Jul 1

Ben Lorica

Ben Lorica

On Friends, Followers, and the Top Twitter Users

An easy way to increase Twitter's signal-to-noise ratio is to follow less people. I'm sure you've heard of Twitter users who follow several thousand twitterers. How they keep up with that many micro-blogs is beyond me. Unfortunately, spammers have discovered that to increase their "following", they simply follow thousands of other users, a small percentage whom will politely start following them. Granted, number of followers is not as informative as the number of conversations a twitter user starts or posts that get retweeted, it is currently the only metric that is generally available.

The Twitterholic ranking does not appear to have been overrun by spammers, although some of the top users follow a substantial number of twitterers. According to Twitterholic, the top users averaged around 10,200 followers. Assuming that each of the top users follows a fraction (say 5%, or 1 in 20) of those following them, that translates to the top users following about 514 other twitterers on average. It turns out about half of the top twitterers follow a larger fraction than that. In the graph below, the green line represents 5% of the number of followers (friends-to-followers ratio equal to 5%). As per Twitterholic, I use the term friends to refer to the number of users a twitterer is following:


I highlighted the users who follow more than 15,000 other users, including one user ("ringernation") who follows more than 100,000 twitterers. 100K is about one-tenth of the twitter user base! Over the second half of June, ringernation increased his following by 61%, going from a followers-to-friends ratio of 3.8% to 6%.

Needless to say, when someone is following that many other users, chances are he is no longer paying close attention to what others are saying: just because Scoble is one of your "followers" doesn't mean he is even speed reading what you're tweeting. As Tim observed a few weeks ago:

For example, among the top twitterers, it's pretty clear that many of them are simply following anyone who follows them, which drives their "popularity." But that makes clear that they aren't actually following any of those people -- the volume is just too great. So ironically, if you follow everyone, you follow no one. (Unless you "friend" them, and only really follow your friends.)

So you can see that there are three categories of twitterers: those who use it for its original purpose, by following and being followed by a small group of friends; those who use it for marketing, by broadcasting to many but following none; and those who recognize the asymmetry, and are followed by many, but follow fewer.

Finally, in case your wondering, 42% of the top users follow more than 666 other twitterers.

tags: social networking, twitter  | comments: 6   | Sphere It

Previous  |  Next

0 TrackBacks

TrackBack URL for this entry: http://blogs.oreilly.com/cgi-bin/mt/mt-t.cgi/6583

Comments: 6

  Logical Extremes [06.30.08 10:29 PM]

A little while back, I guessed at a not-dissimilar graph, but with a more simplistic characterization:


  Mike D. [07.01.08 12:15 AM]

Some of the popular Twitter users have a technique that allows them to give the illusion that they are following tens of thousands of people when they are really only following a few hundred. They use an IM client - e.g. GTalk - to read tweets and only turn device updates on for their close friends - maybe a few hundred people. They can use Track (when it's working) or Summize to do a vanity search on their own name. That allows them to see any tweets from the thousands of people that they are not following on IM and respond accordingly.

  Danny Howard [07.01.08 04:03 PM]

A while back I thought of the spam problem and figured we could calculate a "TwitRank" for users. Basically, you look at the ratio of friends to followers: most users should be around 1. Some famous users will have a positive score, and spammers would have a very negative score.

You'd have to account for fake "follower" networks at some point, probably by weighting the followers based on their basic twitrank. That way "spammy" followers would drag you down.


  iPhone Fan [07.02.08 05:40 AM]

Danny, nice idea... though the spammers will surely find a way around that barrier. As usual it is a cat-and-mouse game... and an end is not in sight.
There are a number of corporate offers regarding cheating the social network structure.

  Mark Drapeau [07.26.08 04:26 PM]

Nice post. Danny, with regard to "Twitrank" you can try http://twitterratio.com/. It would be cool to incorporate that directly into Twitter.

Personally, I have gotten a good number of new followers for various reasons, but am keeping the number of people I follow under 150 (following the 'rule of 150' from The Tipping Point).

Now, I modify that list about every two weeks by periodically adding new, interesting people until I go over 150, and then un-following people who have ceased to be interesting or informative. This works pretty well; I engage with about 75% of my 150.

  kion [04.29.09 07:52 AM]

You can also check this service out: TWIT•RANK, which is based on Google's Page-Rank principles for ranks calculation.

Post A Comment:

 (please be patient, comments may take awhile to post)

Type the characters you see in the picture above.