When I first met Joe Fernandez (@JoeFernandez), CEO and Cofounder of Klout, I thought little more of Klout than a new Twitter score. And to be honest, that is what they were for some time. A few minutes with Joe and you gain a lot of confidence in what they are trying to accomplish.
To reach their goal of becoming “the standard for influence”, they had to start somewhere and Twitter, the realtime back channel of the internet, was an obvious choice. Now they have added Facebook and have plans to add LinkedIn, Foursquare, and maybe Myspace very soon. Additionally, several other online metrics and social networks are on their radar to improve the scope of Klout.
They are trying to build an algorithm (or more accurately a set of algorithms) that will measure ability of someone to drive people to action online. In an attempt to provide insight, let’s break down how Klout produces their main influence score.
Currently Klout measures 30+ variables on Facebook and Twitter to give you a 1 to 100 score “with higher scores representing a wider and stronger sphere of influence”. Having only a Twitter account or Facebook account will not negatively affect your score, but will shift the weight of your score to that network’s variables only.
What are these variables? Well they fall into 3 separate categories.
1. Making People Act
They are pretty much anything that can be determined as an “action” in to response to something you have done. Klout calles this “Amplification Probability” and describes it as the likelihood that your content will be acted upon.- Unique Retweeters
- Unique Messages Retweeted
- Likes Per Post
- Comments Per Post
- Follower Retweet %
- Unique @ Senders
- Follower Mention %
- Inbound Messages Per Outbound Message
- Update Count
2. Network Size
Next, Klout looks at the size of your network. Think of this as the potential size of the audience that may be exposed to your audience. Consider several circles here. First are your direct followers, friends, people following your lists or lists your on, etc. Then consider retweets and replies and comments as they may expand your potential audience greatly. Then consider the meta data of this potential audience. How many are you following? And how many are unique interactions, compared to one person interacting multiple times. Klout refers to this as “True Reach” and describes it as the size of your engaged audience. The official list from their website of variables that control this score are:- Followers
- Mutual Follows
- Friends
- Total Retweets
- Unique Commenters
- Unique Likers
- Follower/Follow Ratio
- Followed Back %
- @ Mention Count
- List Count
- List Followers Count
3. Network Quality
Lastly, Klout looks deeper than just the size of your potential audience and more closely at who makes up that audience. Basically answering the question, how influential are the people you are engaging? They call this “Network Influence” and describe it simply as the influence of your engaged audience. Variables measured here include:- List inclusions
- Follower/Follow Ratio
- Followed Back %
- Unique Senders
- Unique Retweeters
- Unique Commenters
- Unique Likers
- Influence of Followers
- Influence of Retweeters and Mentioners
- Influence of Friends
- Influence of Likers and Commenters
The Secret Sauce
With all these variables to consider, you might start to get a sense of how complicated a task it is to try and measure online influence. Above all the variables and data points is another secret sauce. Klout then has to attach a weight to each variable. This is the part that is not public. How much a retweet affects a score verses a Facebook comment and so on. If you are a doubter, this is likely where a lot of your distrust might creep in. Who is Klout to determine how much a retweet or a new Facebook friend should count toward your influence score? “One thing we really want to avoid is trying to tell people how to use Twitter or Facebook” said Fernandez. When you talk to Joe about the algorithm, it is clear that he is all about the data. About taking clear measurable results and weighing them against other results.Testing For Accuracy
Another item they consider a sample of is click through rates.We perform significant testing to ensure that the average click-through rate on links shared is highly correlated with a person’s Klout Score. (http://klout.com/kscore)To do this they access publicly available click through data via Bit.ly. In fact they do a lot of this “testing” at Klout. At one point, of 22 employees at Klout (likely more now), 18 of their staff were engineers. That is a huge emphasis on the science behind their algorithm and how seriously they take the responsibility measuring influence. When you talk to Fernandez he will emphasize that the engineers working on the algorithm hunt for outliers a lot. They try to find people that have abnormally high or low scores and see if their algorithm has a loophole that are missing these accounts and correct it accordingly.