Japanese Startup Unveils Box-Office Prediction System at Tokyo Content Market
Japanese company Crunchers says it can accurately predict a film's box-office potential using big data
At a promotional seminar on the last day of the Tokyo International Film Festival's TIFFCOM content market, Japanese startup Crunchers introduced a system it has developed to forecast the box-office performance of movies, calling its service the "world's first practical box-office simulator."
Crunchers CEO Daichi Ishii said the high-risk, high-return nature of the film business makes accurate forecasting vitally important for production companies. "Internet content creation is very low-risk, but in the film business, one big-budget flop can cause a crisis in a corporation," Ishii said.
Ishii and Japanese director Junpei Matsumoto (Saigo no Inochi, 2014), the company's CCO, said they created Crunchers with the aim of offering a data-driven resource to production companies.
The Crunchers system uses three main variables to project a film's performance: marketing and promotion, word of mouth or the "viral effect" and the timing of the release. Ishii said the promotional effect and the impact of timing are easiest to predict, as the marketing spend and release timing in relation to holidays and vacation periods are relatively simple to quantify. But forecasting the viral effect presents challenges, he noted.
Ishii said he and his team analyzed 100 recent films to try to identify the elements that determined a movie's viral appeal in the market. His team developed a checklist of 600 yes/no questions that relate to a film's cast, "creative expression," cultural factors (such as whether the film's themes will translate well in Japan), protagonist, conflict, climax and other aspects. "Our hypothesis is that these elements determined the viral effect," he said.
The company than analyzed the 600 factors in relation to the actual performance of some 867 films. Ishii's team then determined that 60 of the factors had a strong correlation with box-office performance. "The rest are somewhat trivial," he said. "We believe we are able to predict the film's viral effect with these simple 60 items."
He added: "We finished the simulator in the middle of September and have been rushing to refine its accurately." Crunchers will be making constant updates to its the system and data. "The more data we add the more accurate we believe it will become," said Matsumoto.
The company's system has yet to be independently tested, but it will offer its services to companies and individuals in Japan in November, followed by a U.S. launch in early 2015.