We social know-it-alls love to poke at score-keeper Klout, the useful idiot of the social world. Klout is is the master of social-media Let’s Pretend (that a simplistic score can accurately capture influence), a fat target, and online snark (like this article) sells. (It sells for the Let’s Pretend currency that most content sites charge and pay, i.e., nada.)
Still, despite its weaknesses, we need Klout and its measurement kin Kred and PeerIndex. Couldn’t the company just make it better? I think so, despite questions about the foundations of the whole of the online-influence business. Here’s my Klout critique, concluding with one point that especially irks me because it’s highly visible and surely easy to fix and another that, from my perspective of a semantics-analytics booster, I see as worthwhile and doable. In the spirit of hoping I can influence (um) Klout’s development priorities, let’s start with the critique and then move to those fixable and fix-worthy points —
The premise behind Klout, that influence can usefully be measured, is sound but Klout’s implementation, a couple of years in, is weak.
Do you buy that social-platform popularity, reduced to a simplistic score, is more than a weak indicator of any real form of influence? That a single score accurately captures someone’s influence on disparate topics? On the input side, despite recent algorithm revisions, Klout does not account for non-social mentions. You were profiled in the New York Times? Feh! Further, Klout doesn’t account for off-line presence. My score is 6 higher than former President George W. Bush’s, which just isn’t correct, regardless what you think of 43’s politics. Further, you can game the system. I know my friend Banafsheh appreciated the +K I recently gave her on “Washington Wizards” and “parenting.” Klout had already deemed her influential on those topics, which don’t exactly fit her real-world profile. All I did was feed a beast not of my own creation. Was I wrong?
We social know-it-alls are vain creatures. We like to know how we stack up. I rarely google myself anymore, but I do watch my Klout score. I want to understand how my on-social activities affect the most and widely accepted, public influence measure. I use Klout as a tool for SEO mark II, for social-engine optimization.
Fixable and Fix-worthy
Let’s be constructive. Two items fall in the “fixable and fix-worthy” category. Call them comparisons and classifications.
Klout facilitates comparisons, and those comparisons expose a stupid but fixable interface flaw. Check it out in the image to the right. You have to log into Klout to explore your scores, and if you do, you’ll see that Klout automatically compares you to yourself. Klout, myself, and I: A bit too solipsistic, even for me.
Klout: Surely this silly UI flaw is easy to fix. Don’t compare someone to him/herself.
Next, classifications. Klout already identifies your particular influence topics. Presumably this is done via some form of content analysis. I post a whole bunch about Technology, Market Research, Analytics, and Business Intelligence (says Klout), and that’s not too far off, even if I would prefer narrower, more precise categories such as sentiment analysis and text analytics. Those topic suggestions may be voted up by others, via the +Ks I mentioned above. It couldn’t be that hard for Klout to compute break-out scores by influence category, scores that would, for instance, tell you that Meg Whitman (score 74) is influential in enterprise IT but isn’t someone folks turn to for political leadership.
PeerIndex does it, provides a break-out of influence by high-level topics. PeerIndex does not, however, allow you to understand which topics that link you to particular people you influence or who influence you, and of course, Klout doesn’t either.
Global and Detailed
So what we really want is an influence measure that’s both global and detailed, incorporating non-social media and off-line contribution and broken out by topics and further, by direction of influence, geographic sphere of influence, and audience influenced. Justin Bieber has over 27 million Twitter followers, but I’m not one of them, and his Klout score might as well be 3 so far I or @supportmusic_ID are concerned. Certain tweeters would have a negative Klout score for me; what I mean by “direction of influence”: I’d reflexively think the opposite of whatever they post.
I’ll close with a couple of pointers, one to Ventana Research CEO Mark Smith’s recent The Stupidity of KPIs in Business Analytics, which covers topics very similar to mine in this article — key performance indicators are to BI what influence, sentiment, and engagement scores are to social-media analytics — and another to my up-coming Sentiment Analysis Symposium, October 30 in San Francisco. A thematic point of this fifth conference go-around is the contribution of sentiment to larger signals, derived from consideration of behaviors (e.g., clickstreams), transactions (recorded in back-end databases), profiles, and analysis of movement and speech. Just as you shouldn’t compute influence solely from social-media measures, sentiment analysis is rightly seen as part of an analytics big picture. Me, Myself, and I make sense only in context, for Klout and for the larger analytics enterprise.