But when I added my personal Facebook account, with about 470 friends, guess what happened? My company's score dropped. And not by a little, but by almost 25%, down to 45. Even more bizarre: It stayed at 45 even after I disconnected my Facebook account from Klout. And that's when I saw the wheels fly off the Klout numeric bus. Think about it. Never mind what mathematical gerrymandering is required to have a cumulative calculation collapse by 25% when I add 5% more users; how can my score not go back up to its original level when I unlink my Facebook account? When I contacted the company, a very nice representative, Lynn Fox, told me, "What you described sounded odd, because adding an account typically does not lead to a score drop ... It appears that there was a Klout glitch." And she said they would restore my account to its previous level. But when I followed up with the deeper questions, like what kind of a model can make such a dubious numeric jump, Fox declined to respond to two emails. So I checked in with experts who manage the type of data Klout deals with and they agreed odd results can be expected. "I am not surprised that there are flaws in this algo," Michael Driscoll explained to me over the phone. Driscoll is the co-founder of Metamarkets, a San Francisco-based big data management and analysis firm. He also is co-chairman of the Bay Area users group for R, a cutting-edge stats computer language. Driscoll, who was clear that he has no direct knowledge of Klout but felt comfortable speaking as an expert, said there are very real practical limits such companies will face. This operation has to pull in data streams from dozens of social media platforms and somehow render all that into a meaningful two-digit number (three if you're Bieber, one if you're dead) -- a monstrously complex job for any company, even one with the backing of Microsoft. It's not just my experience that backs up Driscoll's skepticism. Look around.