Searching for a Hit Movie

In the movie The Internship (involving two ex-salesmen and their internship at Google), Billy McMahon, the character played by Vince Vaughn, notes, “Sometimes, the long shots pay off the biggest.” While this statement is certainly true of movies—who would have predicted the multimillion dollar box-office performance of The Blair Witch Project?—it is also true of other new products. The main challenge facing marketers, of course, is the ability to come up with an accurate forecast of whether a product is going to be a hit or a bomb prior to its introduction or release. Yogi Berra is famously supposed to have said, “It’s tough to make predictions, especially about the future.” (This quote is an example of how it may be difficult to even predict the past: a casual search of the internet reveals the somewhat apocryphal nature of the line—others have attributed it to Niels Bohr, Sam Goldwyn, Dan Quayle, and Mark Twain, among others.)

Nevertheless, marketers invest in methods and processes that enable them to get a better handle on this issue. For example, BASES (the Booz Allen Sales Estimating System), when used in the context of consumer-packaged-goods products, measures trial-and-repeat rates for products prior to launch and then adjusts these numbers on the basis of variables such as awareness (generated via advertising) and distribution to obtain a forecast for the new product. Similarly, other methods have been suggested for products such as durable goods (the BASES model) and technology products. Most of these methods rely on stated preferences, i.e., what consumers feel they are likely to do in the future, combined with knowledge of what the company intends to do to influence these decisions (prices, advertising levels, distribution, etc.).

One industry that is constantly putting out new products and hence faces the repeated challenge of forecasting is the movie business. As Chili Palmer, the character played by John Travolta in the movie Get Shorty puts it, “Rough business, this movie business. I’m gonna have to go back to loan-sharking just to take a rest.” Typically, movie studios commission tracking surveys that begin several weeks prior to the release of the movie to try and forecast box-office performance of the movie on release. (My paper, “A Pre-Diffusion Growth Model of Intentions and Purchase,”published in the Journal of the Academy of Marketing Science, describes this process and constructs a formal statistical model to forecast movie ticket sales.) A challenge with this approach is that sample sizes tend to be limited, the respondents represent limited geographic coverage of the potential market, and stated intentions often do not translate into actual sales.

The internet has allowed studios to potentially gather a lot more information about consumers’ perceptions of a movie prior to its release. In a recent study, described in this NBC News article, researchers at Google discuss how they make use of search patterns on Google to predict the box-office performance of new releases. For example, the number of searches provides a metric of the level of interest in a film. The researchers demonstrate that an index of total box-office receipts during the year 2012 correlated well with the total movie-related search activity. The researchers note that the release of high-profile movies was accompanied by film-specific keyword searches, whereas in periods when there were no high-profile releases, most of the searches tended to be of a more generic nature—i.e., potential consumers looking for movie options. Importantly, the researchers find a strong correlation between movie-specific search activities—search volume and search-ad click volume—in the seven-day period prior to release and opening box-office performance. Once a movie was released, general search volume was no longer significantly correlated with postrelease performance; however, the search-ad click volume continued to correlate strongly, suggesting that this metric was a better reflection of consideration and intent. The researchers conclude that for “film marketers, understanding these patterns can present a substantial opportunity. By adjusting search marketing strategies to these trends, marketers can either capture the attention of the ‘curious’ moviegoer, or deepen audience engagement with a blockbuster title.”

While the Google study certainly casts important light on the role of search, there are several other sources of information, both online and offline, that are correlated with box-office performance. In a recent study in Management Science, my coauthors and I find that movies’ opening box-office performance was related to the level of prerelease advertising and the volume of prerelease blogging activity (as a measure of interest in the film, or “buzz”). However, the “valence” of the blogs (i.e., positive or negative sentiment) did not correlate well with the opening box-office outcome. On the other hand, while advertising continued to be important postrelease, the blog volume no longer mattered; rather it was the valence—i.e., what moviegoers were saying about the film—that related to performance. Postrelease, we also find that the valence of movie reviews of moviegoers (think comments on Yahoo! Movies and imdb.com) related to box-office performance.

Bottom line: the internet provides a wealth of new information about consumers’ reactions to products, which may be indicative of product performance. By carefully monitoring this information, marketers may be able to use their marketing resources more efficiently as well as effectively. Unlike George Valentin, the protagonist of The Artist who tries to resist the onset of the talkies—“I won’t talk! I won’t say a word!” (an Oscar-winning performance by Jean Dujardin), movie studios should be welcoming new sources of information that will enable them to better serve the viewing public: the customer.

Pradeep K. Chintagunta is the Joseph T. and Bernice S. Lewis Distinguished Professor of Marketing at Chicago Booth.