In the data scientist’s world, data is everything. In the business world, however, data is only the beginning. For the 200 attendees that showed up to DataScience.com’s second annual DataScience: Elevate conference in San Francisco last Thursday, this dichotomy would eventually be an opportunity.
The crowd at the Marquis Marriott was a mix of data practitioners, data science managers, and executives interested in bridging the gap between C-level discussions and on-the-ground work. Equally diverse were the companies represented in the crowd, with professionals from Evernote, T-Mobile, Reddit, and Oracle all in attendance.
By the end of the full-day event, everyone would see that engaging beyond the dataset is not only ideal for a data scientist, but necessary.
True digital transformation requires more than even the best team of data scientists. It involves the support of IT, process engineers, and C-level executives, all of whom speak in their own languages and think in their own lines of logic. Throughout the Elevate sessions, speakers from Forrester Research, Salesforce, Levi Strauss & Co., and more, challenged data scientists to take a broader look at their role and rethink their typical approach in delivering data science success at an enterprise company. Rather than just focusing on the data at hand, they stressed the importance of the data scientist considering the company at large.
This shift in thinking changes the way a data scientist even begins building a model. It requires that he or she takes a step back and thinks about the question the model needs to answer. Steve Carter, data science manager at Facebook, in the event’s opening panel, “Data Science Workflows That Lead to Business Results,” referred to this process as figuring out the “people problem.” Kim Larsen, director of marketing analytics at Uber, in his individual session, “Time-Series Analysis for Marketing and Beyond,” called it the “mental model.” While the specifics may look different for every company, the most defining question should always be a business one.
While this isn’t necessarily an innate skills for data scientists, it is something they — like it or not — have to master in order to get their findings to the C-level. One way to ease the burden of this seemingly daunting task, fellow panelist and Director of Data Science at DataScience.com Andrea Trevino suggests, is “making business leaders part of the learning process.”
Similarly, Claire Lebarz, data science manager at Airbnb, recommends an “embedded model” when it comes to structuring a data science team because then “they are able to build trust and have other functions understand what data scientists do.” This will start the process of democratizing data and having the rest of the company speak the data language.
Another tip from the speakers and panelists on how to navigate these struggles is to start small. As Kristen Burton, director of digital community engagement at Cisco System put it, “The way things are going now, we need to deliver ongoing value in bite-sized chunks.” Barbara Murrer, insights lead at Levi Strauss & Co. also suggested to “start small scale first to get people comfortable with the approach” in her “Embedding Insights - The Journey” session. She referred to the format in which data scientists should present their findings to business stakeholders as an “insights mullet” — simplicity in the front, sophistication in the back. After all, she subtly reminded the crowd, “We love data; they don’t.”
Networking and Reflections
Many of the topics covered during the sessions spilled over into the cocktail networking hour. Over a glass or two of wine, industry professionals exchanged their personal experiences of delivering business outcomes and working with business stakeholders.
“What customers don’t realize is that data science is a discovery process and you cannot guarantee the outcomes until you actually research the data,” said Sergey Patsko, engagement leader of data science services at GE Digital. “Once you educate them to the point where they understand that, the majority of them will not be willing to invest money because you cannot guarantee the ROI of this investment.”
“I liked the speaker from Uber because he focused his session on jumping out of the data and thinking about the question you’re trying to solve. Data is just a tool, and the actual question is a business question,” said Lin Li, senior business intelligence analyst and data scientist at LQ Digital.
Indeed, in order to help a business become truly data-driven, it takes a lot of strategy and persuasion. In this sense, data science is not so much a science, but an art.