Interesting discussion (I wrote the reputation algorithm for http://letslunch.com). My main concern with computing online reputation is that you miss the 95% of the people who don't see Twitter as a goal. Is a Senior Vice-President at Apple influential? Klout would say no (no tweets). LetsLunch would say yes :-)
What data does LetsLunch use that allows it to score 95% of people?
On Twitter, if you are the Senior VP at Apple and---hypothetically---you only have 50 followers but they are all movers and shakers, then Twitter analysis should be able to determine that you are influential.
I guess your concern is that there might be many influential people who aren't even on Twitter, or don't have 50 influential followers on Twitter. What is your proposed solution to the problem? Your "How It Works" page (http://letslunch.com/site/page?view=how-it-works) says that you pull signal from LinkedIn and, optionally, Twitter and Hacker News. You arrange lunch dates, collect post-lunch feedback, and then tune people's reputation based upon the feedback.
What about the 99.99% of people who don't use LetsLunch? And why is a single lunch date enough to be able to determine someone's influence?
Yes, my concern is that a lot of influential people just aren't on Twitter. I would almost argue that the really influential people have better things to do than tweet. And I agree they are not on LetsLunch either, that wasn't my point :-)
I gave a fairly specific example. I think it's pretty self-explanatory.
