Data scientists are much in demand. Beyond the domains one might expect — technology, the internet and telecommunications — they are being sought in energy, financial services, manufacturing, healthcare, pharmaceuticals, and other industries, according to recruiting firm Smith Hanley Associates.

But there’s a gender gap. Only 15% to 22% of today’s data science professionals are women, according to recent research from Boston Consulting Group. Moreover, women data analysts tend not to hold managerial roles, comprising only 18% of leadership positions at premier tech companies, according to Forbes. Data science’s appeal is lackluster among female STEM students: In the BCG report, nearly half of them perceived data science to be “overly theoretical and low impact.”

Academic institutions are working to change that. Recently, the University of Pennsylvania hosted its first-annual Women in Data Science conference — WiDS Philadelphia@Penn — to coincide with the annual Global WiDS Conference held at Stanford, and similar events held at approximately 150 colleges and organizations worldwide. The goal of the Penn event, according to the conference organizers — Susan Davidson, computer and information science professor at Penn Engineering; Mary Purk, executive director of Wharton Customer Analytics; and Linda Zhao, Wharton statistics professor — was to “inspire and educate data scientists, regardless of gender, and to support women in the field.” The Penn event was hosted by Analytics at Wharton, Wharton Customer Analytics, Penn Engineering and Wharton Statistics.

An industry panel featured three data analytics leaders from major companies who spoke about their experiences, where the field is headed, and the more general obstacles to working successfully in an area still dauntingly opaque to many businesspeople.

Purk moderated the panel and raised the question of how data scientists in general can play a larger role in terms of a company’s overall goals. To begin with, she asked the panelists: How do you promote business acumen in your teams?

Nicola Blue, vice president at American Express Insights Global Consumer, an analytics and consulting organization within Amex, noted that it’s critical to get her team to work toward actionable results. She called those results the “so what” of any project. “If we’re not focused on the larger environment, we could have the most amazing information in the data, but it won’t have relevance and impact.”

“If we’re not focused on the larger environment, we could have the most amazing information in the data, but it won’t have relevance and impact.” –Nicola Blue

Blue said she coaches her team on being empathetic and understanding the other person’s business needs. “That’s how you get invited back [to meetings and discussions] — showing that you know how to put your expertise into practice.”

Rashmi Patil, solutions expert at McKinsey & Company who previously led initiatives at Amazon and Accenture, said it’s important for data analysts to learn the business context they’re working in so they can think holistically. She stated that it would help them build more sophisticated models that would more closely reflect reality and ultimately yield better forecasting and analysis.

Recognizing that all panelists had a business degree, Ginny Too, executive director of customer and marketing insights at Comcast who has also worked at McKinsey, said that in addition to taking some business coursework, budding data scientists should consider regularly reading the business press — even articles not directly of interest to them — to develop a sense of the real-world business community and its concerns. “Read about airlines. Read about why Coke is launching this or that product. Understand KPIs. I think that kind of content — having it in the back of your mind as you’re building out data models — will really be helpful.”

Building Influence

Ideally, a leader in data science would not only contribute to the business’s goals but become an influencer, Purk noted. She asked the panelists to offer advice on achieving influencer status in their organizations.

Too talked about how to interact with colleagues in a way that keeps the data analytics department from being reduced to a mere supplier of numbers. She recounted how she recently got an email asking for a single statistic — ‘How many of X product did we sell last month?’ — to fit into a PowerPoint slide. Too’s reaction was, “Can we talk? Let’s pick up the phone. What are you actually looking for?” It turned out there was a much larger business problem behind the question. Too said that while being helpful is important, it’s also important to “have the wisdom in those moments to … open the aperture to say, ‘What is the problem, and can we solve it together?’” Taking such action also helps give data analytics a seat at the table, she said.

Picking up the influencer theme, Blue talked about going to meetings with the idea of “active listening … to the body language, to what’s being said and not being said, to the tone of the room.” Thinking about those factors can improve how others perceive you and how effectively you can contribute, she noted.

In addition to one’s behavior at meetings, initiating follow-up conversations or sending colleagues relevant articles or analysis helps build one’s reputation as a thoughtful partner who is not just on the execution side of things but has a strategic point of view, Blue said.

During a question-and-answer session, an audience member asked if the panelists thought data analysts should have to justify their models and results to others in the business. She described a recent occurrence at her own company: After delivering a report, she was asked to explain “what’s really going on” and why the model was reliable. “Do you say, ‘OK, let’s break it down for you,’ or, ‘you just have to trust us?’” she asked.

The panelists agreed that in a situation like that it’s best to assume a positive intention and to take it as an opportunity to educate and inform. First, try to understand why the person is asking the question, said Too. “Is it because they’re going to have to be able to explain it to their boss? Or is there something kind of counter-intuitive in your results?”

“There’s so much we now know about our consumers, their journeys and how they interact with us.” –Ginny Too

Blue added, “Having a dialogue is a good thing. Take it as they’re being curious and interested.”

Patil noted that in explaining one’s data model, it’s useful to be upfront about its limitations — for example, to say, “The accuracy one month out is ‘so much,’ but it degrades to ‘this much’ two months out.” This will help those who will be using the data appreciate that it’s a model and can’t cover all cases, she pointed out.

Looking to the Future

Commenting on the future of data analytics, all the panelists noted increasing interest by consumer-facing businesses. “There’s so much we now know about our consumers, their journeys, and how they interact with us,” said Too. “I think consumer-facing businesses will continue to get sharper about this and require more data scientists.”

Another area of growth involves “bringing others into the fold,” according to Blue. She sees an increasing curiosity and desire for understanding about data science among marketers and product managers, partly because of the growing accessibility of technology visualization tools. “There’s more awareness and less intimidation … so people are curious about it, which is exciting to see.”

Blue outlined a scenario in which data scientist leaders would educate their colleagues “so they can appreciate and speak the language,” giving them the confidence to work with data, albeit in a limited way. Companies would still have a center of data excellence or expertise, but not “ring-fence” it.

Reflecting on how the relatively young field has changed over time, Too observed that when she began her data analytics career in the early 2000s, she was using simple Excel spreadsheets. That’s impossible today, she said, because of the scale, granularity and complexity of data that is available. What hasn’t changed, though, are the problems data scientists are asked to solve. “There’s always going to be a focus on how we frame up the question, the business purpose and the level of sophistication with which we can answer it.”