One of my colleagues in my lab here sent me a link to this paper, by Sitaram Asur and Bernardo Huberman at HP – Predicting the Future with Social Media (PDF). I’m not sure if this was published in any conference proceeding or a journal. Either way, their findings are really cool.

They take data collected from the Twitter Search API and extract data about 24 big Hollywood films — i.e. how many times people talk about the movie, post a link to the film’s website, etc.  Then, they take this data and compare it to data from the Hollywood Exchange and the box office returns.  They show that Twitter chatter is a tremendously accurate predictor (R^2>.90) for the box office returns. Even better, they show that for the films they selected it is a better predictor than the Hollywood Exchange.  They expand on their findings by considering the “sentiment” of the tweets and show that this is also a significant indicator of week 2 movie success. The twitter chatter is a good analog to traditional word-of-mouth marketing.

This is the first quantitative finding that I’ve seen that shows a social network is a good indicator of some product’s success. There are lots of business books out there claiming that viral and social marketing is going to revolutionize the world.  The general interpretation seems to be: if we plaster our product all over social networks than this automatically equates to higher profit. The above findings show that this is generally not always the case; social networks are more of a conduit for traditional word-of-mouth advertising. Simply just plastering social networks with your promotional material does not appear to be a good predictor of your early performance.  Furthermore, if people think your product sucks, negative sentiment among the social network users does translate to poor performance later on.   So, I guess it still pays to make a good product, rather than unnecessarily spending money advertising your promotional material on social networks.