Harvard Study of Twitter Usage is Flawed

A new Harvard study of Twitter usage has surprised some people with its results, including its authors.  The study indicates that 90% of tweets are generated by 10% of the users on Twitter and that the median number of tweets per user is one.   The researchers and the media seem to be surprised by this, but this is because they have not thought about the usage patterns of Twitters users. In other words, they don't understand how people use Twitter.  Here are some common usage patterns that I am familiar with and why they would lead us to expect such a highly bi-model pattern where one group of users post a large number of tweets and the rest do not:

  • overhead : Posting to twitter is "easy" if you use the main website, but most power users don't do that.  The main website is clunky and doens't allow you to filter the firehose of information that is a Twitter feed.  Instead, people use an installed app on their machine like Tweedeck or Tweetie or one of the many external websites integrated to post to twitter such as TwitIQ or bit.ly.  This either requires downloading and installing software or creating accounts on these services and saving bookmarks for posting quickly when a thought comes to mind.  Clearly, many people don't do this and as a result do not post much or at all through the main website.  Those who do invest in setting up a bit of infrastructure to post will understandably do it more often than those who don't.

  • iPhone : In a similar vein, many users post to Twitter via iPhone or some other cell phone, many people do not have these phone or data plans that allow posting for free so we would expect those who do post via cellphone to do it a lot and everyone else to do it not at all.

  • 140 characters : A single tweet does not carry much information, it is common for people to post multiple tweets on the same topic in quick succession to give a more full explanation.  This multiplication of user posts will add up quickly relatives to users not posting at all.

  • conversation threads :  another common pattern is a series of messages back and forth between two or three users on some topic or arranging a meetup.  This would create a larger than average number of posts for a small number of users, further pumping up the number.

  • live blogging : Twitter is an ideal medium for liveblogging events such as conferences, conventions and rallies.  This means that a single user, or the relatively small number of attendees of a conference, post a large number of tweets about the events throughout the day.

  • news/event feed : one of the most useful ways to use twitter is as your own personalized news and events feed.  By searching on different terms of interest to you, you can find users who post regularly with comments, news and links that you want to hear about.  You can then follow them to keep up to date on the topics they seem to be interested in.  This usage pattern relies on the fact that some people post a lot and most people read a lot.  The user reading their feed has no need to respond to all the posts they see and no one would expect them to.

  • games : A recent trend, is for people to create twitter trend games where a hashtag keyword is defined, such as #crapsuperpowers, and people post back with the hashtag and their input.  Some games involve constructing sentences using movie names or based on some quote in the news.  The point is that there are people who play games and once they play one they are likely to play many, probably multiple times as they think of better answers and respond to other great answers by others playing the game to vote on the winner.  This encourages a small group of users who like playing games to post many tweets that non-gameplaying users would not post.


For all these reasons, we would not expect the usage patterns of twitter users as measured by tweets per user to be a continuous, smooth scale such as  (none, once a month, twice a month, thrice a month, etc..., Ashton Kutcher).  We would expect a large number of users who logged on and didn't get it, another group who post very rarely using the clunky web interface and a smaller group of high users ranging from news feed readings on the low end to  new-posting, games-playing, live-blogging, iPhone-using super-tweeters at the high end.

That is Twitter, that's how it works and how it's supposed to work.  There is no need to change the service to increase its usage based on a flawed measure such as tweets per user.