria was right and they needed to be adjusted.

The curve has got to turn S-shaped at some point, although that point will be delayed by the effects of churn - people moving from one blogger site to another - and the bulking effect of dormant blogs. Whatever the trajectory of the curve, the blogging population still some way behind the online population of some 1 billion people. The open question is what is the proportion of people who are willing to launch and maintain a blog? If that number were to move much nearer to the online total, then link love might indeed become a good proxy for attention. But, right now, it is going to give very skewed results.


The US obsession with measuring online influence by measuring links, and the quality of the links, misses a huge hole: the influence that is not measured by links but by readership :-)

And what about the secondary and tertiary discussions that are created that no longer link back to the original postings? Technorati is one of the biggest sellers of the blogosphere, it purports to have the tools to let corporate America know who their friends and enemies are, where they blog and who is in their circle. Technorati is one of several Big Brothers spying on the blogosphere and selling that information to anyone with a large check. I know because I had the unfortunate experience to be on a panel with one of Technorati's senior people and all he did was pitch, pitch, pitch.

Good old readership: I was wondering whether that got tipped out with the rest of the Web 1.0 bathwater. Is this possibly where bloggers can take back what is their own? Rather than have Technorati make up some cod metrics for them and making do with Alexa, why not just publish potted webstats? Tim Bray sticks his up every Sunday as I recall and I think Boing Boing has been doing it. If there was a common way of reporting them, it might be a way of demonstrating that there is more to it than link love.

It would certainly help remove Technorati and others' claim to monitor the blogosphere accurately if it could be shown that high traffic, but low trackback posts were popping up (and having a greater 'influence' overall).

Comment and link ratios might form part of an expanded set of metrics. I'd be interested to see how big the deviations are between blogs in terms of traffic and comment ratios for one. It might also help the advertising situation for those who want to take ads as a common format would help remove suspicion about self-provided stats (although that might not help with deliberate traffic boosting using bots etc).