Understanding the Klout Score Part I
November 18th, 2011 by Joe FernandezThis post begins a new “Understanding your Klout Score” series. Today, we review our recent changes and in the future we’ll dive more into your Score, networks, and tips on how to improve your Score.
It’s been an interesting few weeks here Klout. Now that we are finally catching our breath, I think this is a good opportunity to look back at what we’ve learned and discuss the changes we’ve made.
The way influence is signaled online is constantly changing. New networks are born and new behaviors emerge overnight. The Klout Score will continue to evolve to support this change. The Klout Score and Topics will always exist in a dynamic state of improvement.
We will be more transparent
Our biggest priority with the new scoring model was to increase transparency. We added some insights to show why your Score changed but this isn’t nearly enough. Given the passion our users have for their Klout scores, it is clear that we need to do more to help them understand why the Score has changed and what that Score means. To accomplish this, we are focusing nearly all of our efforts on projects that relate directly to the transparency of the Klout Score. The team is really fired up to share the tremendous amount of data and thought that goes into creating the Klout Score.
Our biggest priority with the new scoring model was to increase transparency. We added some insights to show why your Score changed but this isn’t nearly enough. Given the passion our users have for their Klout scores, it is clear that we need to do more to help them understand why the Score has changed and what that Score means. To accomplish this, we are focusing nearly all of our efforts on projects that relate directly to the transparency of the Klout Score. The team is really fired up to share the tremendous amount of data and thought that goes into creating the Klout Score.In the spirit of greater transparency, here are some in-depth examples to illustrate the three, primary improvements we made to the Klout algorithm on October 26th.
Greater equality of networks
In our previous Scoring model, the main driver of your Klout was a primary network (the one you’re best on) and, to be honest, your influence on secondary networks was too small a part of your Score. Now, a user who has two networks that are fairly equal in terms of participation and influence will see a greater parity in the way we score those two. Certainly, there may be more potential to be influential on a network with many millions of users like Twitter or YouTube, but we measure that influence equally wherever it occurs. That said, there is no score reward for just adding networks that you do not participate in.
In our previous Scoring model, the main driver of your Klout was a primary network (the one you’re best on) and, to be honest, your influence on secondary networks was too small a part of your Score. Now, a user who has two networks that are fairly equal in terms of participation and influence will see a greater parity in the way we score those two. Certainly, there may be more potential to be influential on a network with many millions of users like Twitter or YouTube, but we measure that influence equally wherever it occurs. That said, there is no score reward for just adding networks that you do not participate in.
Example: Consider two people who influence the same 100 people to the same extent. One person influences their network exclusively on Twitter. The other person influences two audiences of 50 equally on Twitter and Facebook. In practice, they have the same level of influence, and now they will have the same Klout Score as well.
Interactions must be taken in context
Likes, Retweets, and other interactions have always played a prominent role in the Klout algorithm. We believe these are valuable signals of influence. What we found though is that some people are extremely generous with these interactions. People should Like and Retweet to their heart’s content, but we believe that interactions need to be measured in the context of the person interacting. This was the most prominent reason why some scores dropped.
Likes, Retweets, and other interactions have always played a prominent role in the Klout algorithm. We believe these are valuable signals of influence. What we found though is that some people are extremely generous with these interactions. People should Like and Retweet to their heart’s content, but we believe that interactions need to be measured in the context of the person interacting. This was the most prominent reason why some scores dropped.
Example: Consider two users who Retweet my Tweet. User A Retweets me but she also Retweets 100 others in the same day. User B Retweets me and only me. We now consider these ratios in our algorithm and consider the singular Retweet as a greater sign of influence. Similarly, if you selectively only give out one Facebook Like a week and you choose to do so for my content, that is much more meaningful than if you Like 50 times a day.
Stability and consistency
Seeing the ebb and flow of your influence on a daily basis is helpful, but we also understand that your influence rarely makes huge jumps in short intervals of time. We considered massive spikes and steep drops as problems in the way our algorithm behaved. Our new algorithm makes the Klout Score more stable by taking a longer window of time (90 days instead of 30) into account when measuring your influence.
Seeing the ebb and flow of your influence on a daily basis is helpful, but we also understand that your influence rarely makes huge jumps in short intervals of time. We considered massive spikes and steep drops as problems in the way our algorithm behaved. Our new algorithm makes the Klout Score more stable by taking a longer window of time (90 days instead of 30) into account when measuring your influence.
Example: We used to hear about “the vacation problem” where users saw a steep drop in their Klout if they took a break from social networks while they were off the grid. Now the Score will remain much more stable during short breaks from social media.
These are three of the main improvements in this algorithm change, but there are many more small improvements in this release. With this release, the average Klout Score is close to 20 and a Score of 50+ puts you in the 95th percentile. We now analyze 2.7 billion pieces of content and connections a day.
We are continually improving and solving new problems with our science team. We appreciate all your feedback and are working to help you better understand what goes into the Klout Score with this new series. Let us know what you think!
