dorkitude

Klout’s Score Decay Problem (Do I really have more Klout than Hipster Hacker?)

Klout is a service that tries to distill your online influence (mostly your Twitter influence) down to a number.

But I believe their scoring algorithm is profoundly broken, because it puts an undue emphasis on recency. With Klout, it’s not “what have you done?”, but “what have you done lately?” I believe recency should be weighed much lower than they weigh it today — it might even be a counter-signal!

To demonstrate the problem, let’s look at two accounts with similar audience profiles (software developers and startup geeks). One is my personal account (@dorkitude), and the other is a popular comedy account, Hipster Hacker (@hipsterhacker), which is also geared toward programmers.

I tweet pretty sporadically, but I haven’t been as active as usual over the last three to six months (although I’ve gotten better in December). My Klout score was as high as 62 when I attended a party at their headquarters ~7 months ago.

Hipster Hacker tweets very sparingly, and he has been particularly inactive over the last three months, posting only three total tweets during that period. The highest I’ve ever seen his Klout score was 69, also ~7 months ago.

At the time of this writing, @dorkitude has:

• 1,453 followers
• 8,814 tweets
• 50 Klout

And @hipsterhacker has:

• 9,912 followers
• 89 tweets
• 42 Klout

Both Klout scores have decayed (and Hipster Hacker’s far more than mine), but it should be obvious whose tweets actually have more influence.

Even if we ignore the 7x advantage in sheer follower count, @hipsterhacker has a ridiculously high follower-to-tweet ratio of 111.4 to 1, while @dorkitude’s is a paltry 0.165 to 1. His most retweeted single tweet was retweeted 1172 times and number two 768 times. Meanwhile all of my tweets combined have been retweeted around 1200 times, according to my Klout profile.

If anything, Hipster Hacker’s tweets have garnered more attention on average as they have grown more scarce, precisely because they have grown more scarce: people aren’t used to seeing his name jamming their feeds, so they perk up when they see him writing again.

I’ll close with two questions:

1. Has anyone else seen this pattern on Klout? Counterexamples?

2. What does Klout’s business model look like? Is there a core business incentive to make the algorithm as good as possible, or just good enough?