Days, sometimes weeks before Hollywood blockbusters hit the theaters, fans are clicking away, surfing the web for movie-related details. Taken collectively, these search clicks could predict box office hits (or misses), music and video purchases, and other consumer activity, a new study finds.
Web search volume has been used in the past to get statistics about the present, for instance, to quantify the number of seasonal flu cases.
“There have been a few examples where people have done what’s called ‘predict the present,’ where the objective isn’t to predict flu levels in a couple weeks from now … [but] to predict what’s on the ground right now as a means of faster reporting,” said study co-author Sharad Goel, a scientist at Yahoo! Research, the research arm of Internet giant Yahoo.
“We think of [the Internet] as a tool for helping us find information, but it could also be interpreted as a way of capturing human activity,” Goel told TechNewsDaily.
Goel and his colleagues used web search volumes across the United States to predict opening-weekend box office revenue for 119 films, first month sales for 106 video games, and the weekly rank of 307 songs. To quantify web search activity, the researchers used special algorithms to count the number of times someone’s search results pulled up one of these products.
They found that these “clicks,” when taken collectively, tell a reasonably accurate story of the future.
For instance, they found that the number of web searches for the movie "Transformers 2" and the video game "Tom Clancy’s H.A.W.X.", for example, began to jump days to weeks before their release. And the ranking of the song "Right Round" by Flo Rida closely tracked with its ranking on the Billboard hit list.
The results aren’t necessarily better than predictions found through traditional routes, though. For example, information such as movie production budgets, critics’ reviews, and prequel sales also do a good job of predicting future revenues.
The real advantage is when this traditional information isn’t available, or when you want to make predictions “pretty quickly and cheaply,” said Jake Hofman, one of the study authors and Yahoo! Research scientist.
The research was detailed online Sept. 27 in the Proceedings of the National Academy of Sciences.