Since the introduction of Lists in Twitter, there has been some speculation about how Twitter Lists could help indicate Influence. See the following for some background:
- Using Twitter Lists to Judge Influence
- How Twitter Lists Influence…Influence
- Twitter Lists: What Is Your Respect Ratio?
- Twitter Lists to Followers Ratio: The New Social Media Metric
It is clear that interest focuses on the ratio of your Lists to Followers.
I decided to assess whether this new metric correlates in any way to existing influence measurement tools. The objective was to assess whether the metric has any correlation with influence ranking algorithms that do not use Lists information. For my experiment, I considered influence measurement tools like Twinfluence, Twitalyzer, and Klout.
Is this a Big Deal?
Not for casual users. There can be important implications for serious users. Since the advent of Twitter, the number of followers has been considered to be a rough indicator of influence. As a result, very few have taken pains to actually filter their followers and weed out spammers and bots. In 12 Tips to Enhance Your Twitter Reputation, I had discussed how you should do this. If the Lists-Follower metric is widely used for influence measurement, you will see people actually scanning their Followers.
This can also become important because your influence may determine the ranking of your tweets in search results.
Influence Ranking Tool
My tool of choice was Klout, for the following reasons:
- Speed. The tool had to process and rank influence for each member of my sample set quickly.
- Twitalyzer gave unlikely influence ranks for some people I knew.
- Klout is transparent in revealing what factors it considers and changes to their algorithm. This will be useful in revisiting this after it incorporates Lists information.
- Klout Score uses 25-30 variables to be comprehensive, unlike Twitalyzer, which uses only 5.
Sample Selection
I used 40 Twitter users I follow for creating my dataset. I only considered accounts that represented people, and not brands. For my dataset, I selected:
- Those with more than 10,000 followers
- Those with a ratio of Followers:Friends > 10:1
- Some more users at random to form a long tail for the analysis, all of whom have more than 1000 followers
- I couldn’t resist including myself, as one user with <900 followers
The result of my experiment looks like this, with the accounts ordered by decreasing no. of followers:
