A Computer Is Deciding What Movies You’re Going to Watch Next

April 21, 2017, 3:00 PM UTC

Hollywood allocates considerable brainpower and capital toward crafting the perfect formula for a blockbuster. And now a growing number of tech startups are taking that idea literally.

Take Vault, an Israel-based artificial-intelligence startup that’s one of the newest entrants using analytics and algorithms to predict ticket sales.

Vault CEO and co-founder David Stiff says his company’s 4CAST platform can analyze the box office potential of a film based only on the “core story DNA” gleaned from a raw screenplay or a movie trailer. Founded in 2015, Vault has spent two years honing a neural-network algorithm that relies on 30 years worth of box office revenues, film budgets, audience demographics, and casting information to help determine box office potential.

And it’s not alone. Other companies in the space include the U.K.’s Epagogix, founded in 2003, as well as ScriptBook and Boston-based Pilot. Forward-thinking studios like Legendary Pictures (now owned by China’s Dalian Wanda) also rely on their own in-house analytics teams to devise data-informed marketing strategies.

Stiff admits his platform whiffed on the recent surprise box office success of the groundbreaking satirical horror movie Get Out after underestimating the film’s social media buzz. But ultimately, he says roughly 75 percent of Vault’s predictions come “pretty close” to the films’ actual opening grosses. The company recently predicted a $16.1 million domestic opening for the sci-fi thriller Life, which debuted at $12.5 million, according to Box Office Mojo. A little farther afield? Buddy cop comedy CHiPs premiered to $7.7 million, versus Vault’s $15 million prediction.

While Hollywood tweaks the code for the box office battle, a larger war for audiences’ attention looms. Box office forecasting has always been en vogue in Hollywood, but the movie industry’s eternal conundrum has a new element of urgency thanks to the ascendance of streaming platforms like Netflix and Amazon Video.

With their treasure troves of user data, those online distributors have access to a wealth of information that can help gauge exactly what audiences want to watch—and how and when. And, while traditional movie studios don’t have access to that same data (which Netflix and Amazon do not share), box office results are mostly public knowledge.

In other words, traditional Hollywood’s search for the perfect algorithm could pale in comparison to the Big Data battles to come.


AI Predicts This Summer’s Opening Weekend Box-Office Hauls

    $21.58m on 3,500 screens
    $17.3m on 3,400 screens
    $15.6m on 3,000 screens
    $13.3m on 3,000 screens
    Algorithms can use inputs like Amy Schumer’s popularity to project ticket sales.
    $9.2m on 2,900 screens

Source: Vault

A version of this article appears in the May 1, 2017 issue of Fortune with the headline “Hollywood’s Search for a Blockbuster Algorithm.”