A year ago some researchers at MIT believed they had stumbled across the secret behind why tweets get shared. They narrowed it down to a list of ten things that go a long way to ensuring your tweets get shared, including things such as using less than 70 characters and revealing the inner you.
MIT are back with a fresh look at the business of retweets. This time they’ve created a model that they believe lets them predict with a good degree of accuracy how many times a tweet will be retweeted. They’ve made a site to showcase the model, using famous Twitter users as live guinea pigs. You can see an individuals latest tweet, together with a prediction for how many times it will be retweeted.
The model was created by collecting retweets across a range of topics. They then explored the time of the original tweet and how fast it spread via retweets. They used this data on popularity to build their predictive model.
The researchers’ findings were explained in a paper submitted to the Annals of Applied Statistics. In the paper, the authors note that “understanding retweet behavior could lead to a better understanding of how broader ideas spread in Twitter and in other social networks,” and such data may be helpful in a number of areas, like marketing and political campaigning.
I’m kinda in the sceptic camp when it comes to things like this, as I don’t believe the viral nature of the web is something that can be easily predicted. For instance Sharad Goel looked at how ideas diffuse through social networks and found that the vast majority struggle to achieve one ‘hop’, or in other words, they might get passed on once if they’re lucky, but seldom beyond that first generation of shares.
This finding was then replicated by Microsoft a few months later. They analysed billions of pages of content, and found that those that were passed through 20 generations were literally one in a million, and that little in these lucky few could tie them together enough to form a viral recipe book. So I’m inclined to think that Twouija is a fun toy to play around with, but of little use in predicting when tweets will go viral.
Hmm, interesting. I'd like to see this tested on a much wider range of accounts though I think.
Hmmm, I've always been skeptical about predictive analysis, even though I've seen some pretty good use of it with consumer behavior research, among others. But trying to figure out how many retweets a tweet will get? Tough nut to crack, for sure.
Yesterday, @Newscientist re tweeted one of my tweets @victorantos to its followers,975k, as a result another 50 retweets happened.amazing, but i ve got only 3 new followers after the event, why?