How Data Science At Netflix Turned Hollywood On Its Head

Approaching movie-making like a complex logistical and creative puzzle which can be optimized by data and technology. JL


S.C. Stuart reports in PC magazine:

"Using data science, analytics, machine learning, and optimization, we can support content creators' decisions from pre-production through principal photography and editing, VFX, sound mixing, as well as quality control.Then we use data to help [choose] good options, rather than defining solutions from scratch. We found [the logistics] to be similar to a very complex supply chain."
In the September issue of GQ, Cary Fukunaga, the newly minted director of the James Bond franchise, says he doesn't mind taking notes from Netflix's algorithm. In story meetings for the streaming giant's latest series, Maniac, for which he is an executive producer, Netflix "can look at something you're writing and say, 'We know based on our data that if you do this, we will lose many viewers,' [and] I have no doubt the algorithm will be right," Fukunaga says.
With over 130 million global members, Netflix is bringing tech industry efficiencies to Tinseltown to make sure its system is future-proofed and can scale, fast. One of these Northern California transplants is Jen Walraven, who joined the streaming giant as a senior analytics engineer last year and was promoted to manager of Science and Analytics in March.
A former Deloitte analyst who studied computer science at University of California, Berkeley, Walraven spoke at the recent Data Science Salon, where she detailed how data can be used to help creative teams tackle the operational and logistical challenges of content production.

But Walraven's talk wasn't about the algorithms that magically bind you to the screen, hour after hour, pausing only for bio-breaks and dubiously nutritious snacks (that's someone else's responsibility). Instead, she explained how her team's work underpins the complex operation behind production, giving content creators the freedom and resources to create their stories globally.
"Using data science, analytics, machine learning, and optimization, we can support content creators' decisions from pre-production through principal photography and editing, VFX, sound mixing, as well as localization and quality control," Walraven said. "Then we use data to help teams [choose from] good options, rather than defining solutions [to operational challenges] from scratch."
Through a mercifully snappy set of slides, Walraven gave a sneak peek into how assistant directors used to "break down" the script using strips of paper (the data scientists in the audience audibly groaned at such an outmoded process) to plan a production shooting schedule.
"Script are like the blueprint of a production," she explained. "From there, producers will then reach out to their network and start to assemble all the moving parts. We looked behind the scenes and found [the logistics] to be similar to a very complex supply chain and not without challenges."
Walraven quickly clicked through the problem on stage—the multiple formats of digital software involved (script software, scheduling and more); paper-based communications (we all shuddered again) and so on.
Then (we saw this coming), she showed how her team is diving in headfirst to build tools that can ingest information to support and inform decision-making in an understandable, coherent, and transparent way.

"What we think about now," she continued, "is using data from the entire production life cycle. This spans planning and logistics in pre-production, principal photography—filming, tracking daily schedules and budgets, managing vendors—and post production; every day making sure we can help support content creators as they lead productions on time and on budget to make content for our global audience."
Walraven demonstrated sample visualizations of things like soundstage bookings, subtitles, travel, and equipment. "A lot of effort goes into making [the content our members see] true to the original creative vision," Walraven told us. "So we use data to build tools to enable content creators to develop their work across geographic boundaries."
I looked around the audience but couldn't spot anyone I knew from production. Safe to say, and I've spent time both working on film sets (post university) and visiting them as a journalist, what Walraven was outlining is a smooth and efficient system I've never seen IRL.
Productions are a notorious time-suck—a lot of hanging around, running into logistical errors constantly, and that's after sometimes years of development hell, before you even get to the set."As our investment in original content continues to grow," Walraven added, "data can help us make better decisions, more effectively. [By using] data alongside human judgment, we're researching ways to develop interactive tools that can inform discussion among creative teams as they discuss things from beginning to end like cast availability, sound stages, union requirements, and schedules."
So creatives better start doing their homework if they want to use the right buzzwords when pitching the newest studio heads. The geeks are here and they're tracking, analyzing, and smartening up this crazy town.