The first Women in Data Science Conference, held at Stanford University, brought together 500 women in data-related industries. Women from all over the country met to hear about the latest data science-related research and meet others in the ecosystem.

Caitlin Smallwood, vice president of science and algorithms for Netflix, Inc., met up with Jeff Frick, cohost of theCUBE, from the Silicon ANGLE Media team, to talk about how the international provider of on-demand Internet streaming media uses algorithms and recommendation engines to improve the customer experience.

Value of slicing and dicing

In her presentation at the event, Smallwood demonstrated how over 50,000 combinations of row labels contain specific criteria for each movie’s characteristics play a part in customer’s movie choices.

It is a laborious process, however. Netflix uses the data to view patterns and analyze various influences and customer behaviors.

“It is interesting to see how small the head is and how long and flat the tail is” when reviewing the patterns, Smallwood stated.

While knowing the company cannot study every one of its 69 million customers, Smallwood’s work involves using different algorithms to compare the different labeling techniques or mechanisms to show what labels are working. Smallwood considers the algorithms as unsupervised learning that enables Netflix to view the raw structure patterns in customer consumption.

Cause and effect vs. correlation

According to Smallwood, Netflix as a company is dedicated to experimentation. “It’s the one thing that helps you get a causality instead of correlation,” she said, maintaining that even the best statistical models and data produce correlation rather than causality. Therefore, only a controlled experiment is the only reliable way to get a causality.

Experimentation is not always possible, and for those instances Smallwood believes that using the models and algorithms is certainly better than human judgment alone. However, they don’t hold a candle to what you can learn through the practice of experimentation, she added.

Content development based on data

Data is now playing a role in the decision-making process for the network’s original series content. While it is not the ultimate factor in the process, the organization can review the data to determine what show names and scripts could work, what value it will bring the audience and if it is worth producing.

Smallwood pointed out that the process is to leverage the data to enable a talented content team to make the determinations. “We are here to empower the decisions and not get in the way of creative,” she said.

Watch the full interview below, and be sure to check out more of SiliconANGLE and theCUBE’s coverage of Women in Data Science 2015.

Photo by SiliconANGLE