A study released yesterday by Pear Analytics analyzed 2,000 tweets in English, from U.S. users, randomly captured in half-hour increments between 11 am and 5 pm CST, over a two week period. Each tweet was categorized into one of six categories: News, Spam, Self-Promotion, Pointless Babble, Conversational and Pass-Along Value.
The results: 40.55% of the tweets were pointless babble. Conversational tweets (conversations between people) represented 37.55% of the 2000 tweets, and Pass-Along Value (any tweets with RT) represented a very distant third – at 8.7%.
Anyone who conducts substantive original research deserves a lot of credit. When the original research looks at the data in ways that others haven’t measured before – this credit is well deserved. Kudos to Ryan Kelly and his team on a job well done.
When I initially read the study, I wondered whether the sample of 2,000 tweets was statistically significant. I asked Ryan (on twitter). Ryan said that his team knew that there were 3 million tweets per day on twitter, and that a “inferences of a billion is estimated with data of a few thousand.” Ryan also added that the trends were pointing in the same direction, so it was not necessary to sample more data.
I hope Pear Analytics considers slightly expanding its analysis in future studies by assessing how their measured data is impacted by some or all of the following:
- Five percent of Twitter users create 75 percent of the tweets (according to Sysomos). It would be interesting to compare the five percent to the other 95 percent, and categorize the results. Maybe we’d find that the pointless babble is created primarily by the thousands of “social media” experts on Twitter. Or the 24 percent of Twitter users who are bots (and presumably, also social media experts).