3 Reasons Klout’s Algorithm Is Bogus

Labeling itself “The Standard for Influence,” Klout has worked to become a relevant influence-measurement tool for businesses and individuals.

While every marketer would love a way to quantify social media influence, is there really any substance behind Klout’s numbers? Does a high (or low) Klout score really have an impact on whether you can influence others’ behavior — or is Klout simply an imprecise  measurement of one’s social media prowess? Here are three reasons Klout’s algorithm is probably more alchemy than science.

Measuring the Wrong Things


One of the main problems of Klout is that it relies on third-party APIs (Application Programming Interfaces) to get data from the networks it scans — and those APIs have limits. A good example of where this falls short is that Klout has no way to monitor click-through rates. The service has no way of knowing what volume of traffic flows through the items a member posts, or if these posts actually lead to any action offline. How can a service claim to adequately monitor influence with such a huge blind spot?

Another issue is what Klout actually tracks and how it tracks it. With Twitter, scores are determined by monitoring things like follower counts, mentions, and retweets. However, it doesn’t give credit very well for those using Twitter’s native retweet system. When a member retweets something, Klout gives credit back to the original account, even if another user’s retweet exposed it to a larger audience.

Some Twitter members don’t like this and have devised ways to game Klout’s handling of this metric. These individuals choose to use the older manual “RT @name” style retweets instead of the native Twitter system. This way, when the post is retweeted by their audiences, they get the score boost and increased visibility instead of the original user. This is essentially a way of stealing influence and Klout’s algorithm (as it is today) encourages it.

An Ever-Shifting House of Cards

In February of last year, an infographic made the rounds on the web highlighting a Justin Bieber Twitter spam account with an astoundingly high Klout score. This illustrated some major flaws with Klout and served to embarrass the company briefly.

To combat the loss of confidence in its service, Klout has attempted to iterate and improve its scoring dramatically over the past year. However, it has often done this without notice or explanation. It was not uncommon for members to log in and find that their scores had plummeted by tens of points without explanation. These changes, while aimed at improving the service, essentially stomped all over Klout’s credibility and gave the company a boy-who-cried-wolf reputation with each new algorithm upgrade.

Perhaps most damning is the way Klout has de-emphasized and removed certain metrics in its latest redesign. For example, the latest revision has removed things like user classifications and score analysis. The company has even hidden the ability to track changes to certain metrics over time. If Klout felt more confident in its algorithm, it would expose more of this data, not less.

Topical Misunderstanding

Klout’s system of topics are, in many cases, baffling. For example, having one tweet including the words “Paparazzi” and “Matt Damon” retweeted by someone Klout deems of greater influence can get you labeled as influential on those topics. This can happen despite that tweet being the only time you ever mention those topics. It doesn’t matter if you have many other tweets pertaining to broader topics, such as politics or technology, that achieve a broader reach.

It appears that Klout matches its “topics” to keywords used in social media posts alone, instead of using actual, objective, topics. It also appears that Klout’s algorithm is not smart enough to understand and classify posts based on context — without the usage of such keywords directly. These are some of the flaws of automation. Algorithms cannot understand things like sarcasm and tone, and they have an especially hard time with nuances of language and meaning that only humans would pick up.

At the end of the day, Klout is trying to quantify something as inherently subjective as influence. But while some metrics can be helpful to understand audience reach, influence cannot be distilled into single number. While Klout’s promise is tantalizing for marketers, it will always need to rely on a certain amount of assumption and fabrication to seem legit, thus leading its value to be questionable at best, and bogus at worst.

[Image credit: Peter Roberts]

An Event Apart Chicago 2012 in Impossible Images

Last week Jeffrey Zeldman and Eric Meyer brought their excellent design conference, An Event Apart, back to Chicago for the first time since 2009. As in previous years I had the opportunity to play the role of photographer for the event. However, this year I decided to mix it up a bit and also take Impossible images to complement the regular digital photos and coincide with my ongoing Impossible Year project.

Without further ado, An Event Apart Chicago 2012 in Impossible images:

Steve Jobs on Speech Technology as Transcribed by an iPhone 4S

In 2003 Steve Jobs sat down with Walt Mossberg at the first ever D: All Things Digital conference. The end of the session included a rare open Q&A with Jobs by audience members in which the following question was asked:

“You talk about handwriting and keyboard, how about speech as a human interface?”

Here is Steve’s answer transcribed by an iPhone 4S using iOS 5′s Dictation feature:

“You know that the last part of that question is is exactly right I I I’ve been almost 30 years and it’s speech is always been five years away because I’m trying to completions
But it went most of the really smart people of speech I know have gotten out of Beelitz it’s like nuclear fusion a lot of bright people went into it and and and people now think it’s a ways away you can do what you can do adequate speech today to meet your correcting a lot of stuff if you just have to know even 1% area try cannot as it turns out you can speak a lot of words in a few terrible so it’s got to be very accurate and so far no one’s come up with the technology Apple’s got a very we got to speak working on stuff Microsoft has a group you want GreatWorks be done in academia but it doesn’t look like it’s been a be real anytime soon I wish it was different”

Here is what he actually said:

“Ya know, the last part of that question is exactly right. I’ve been in this industry almost 30 years and speech has always been five years away. It’s been constant time to completion, just moving along 5 years away. Most of the really smart people in speech I know have gotten out of the field. It’s like nuclear fusion a lot of bright people went into it and people now think it’s a ways away. You can do you can do adequate speech today but it means you’re correcting a lot of stuff. If you just have even 1% error it drives you nuts. Because it turns out you can speak a lot of words in a five minute interval. So, it’s gotta be very accurate and so far no ones come up with the technology. Apple’s got a speech group working on stuff. Microsoft has a group. A lot of great work is being done in academia but it doesn’t look like it’s going to be real any time soon. I wish it was different.”

Looks like Steve was right.

ILD Bike MS: Tour De Farms 2012

I should have posted this much sooner but I’ve been busy. Perhaps too busy, but it’s never too late:

This weekend I will be riding, with my good friend Laura, in the ILD Bike MS: Tour De Farms 2012 in an effort to raise money to fight MS.

This is a cause that is very important to me. Growing up I saw MS devastate and eventually take the life of my “Aunt” Kathy , my mother’s best friend.

I’ve set a small goal of $400 but I’d like to blow it out of the water. If you’re interested in helping out you can donate here. It’s tax deductible and goes to a very good cause.

Thanks.