Measuring Influence – My Algorithm

Over the weekend I had some great dialog with some Twitter folks (@peoplefw, @TedRubin)  about the value of Klout, Kred, etc. Spawning from my previous post about “Rewarding Influential Customers” the discussion quickly became one of the value of such applications and are they really accurate. I spent some time thinking about an algorithm that “could” suffice for establishing true influence.

In my simple mind the concept of sharing should not even be in the equation of an influence score. Anyone can “share” something. Real influence is how many people read it, “liked it”, commented on it, and re-shared it. And that’s where I think the focus should be.

Influence also changes over time, so having the algorithm time boxed is key. You may be influential today but not tomorrow. So from the aspect of “sharing” you must at least share something to get an influence score – so “sharing” does have an impact.  What won’t have an impact is the number of times you share.

The points would be calculated in a tree model based on likes, shares, and comments. Let’s keep it simple and have a similar scale as Klout, 0-100 points total. The points would only really be calculated on the “best” share you did. So if you had a share that was re-tweeted 200 and another re-tweeted 2 times you get credit for only the most shared content and you don’t “add” the two together.

The same scoring could be used for Facebook, LinkedIn, Google+, Pinterest, Instagram, etc. As you can see you would have to have thousands, if not millions of shares to achieve a score of 100. Most of us, people like me, would be happy with a score of 10.

My scoring system would be based on a point system for different periods, 30, 60, 90 day periods. In the graphic below you can see this persons influence has diminished in the past 90 days. Could be lack of sharing or lack of sustained influence.

The problem with this scoring system is the system would need access into a persons account and all of their children accounts and their children accounts down the line. This is what makes a true scoring system almost impossible due to the OAuth 2.0 model and access into the child accounts data.

Now, if all of these social networks subscribed to this service it could work. The systems could then show your global “influence score” on your profile page and the score would not come from another site. Don’t be surprised if the big boys come out with their own influence score in the future. Getting them all to buy into this is most likely impossible.

This scoring system is only a few hours worth of thinking and probably has many holes in it. I would be interested in hearing your opinions about influence scores in general or if you have your own idea of scoring. Maybe this would be a great open source project – OpenInfluence!

 

Short URL: http://bit.ly/Y9u2kP