November 17, 2008

Twitterank's Leaderboard: Odd, Mysterious and Broken

The launch of a leaderboard for the once-feared Twitterank was inevitable. After all, in the online world, if you can measure something and give it a score, then by all means the next step is to rank people from high to low, and provide a leaderboard. It's happened with blog "influence" (Technorati). It's happened with mentions on Techmeme. It's even happened with how frequently people's items are shared on Google Reader (Feedheads, RSSmeme and ReadBurner). As ranking one's Twitter influence has been tried several times by a bunch of different sites, from Twinfluence to Twitter Grader, Twitterank was practically destined to join the crowd. On Friday, the site launched a "Top 50" list and after watching the dust settle a bit, I have to be extremely amused by the results.

Every ranking system has its flaws. And considering Twitterank's algorithm is both secret and changing, according to its author, Ryo Chijiiwa, initial hiccups are no surprise. But glancing at the top 50 tells me that Twitterank must measure influence in a very odd way, contrary to just about every other measure I've seen out there.

For example, according to Twitterank, the #1, highest scoring person in all the world is Scott Beale of Laughing Squid. (@laughingsquid) Scott's account garners a score of 237.591. His own Twitter account shows he (as of Monday after midnight) is following 1,636 people, has 19,307 followers, and has made 5,285 updates. This does not rank him among the top 50 on Twinfluence in total reach, but he does reach #20 on Twitter Grader.

In the #2 position on Twitterank is Brian Solis (@briansolis), who weighs in with a score of 235.847, and Twitter activity of 582 following, 8,033 followers and 3,524 updates. This activity garners him the #43 position on Twinfluence and #22 overall on Twitter Grader.

While Twitterater's top list does have a lot of "household names" like Dave Winer, Michael Arrington, Jeremiah Owyang and Steve Rubel, there are some big oddities, including at least one account that has never sent a message on Twitter at all.

Let's be honest, there's no way I should be this high.

For example, Loic Lemeur (not pictured, but at 226.91) actually ranks below me in the rankings, despite his following and being followed by almost five times as many people, and sending ten times the amount of updates. Meanwhile, Leo Laporte gets a 179.87 ranking, well off the top 50 list, despite having more than 60,000 followers, behind only president-elect Barack Obama and Kevin Rose of Digg (that I know of). And the ever-present Robert Scoble gets only a 188.63, also keeping him off the Top 50.

Leo Laporte, with 60,000 followers, misses the leaderboard?

And Scoble, Mr. Twitter, doesn't break 200 either?

So how does that make any sense? I was going to guess that Scott Beale ranked highly thanks to his high followers to following ratio, but Leo Laporte's ratio is an astonishing 120 to 1, so that, in theory would rank higher. And Scoble's real numbers are off the charts in almost every metric.

Another canary in the coal mine - the account of @google, which ranks #13 overall, according to Twitterank's Top 50, but has only 366 followers, isn't following anyone and has never updated their Twitter account.

So... @google, a user with no updates, has a higher Twitterank than does Scoble, who tops out at 39,000 followers, and more than 15,000 updates. Whatever you think about the content of Robert's tweets, whether they be too frequent or too off-topic, to say that an unused account is among the top in the world is as they say in the Web world... a big FAIL.

That Twitterank has an algorithm which measures something is clear as it gets some of the names you'd expect, but there are still a lot of questions around this service. Right now, it's basically a toy, and has little value.