Author-screenwriter William Goldman summed up the film industry by saying, “Nobody knows anything.” However, Concourse Media and Productivity Media are betting that reverse-engineering scripts via a smart computer process may prove that adage wrong. Take the fact that their mock trailer for the horror film Impossible Things, which is only at the script stage, has been viewed 562,000 times and already has 33K followers on Facebook — without a single cast member attached and not even one frame of film to show.
The script for the film was basically developed by relying on computer analytics in a new system that uses machine learning (artificial intelligence) to develop a project targeted by genre, story, demographics. It can even calculate in particular cities where that genre of film worked best — defining that audience and building a project finely tailored for them.
Developed by Jack Zhang through his company Greenlight Essentials, the system will be on-site in Toronto to unveil the A.I. program to agents, distributors and producers. Executives from all three companies will be on hand to discuss this new way they say takes a fair amount of the risk and guesswork out of financial decisions when it comes to making films.
The program uses a Lowess Analysis which is adaptive so as you change variables, the outcome changes. Zhang databased 3,000 titles over 15 years, then curated it by genre, director, and key plot attributes and matched those up with consumer demographic profiles gleaned from user behavior on social media.
“Jack was at the University of Waterloo in a machine-learning course and always was interested in film, but was curious to see if there was a way he could find evergreen elements within (the films themselves),” said Productivity Media’s William Santor. “He developed an algorithm that was able to pull and curate all this data and how it correlates to the success or failure of a film.”
By doing that Zhang pinpointed the evergreen elements that could be used for Impossible Things. They extracted just under 20 “genes” (as they call it) for the horror film. Based on those analytics, they were able to draft a screenplay using those evergreens. Said Santor: “After identifying 20 genes, we took it over to a horror writer to put all these elements into the story and work on the shape of the film. The result is Impossible Things.”
Concourse’s Andrew Felts said the company had been using this computer data system since December for help on “validating their intuition” for other areas of their business, but noted: “This is the first time that we used this to develop a screenplay.”
“Everyone has their own process for deciding on a title,” Felts says. “We’ve seen how people run comps but nothing has compared to this. Not only are we seeing what the market is and historical data, it backs it up with actual consumer data that is current and then it allows you to make a current and informed decision.”
Added Concourse’s Matthew Shreder: “Executives can now make better financial decisions by using the data and audience tools that (this system) provides. For Concourse, it mitigates risk for us as we package, finance and sell movies.”
The data dive is deeper than just using the 3,000 titles as it also evaluates current moviegoer behavior and consumer tastes gleaned from social media. For instance, the computer can identify key territories overseas and key cities in the U.S. where consumers embrace a particular genre. For the macro horror genre, for instance, the data points to China as the No. 1 market, the UK as No. 2 and Japan No. 3. However, when you change it up and ask about the horror genre that feature zombies, Japan becomes the No. 1 market, followed by the UK then South Korea.
Likewise, in the U.S., the analytics will tell you what top cities the macro-horror genre will perform best — they are heavily based in Texas, followed by a town in New York and then Colorado. “So what the data tells us is that if we released a macro-horror film, we would get a better per-screen average by releasing the film in those particular cities that historically have embraced this genre.”
With that kind of information, advertising spends could be much more targeted as well. Zhang, who was not available for an interview as he was in China, told Concourse and Productivity that he could use the information to start building an audience for the film that didn’t exist.
“I said, yeah, good luck,” said Felts. “But in a matter of two months, we saw it grow from zero followers to 30,000 followers for Impossible Things and he spent only $500 in advertising. In order for us to do that, we would have had to spend $5,000. In order to get followers in today’s marketplace on social media, you do have to spend money. It happens one of two ways: You can pay or it can be organic. What he showed us is that Impossible Things can flourish organically after a small ad spend.”
The mock teaser trailer for Impossible Things has now received 562K views with positive comments from females under age 25. The Facebook post has been shared almost 7,000 times in less than two months.
Impossible Things centers on a mother’s disturbance over her daughter’s death. As her out-of-work husband leaves the city and moves to an idyllic and secluded country home, the mother stays to care for her two other children. When they begin to hear voices and see visions of a deranged woman and child ghost, reality becomes skewed and insanity grows and she succumbs to the house’s demonic traps.
The program Zhang developed specifically uses public social media data on a global scale and then analyzes thousands of indicators as they relate to the relationship between moviegoer tastes, film-plot attributes, and industry trends. Results can be as specific as the user wants, ranging from talent trends over time, specific genre storylines, locations in the script, and casting certain demographics or age groups.
Managed by Santor, John Hills and Andrew Chang-Sang, Productivity Media recently financed and executive produced Born To Be Blue starring Ethan Hawke, and The Little Hours starring Alison Brie, Dave Franco, Kate Micucci, Aubrey Plaza, John C. Reilly and Molly Shannon. The system was not used for those films.
“By curating the social media and consumer data, we now can actually profile what the audience is going to look like,” said Santor. “We further were able to pull up consumer profiles by gender, age, even taste in products.”
The computer does that by monitoring what brands fans follow and engage with online. For instance, Santor says, “On the macro-horror level, we were able to find out that Coca-Cola is probably putting the wrong product into the movie for placement for that particular genre. We found that those fans respond more to Sprite.”
Makes you kinda rethink being on social media, no?
Zhang said in a statement: “With the use of our modern technology, we believe we can help take the entertainment industry to the next level.”