If you haven't seen Klout, take a look. Here's how it works: you give Klout access to Twitter, Facebook, LinkedIn via APIs, Klout does some magic analytics on your social media streams and then gives you a score. The score is supposed to encapsulate how influential you are on a given topic. My Klout score is a modest 37.

Ok, great. I love this effort to cut through the cruft and identify who matters with respect to a given topic. I guess I'm just having trouble answering the question: matters for what? Recently, some media outlets (e.g. The Guardian) have made a big deal of Klout's claim that Justin Bieber is more influential than Barak Obama. (OMG!) Klout CEO Joe Hernandez responds on the Klout Blog. Joe says:

At Klout we measure influence across the social web. The point isn’t that Justin Bieber is more influential in the world than Obama, but that he is using social media more effectively to drive more actions from his network than anyone else right now… In the last 24 hours both Obama and Bieber have tweeted links to Youtube videos. The video shared by Bieber has generated nearly 10x the number of views the Obama video has.

Ok, here's my axe to grind, and here is why measuring influence with algorithms, without much context (except a vague topic area like "iPhone"), and without real social connections and interactions is a problem. Hernandez thinks Bieber is more influential because he gets teens to tweet more or watch YouTube videos more. But convincing a teen to Tweet is like convincing a pothead to eat Cheetos: EASY. Likewise, if I run into you in the parking lot of 7-11 and suggest that you go in, you wouldn't call me very influential, even though you'll have a Slurpie in your hand in 90 seconds. Noticing a statistical correlation between these things will get you a "Master of the Obvious" badge.

I don't make any claims about Klout's algorithms, or about the statistics of influence. But I do know that an influential person is someone who persuades you to do something you would not otherwise have done. So any serious discussion of my influence has to compare the likelihood of your doing something in the presence of my influence and in its absence.

It does not appear that Klout is doing this, which leaves me confused about what they are really measuring. (In fairness, maybe they are doing this, but this is not how they're pitching it…) I have a feeling the people they identify are more like the people at the top of information cascades. I guess that's a kind of influence – you get people all lined up in social network channels according to their interests, and then try to figure out where the cascades usually begin. This sort of thing is very important to marketers and brands, because it helps them know who to go to if they want to get something started. Likewise finding these people can be useful to you and me if we want to cut to the chase and hear it from the horse's mouth. But this is really not very much like offline influence.