Skip to Content, Navigation, or Footer.
Thursday, Feb. 29, 2024
The Observer

Expert examines the role of data analysis at Netflix

Expert in Data Science at Netflix spoke on Wednesday in Little Theatre at Saint Mary's.
Douglas Twisselmann, expert in Data Science at Netflix, spoke on Wednesday in Little Theatre at Saint Mary's.

Saint Mary’s hosted senior data scientist Douglas Twisselmann on Wednesday night in Little Moreau Theatre. Twisselmann is a member of the science and algorithms group at Netflix and works with the branch dealing with the media content Netflix provides for its viewers. Twisselmann’s talk focused on Netflix’s goals of identifying characteristics of an “ideal” content library, predicting demand for content that Netflix does not have and divining the next Netflix original series.

Twisselmann works with the Netflix content team to license, purchase and develop the movies and television shows that will be featured on the streaming service. Netflix has more than 60 million viewers across more than 40 different countries, and it falls to the content team to predict the material viewers want to watch and create a content library to fit that criteria.

“We always want to keep the viewers happy,” Twisselmann said.

All aspects of data science are modeled around consumer science testing, which allows Netflix to have personalized content libraries for all of its viewers, Twisselmann said. Netflix does not base its suggested libraries based on gender or age but on the content that one watches, Twisselmann said. This way, Netflix can send users personal updates and messages about one’s favorite shows or potential options that they may enjoy, Twisselmann said.

“With a lot of our content acquisition, because we tend to buy for long periods of time, we’re less concerned if someone is going to watch it tomorrow; we’re more interested, on average, who is going to watch it,” Twisselmann said.

“Our home pages are very personalized based on what people watch,” Twisselmann said. “We’re one of the few companies that doesn’t ask background information. If we don’t know a lot about the individuals, then we can’t make assumptions about them.”

When looking for content to add to a Netflix library, the data scientists have to determine what the ideal consumer catalog would look like, the span of the content and the depth of the content, Twisselmann said. The content cannot be too repetitive and it must be worth the cost, Twisselmann said. He said the value of each potential program is not based on revenue but the number of viewers and how much the viewers want to watch over a certain period of time.

The key to predicting is to analyze data categories such as how well the program did in the theaters, what the program was rated, the time since it has been released and the actors and actresses performing in it, Twisselmann said. To be able to do so quickly, Netflix has created a linear predictive system called “The Crystal Ball,” based on a simple y=mx+b formula, Twisselmann said.

“It’s pretty straightforward in actually doing it,” said Twisselmann. “It’s interpreting [the data], that’s the hard part.”

Netflix's goal is to eventually perfect the system to work globally, Twisselmann said.

There a certain culture attributed to Netflix — one that involves freedom, hard work and high performance.

“I love working there,” Twisselmann said. “It’s a culture with freedom and responsibility — freedom to do what’s right and the responsibility to do what we know is right.”

With all of that freedom, there is a certain level of standard within the company, as well, Twisselmann said.

“We spend most of our time in meetings to make sure everybody is in line and on the same page, and then we go off and do our own thing,” Twisselmann said. “However, we have a high performance culture where trying isn’t good enough — it is very demanding.”

Twisselmann encouraged students to consider data science as a career.

“Everybody’s hiring in data science — Netflix, Facebook, Google, Yelp! — you name it,” Twisselmann said. “It’s a young business.”