Are you a follower or a leader ?
Calculating Twitter authority is quite tricky. One needs to make choices on how he understands the word « authority ».
I decided to interpret « authority » in « leadership power ». In other words, I didn’t try to calculate how many people you are influent on (that’s simply your followers number), but to measure your leadership power : the weight of people CHOOSING to follow you ALTHOUGH you are maybe not following them, vs the number of people you follow…
These are the options I took in defining the algorithm :
A. Re-twitts of someone’s twitts are not a relevant parameter regarding Authority.
- Re-twittering relates to the « efficiency » of the original twitt. Not about the genuine authority of its author.
- On which time frame should these re-twitts be counted ? Last week ? Today ? Last month ? This is far too vague.
B. The number of Twitts is not a relevant Authority parameter.
- Using Yahoo! Pipes & TwitterFeed, you can create a giant feed-of-all-feeds and automatically publish hundreds of twitts per day. Does this have something to do with your genuine authority ?
C. The calculation should annihilate Bots.
- It means that users with « balanced » accounts (approximate same number of friends and number of followers) don’t have an important leadership. They just have friends. We don’t measure your friendship ability, here
D. Still, the number of followers, even if « balanced » by an automatic following, should have some weight.
- Because, even if your followers/friend ratio is balanced, there is still a difference of influence in having 10.000 followers rather than 100.
To test my formula, I included 6 types of test data.
- A major Twitter user (techcrunch)
- A Twitter user with a large number of followers and who manually chooses who he follows (loic)
- A Twitter user with a huge number of followers that he automatically follows (scobleizer)
- A Powerful blogger (ourielohayon)
- A typical average blogger (denisflorent) – that’s me
- A suite of different theoretical Twitter users with different values.
Here are the results :
These results comply with the way I understand Tribe Leadership Power.
What’s your opinion ?
A lire également / Also :
{ 11 comments… read them below or add one }
Interesting, but the average could be easily biased no ? Your options are brilliant !
Very interesting indeed work, Denis. Thank you!
Are you planning on posting the algorithm so that we can actually look at how your assumptions A-D are implemented?
Also, it would be great if we as a community developed a sort of algorithm which is open and agreed upon by a large number of people/companies. We could then use it in different applications and still share a same common understanding of what we’re talking about.
Obviously, I am referring to Magpie where we need a means to calculate the price of advertising messages. Right now, we use the follower count but are always looking for ways to improve…
Thanks for taking on this discussion.
Jan
I must have missed the formula somewhere, or at least indications. You don't consider someone who is frequently re-twetted as authoritative, for instance?
Denis, could I point out my TwitRank project (see link)… it already does what you want and includes authority of followers..
I’ll try to explain one thing by example you seem to have ignored in your authority algorithm.
Twitterer A has 100 followers. He follows everyone back.
Each of his followers follows 100 different Twitterers and A is one of these.
Twitterer B has 100 followers, too. He follows everyone back.
Each of B’s followers follows only 50 different people.
With your algorithm should the authority of A equal B’s authority.
I would argue that B’s authority is bigger than A’s because the attention of B’s followers isn’t parted that much.
What do you think?
Richard
Thanks for sharing this. I wonder how my « twitter authority » will look like based on your formula. Kindly advise.
Just tested. You get a 40 ! Congratulations
Anyway I will post a script as soon as possible where people will be able to test their Twitter Authority according to my algo.
It will be fun.
Hello Yvo.
Very interesting website, indeed.
Hello Jan.
Thanks for your nice comment.
About Magpie, I took the time to think about your model and I still think that using the simple follower count to calculate the price is far more logical than using any Authority count.
Or maybe, it should be an option chosen by the advertiser, like (Do you want to be broadcast on accounts with huge numbers of followers or on accounts with huge authority ?)
I'll contact you directly, if you don't mind.
Hi Denis. Thank you for computing my « twitter authority ». I also suggest coming up with a range and what does it mean. Example, those of us who got 50 and below, does it mean we are twitter authority newbies?
Cheers!
Well, that's a very good question !
I think that when the script will be online, it will allow us to rank a lot of accounts and then we'll be able to understand better the « meaning » of our own results.
I'll try to put a first version online – very rough – over the week-end.