I recently caught up with Emily Glassberg Sands, Data Science Manager at Coursera, to talk about applying machine learning, neural networks, natural language processing, and big data analytics to the retail industry. Emily leads an awesome team of data scientists and data science managers working on growth, discovery, the learning experience, and a new enterprise offering. The team’s job is to help build a better Coursera through data-driven decisions and products. She also oversees the business intelligence engineering group, which drives the data modeling, data standardization, and self-service reporting foundational to analytical work at Coursera.
insideBIGDATA: Can you tell us about how Coursera uses data science, for example understanding how people learn and ultimately helping improve access to education? Anything else?
Emily Glassberg Sands: There are two primary ways we use data science to advance our mission of transforming lives through access to high-quality education.
The first is our decision science work – developing and testing hypotheses that are key to our product and business direction. Data helps us make decisions that lead to a better experience for learners – whether we’re deciding which course topics to source for our catalog, how to prioritize outreach to learners across markets and languages, or what product changes we can make to help learners stay motivated through the learning journey.
The second is our data products work – building and iterating on discovery and learning products powered by data. For example, consider the challenge of course discovery. We have more than 24 million learners on Coursera, representing a diverse range of geographic, cultural, and academic backgrounds. And we have more than 2,000 courses, from universities around the world, in just about every subject and at a range of difficulty levels. Matching our learners with the courses that will help them meet their goals demands sophisticated search algorithms, customized learning pathways, and personalized recommendations.
insideBIGDATA: Please tell us a little about your background and how you came to practice data science at Coursera? What’s your typical day like?
Emily Glassberg Sands: I joined Coursera in early 2014 as the first non-engineering hire on the Analytics team. I had just received my Ph.D. in Economics from Harvard where my research focused on experimental and applied methods to better understand labor markets and consumer behavior. In grad school, I had married machine learning and causal inference techniques to understand why people do what those around them are doing (network effects), and I had run experiments on oDesk, an online labor market, to understand why companies hire referrals (they are more productive). After four years of grad school, I was itching to contribute directly to helping people succeed in navigating the labor markets, and there was no better place to do that at scale than Coursera.
In my first few months with the company, I drove the analysis and experimentation behind Coursera’s early user growth and monetization strategies. As the company and team scaled, I started managing a small group of data scientists working on growth, monetization, marketing, and operations challenges. Today, I lead a larger team that works across the organization to develop products that empower our online learners. We work on decision science and data products for platform growth, content discovery, the learning experience, and Coursera’s new enterprise offering. I also lead the business intelligence engineering group, which drives the data modeling, data standardization, and self-service reporting that is foundational to our analytical work at Coursera.
My typical day is oriented around empowering my team to have maximal impact. This includes setting clear long-run vision and near-term direction, aligning with cross-functional partners on major shared initiatives, and helping team members grow in their roles. Much of the workday is face-to-face meetings (plus some recruiting) so I reserve early mornings before the office starts buzzing for deeper reading and writing. There’s often some admin or reactive work to do, too, but unless it’s time-sensitive I try to leave it to the evenings when I have less mental bandwidth for deeper work.
insideBIGDATA: As a female, can you give us any observations about the importance of diversity in the profession?
Emily Glassberg Sands: Diversity is critically important in data science. Empirical and computational skills are tools in the data scientist’s toolbox: necessary to be good, but not sufficient to be great. The heart of the discipline is in creatively identifying, framing, and answering questions about why humans (or other diverse actors) do what they do – and, as the literature consistently reminds us, diverse teams are more creative.
Our own diversity can also facilitate empathy for our diverse users. At Coursera, we have an ambitious goal – to be a place where anyone, anywhere can transform their life through access to the world’s best learning experience. We’re trying to answer big questions about learners’ motivations, about the challenges they face in the learning journey, and about the tangible and intangible outcomes they’ll get from the learning experience. And then we’re trying to align our product with their goals, and to build it in a way that helps them overcome the challenges and maximize the positive outcomes. We have data that can help – but even when the data is analyzed, the answers aren’t always obvious. To look beyond the numbers to the human stories, we need a creative and diverse team in which everyone brings their own perspectives and insights to the table. We need to understand our learners, and there’s no better way to do that than to have been in their shoes.
Although only 16% of technical roles at major tech companies are held by women, at Coursera we’re proud to have a data science team that is nearly half female. Building such a team takes conscious effort, and we’ve learned important lessons about sourcing thoughtfully, removing bias in screening and interviewing, treating all employees as equal in the workplace, and growing and empowering each individual in his or her role.
insideBIGDATA: What’s out there on the horizon with respect to the use of data science at Coursera? Any plans for the future?
Emily Glassberg Sands: I can’t say too much but I will say that we’re doubling down our investments in discovery and learning products powered by data. These include data products designed to better match users with the right learning content to meet their goals; to personalize the learning experience with diagnostics, assessments, behavioral interventions, and feedback; and, ultimately, to provide precisely what each of our more than 24 million learners needs to succeed at each moment in their learning journey on Coursera.
Sign up for the free insideBIGDATA newsletter.