Using Twitter for box office predictions has proved to be a valuable tool for the film industry in recent years, and that’s certainly the case with our team at BoxofficePRO.com as well. The way the process traditionally works involves a formula or algorithm that can be spread over the masses of tweets worldwide about specific films, allowing analysts to track sentiment and sift out other similar nuggets of information ahead of a film’s release. One often overlooked aspect of Twitter analysis, however, is the ability to predict how well a movie will perform after its release. This is especially valuable for limited or awards-season releases, films that expand their screen count over a course of weeks and whose box office fortune can be drastically improved by word of mouth.
But what about films that aren’t launched as platform releases? This is the experiment we decided to take up with our social media analysis this month: looking at how Twitter performance affect a film’s box office after its initial release. Taking advantage of the tentpole-heavy summer season, we decided to track recent titles on Twitter after their initial release, setting up a simple data set that looks at the first three days of tweets for a film and comparing that to the total box office for said film. Our findings confirmed the power of word of mouth, a good indicator of a film’s long-term potential regardless the season.
Our analysis used two variables. The first was the tweet expansion over the first three days of release: (Friday tweets + Saturday tweets + Sunday tweets) / Friday tweets. This gave us a quick idea of how well word of mouth on Twitter expanded over its opening weekend. The second variable we looked at was its box office “legs” as calculated by the total box office for a film / opening-weekend gross. Any early award-season contenders that were expanding over the past couple of months were excluded from this study.
Our data set included all wide release films for 2015 and up to the weekend of April 8, 2016, a total of 152 films. These two data sets had a correlation of 0.40, which basically means that 40 percent of the variation in the box office could be determined by the tweet expansion over its first three days. That is one of the largest single determinants of total box office we have ever come across. To put it in perspective a bit, given the same list of movies, Cinemascore has a correlation of less than 0.05 or 5 percent; theater count is less than 0.02 or 2 percent.
Big performers using this method include the blockbusters Avengers: Age of Ultron, Deadpool, The Hunger Games: Mockingjay–Part 2, and Furious 7; they all had a near perfect prediction of their final gross by using the tweet returns of their first three days.
While this isn’t a perfect model, and could certainly be improved with more complex data modeling or adding factors such as sentiment or additional markets, it proved to be a valuable method to quickly gauge long-term performance. Our methodology for this analysis, admittedly, would prove flawed if tackling award-season holiday releases, considering the unique release patterns associated with them, or genres like family films, which are most frequented by demographics who aren’t particularly active Twitter users.
Word of mouth has previously been a notoriously difficult factor to properly analyze or quantify, until now. Despite the limitations mentioned above, Twitter analysis of word of mouth proved to be exceptionally insightful when dealing with the level playing field of wide-release films in our study.