New algorithms drive film marketers
Leaving a “hot” screening the other day, I had to pass through a solid wall of people who were avidly tweeting their reports on the film. My instant question: Will anyone actually read these tweets? Everyone seems to be sending but no one’s reading.This week I actually found one dedicated tweet reader, but here’s the problem: It’s a computer. It resides at the Annenberg Innovation Lab and it has been taught how to analyze as many as four million tweets at a time in its assignment to study the shifting sentiment of the social media. The Annenberg computer (at USC) has dabbled in film analysis. Its “bubble charts” predicted trouble for “Larry Crown” and “Green Lantern” and suggested that there are serious marketing challenges for “Cowboys & Aliens,” which seems to be generating remarkably little buzz. But the study’s aims are broader. According to Jonathan Taplin, an Annenberg professor who also is managing director of the lab, the IBM technology is also tracing attitudes toward Republican presidential hopefuls (Michele Bachmann keeps gaining ground even though many of her followers aren’t sharp enough to know how to spell her name). Also under study are the contentious presidential candidates in Egypt — insights that are being fed to the State Dept. While the lab’s computers have been taught to read, they’re still not strong on nuance, which frustrates those denizens of twitterdom who are prone to irony. Nonetheless, they likely will provide a valuable resource for institutions and businesses that are all but frothing at the mouth to understand moment-by-moment shifts in the social media. Several intriguing businesses have sprung up that focus on the entertainment community. A company called Fizziology with ad agency roots and based in Indianapolis, relies on a corps of freelancers who peruse not only Twitter but also Facebook and blogs with the aim of forecasting box office performance. It made some good calls this summer, predicting a much bigger opening for “Super 8” than did the more traditional tracking studies. Ben Carlson, a partner in Fizziology, says his company also provides insights on casting decisions. “If a director is looking for a ‘hunky’ leading man, we can tell them who’s most widely described as ‘hunky,’ ” he says. Another entity, with the name of Crimson Hexagon, mobilizes its proprietary algorithms to forecast box office and provide insights on casting and marketing — with an exotic name like that, the company, obviously, like Facebook, was conjured up by Harvard students. Crimson Hexagon became a client of the UTA talent agency earlier this year, except that UTA has now turned around to hire the company as a resource. Inevitably, the veteran players in the tracking and research business are skeptical about newcomers, as are the senior studio executives. Companies like NRG and Marketcast (a Variety sister company) have long specialized in tracking, but their biggest profits stem from testing trailers and television spots. According to the old pros of the testing business, the newcomers don’t have enough sophistication in breaking down the demographic appeal of the product that they analyze. With Hollywood’s tentpoles increasingly reliant on international box office, this arena, too, is a topic of controversy. Studio insiders claim that they understand the foibles of foreign markets while the new players argue that Hollywood is still clumsy in exporting advertising material en masse to countries without regard to specific national tastes and idiosyncrasies. The one area of agreement: Staying tuned to the social media around the world is mandatory. Marketers desperately search those key words that would suggest they have tapped into an iconic subject. Amid the global marketing melee, their holy grail is brand awareness. I would like to have a private conversation with that Annenberg computer that has learned how to read. Putting key words aside, I would ask, How do you assess the intelligence of your tweeters? Are Americans as illiterate as they seem? I don’t think I would get a nuanced response.