Twitter jokes
(The image above doesn’t really relate to the article – I just found it amusing).

 

A friend of mine sent me a link to an article that discusses Twitter and influence. It argues that number of followers isn’t the best measure of Twitter influence. I think he expected me to argue that the article is wrong. However, I fully agree with the article’s premise. Here’s an anecdote: there is a Twitter user I recently stumbled onto who has – if I recall – ~135,000 followers and 0 tweets. Clearly that user is not influential in any way. How he got to this number of followers is a mystery.

 
An additional commonly regarded measure is number of times a user is being retweeted. While I think there is some merit to this idea, I think it fails too. There are quite a lot of users – I can name 5 off the top of my head – who have an auto-tweet mechanism that just tweets quotes and/or facts and/or news all day long. Some are bots – there isn’t even a real user behind them. Personally, I find them somewhat annoying, but manage by mainly ignoring them – though a few friends of mine detest those passionately and argue that these are what will bring Twitter’s downfall. I do have a problem with the quotes being repetitive and cliche, as well as the facts being erroneous and misleading, but that’s a different matter.

 
My point is, these users get retweeted lot. This is unsurprising: corny, cliche quotes would be retweeted since they appeal to the lowest common denominator, particularly if they are repeated countless times. It’s like popular commercials: the jingle sticks in your head no matter what, simply because you hear it all. the. time.

 
Another aspect I haven’t read about and yet would consider relevant is the number of lists one appears in. This appears to be somewhat correlated with influence. However, the more followers and/or the more active a user is, the more lists a user tends to appear in. I wouldn’t say that the more lists a user appears in, the more influential he or she is. However, I would say that a user that is not listed a lot is probably not influential.

 
Yet another aspect is interaction: how much a user interacts with fellow users. Why is this relevant? Because interaction creates relationships, relationships result in paying attention to what one says, and that is influence.

 
Ok, now that we’ve seen this is problematic, how would I measure influence on Twitter? As the article argues, it’s not easy.

 
If I had to devise my own algorithm, and a few months ago I actually considered doing just that, I would take several factors into account. The following are not sorted in order of importance:

  1. Number of followers.
     
  2. Number of followees.
     
  3. Number of tweets.
     
  4. Number of lists the user is included in.
     
  5. Number of retweets.
     
  6. Number of interactions a user has with other users (responses) – both the number of users he/she interacts with and the actual number of interactions (i.e. 100 responses to many different users vs. 100 responses to a single user).
     
  7. The characteristics of those who retweet the user’s tweets (i.e. their number of followers, followees, lists, tweets, and retweets).
     
  8. The characteristics of those who interact with the user.
     
  9. A measure that will indicate whether a user is real person or not. If a user tweet 24 hours a day, in set intervals, and never interacts in any meaningful way (i.e. it may say “thank you for the RTs” in a mechanical manner) it’s not a real person, period. I would remove those from the list completely.

What will this show me? I believe it will – generally – demonstrate who is an influential presence in Twitter. Of course, some people are exempt from this. Conan O’Brien doesn’t need to respond to anyone to be influential. But he’s influential outside of Twitter and this influence carries into Twitter.

 
Why is Twitter influence important? The article stated that companies are paying users with many followers to tweet. This is true to a degree. Companies such as SponsoredTweets and ad.ly (I believe) do factor influence measures into the decision whether to pay a user to tweet. It’s not just number of followers.

 
I don’t intend to do a full survey of Twitter influence tools, but here are three I’ve used.

 
Twitter Grader used to do an adequate job in my opinion, but it was changed so much its results are completely meaningless nowadays. I stopped paying serious attention to it a long time ago, as it is so fundamentally flawed.

 
For example, if you look for major influencers in the state of New Jersey, the #1 – consistently – on the list is a bot (it RTs useful pieces of information – but a bot, nonetheless. Until I see it interact with anyone I won’t believe otherwise). Clearly it assigns excessive importance to retweeting. Also, strangely it includes users that have tweeted a ridiculously low number of tweets and don’t have a lot of followers either (i.e. one user tweeted a total of 172 tweets – seriously, that user is a major influencer? I’m not buying that). Twitter grader also doesn’t explain how it works – it feels like a random generator of numbers. There are more issues but I’ll stop here.

 
Twitterholic (which seems to be in the process of changing its name to Twitaholic) simply shows the number of followers, followees and tweets for a specific area. That, as stated, is not an accurate measure, but it is far superior to the Twitter grader whose results are meaningless. Even this limited number of characteristics says a lot about the user.

 
Klout, I believe, does the best job. What I like about this tool is that it takes a lot of the above factors into consideration and even explains its reasoning why a user is influential or not, and classifies it into one of several categories. It’s really the only tool that does an adequate job in my opinion.

 
And there are many more. Many more. More seem to appear every day.

 
Any thoughts?