2011-06-28

Understanding Klout Topics

The most widely known social media measurement tool right now in the market is most likely Klout. They have gained notoriety through their API which have allow hotels to look up your Klout score at check-in (and then providing special treatment to those with high scores), and other applications to help people evaluate other people's influence in the land of social media.

Yesterday I had the opportunity to visit Klout's head office in San Francisco and to sit down with Megan Berry (@meganberry) their Marketing Manager. We had a great 30 minute chat, where we covered a variety of topics from Klout Perks to scoring, to Klout Topics. Klout Topics have dominated my Twitter stream for months now with many people questioning how these topics are chosen by Klout especially when in some cases they seem completely out of character or have nothing to do with what these people tweet about.
People hard at work at Klout


Here is a summary of what I learned and it now is starting to make sense.

1. Klout first looks for specific keywords/themes in your tweets that generated lots of engagement. This can be based on lots of replies to your tweet or retweets of your tweet;
2. Secondly, if you replied to someone's tweet and your response generated lots of engagement then they will look back to the original tweet for keywords/themes;
3. Once they have the keywords/themes where you yield influence, they use semantec analysis to identify   standardized and relevant terms
4. Klout then compares your influence on these standardized terms to see if you are yielding significant influence within your circle and within their user base
5. If you are deemed to yield influence on a specific term it will appear in your list.

Take these two examples:

On person tweets about personal "privacy" 20 times a day. If no-one ever replies to their tweet or retweets it, the term "privaciy" will not show up on their list.

Another person publishes a single tweet on pesonal "privacy", it generates 20 replies and is retweeted 30 times plus many retweets of the retweets. In this case a single tweet has generated influence and it will appear in their list (assuming that they are not more influential on another 10 other topics - the maximum displayed).

This can cause problems when the use of the Symantec dictionary presents terms that don't make senses to you. For example, in my case for more than the past month Klout has told me (http://klout.com/#/aknecht/topics) that I'm influential on the term "tools". I rarely use the word "tool" or "tools" in my tweets and I took it to mean items like hammers. Megan explained to me that I tweet a lot about  analytics software, utilities for measuring social media, etc. In general I tweet a lot about "social media tools" and "analytic measurement tools" so Klout has simplified these items down to just the word "tools".

I have theorized that some of the problems people are having occurs if they click the reply button on a tweet and then start a different conversation with the person not realizing that while in their minds this is a new tweet,  yet in Klout's eyes it is still connected to the theme and keywords of the original tweet that started the conversation.

So that is Klout Topics in a nutshell. Do you think they are on the right track or they off-base?