New master’s degree program in data science through the UW System aims to ease the shortage of professionals trained to harness the power of business data
Story by Rick Berg
The promised payoff for all that data the business world has been collecting remains elusive for most companies – largely because there are relatively few people qualified to pull useful information from that massive collection of data.
A 2015 survey by researchers at M.I.T. found 40 percent of U.S. companies were struggling to find and retain data analytics talent, and Stanford University reported last year that its data science graduates were earning salaries above $200,000 – in large part because of the talent gap. Job-search website Glassdoor.com estimates the average data scientist salary is $113,000 nationally.
To help bridge that talent gap for employers, the University of Wisconsin System launched a master’s degree program in data science a year ago which includes UW Oshkosh and UW Green Bay among the six campuses engaged in the program. The concept for the program grew out of a series of 2014 meetings between educators and business leaders, said Erik Krohn, an associate professor of data science at UW Oshkosh.
“Many of our industry contacts told us that they had data analyst openings but didn’t have qualified applicants to fill them,” Krohn said. “They told us that many of the people filling those positions were partially-trained data scientists. They were hiring people with computer science degrees who had little statistical knowledge or they were hiring statisticians with little programming knowledge.
“These new hires had to learn a new skill with varying levels of success. Instead of having to do on-the-job training for many of these new hires, they were looking to hire someone who had the whole package: a programmer who could do statistical analysis and could present the findings to his or her supervisors.”
Shane Miller, chief information officer at Green Bay-based Prevea Health, was one of those business leaders concerned about the shortage of talented and trained data scientists.
“I’ve found that the only thing exceeding our need to analyze our data is the volume of data itself,” Miller said. “There are lots of people that can query data and generate reports on request. However, you really need someone that can see the possibilities within the data itself and find a meaningful way of telling its story. Without that, we aren’t able to take full advantage of what we have. Finding someone like that, that can utilize the tools, is very difficult.”
The new data science master’s degree program aims to provide such training.
“The data science master’s program will teach you how to harness the power of big data using the latest tools and analytical methods,” Krohn said. “Our program is interdisciplinary and we don’t just focus on one aspect of data science. Our curriculum has courses related to computer science, math, statistics, management and communication.”
“Our students have a wide range of backgrounds. This is one of the reasons we didn’t want to require a specific type of bachelor’s degree to get into our program,” Krohn said. “Any bachelor’s degree is sufficient as long as you meet some of our other prerequisites, such as a programming course and a database course. A good number of our students have a degree in an area you wouldn’t associate with data science. We would not want to exclude a qualified candidate just because he didn’t have a degree in statistics or computer science.”
The art of data science
Despite its name, data science is at least as much art as it is science, which is why human skills are needed to translate mountains of data into useful information. Gaurav Bansal, an associate professor of management information systems at UW Green Bay, said that a defining characteristic for a successful data scientist is curiosity.
“Curiosity is what separates humans from computers,” Bansal said. “The challenge of managing data is being able to slice it and dice it to learn from it, and computers can’t do that. Computers don’t have curiosity.”
Not even all humans are able to display that curiosity, noted Jon Biskner, vice president of information technology at Green Bay-based Nicolet Bank.
“There are a lot of statisticians, a lot of people that work with data, but it’s not just statistics. You need to be able to relate that back to the business,” Biskner said. “You need the curiosity to asks questions about that data, to look for those correlation points and see where that leads. Yes, that’s data science, but it’s also an art.”
“The bigger the dataset is, the more potential there is that it’s hiding solutions to problems you may not even be focused on yet,” said Prevea’s Miller. “Having someone who can interpret the data and visualize it in a way that makes sense is a challenge. At the same time, you need people who can find the hidden correlations and opportunities within the data. They need to find the solutions to problems that haven’t been identified yet.”
Bansal and Krohn said the program was designed to meet that critical need for multi-dimensional talent.
“The people we talked with at the beginning were very explicit about that fact that they were looking for individuals with analytical skills that combined statistics, programming and database management,” Bansal said. “They have to be able to find useful patterns in the data and make business sense of it.”
“Our industry leaders told us early in our designing of this degree that there are just not enough people who have the ‘whole package’ when it comes to data science,” Krohn said. “There are many people trained in one or two areas of data science but there are very few people who are trained in all areas. The interdisciplinary approach to this degree is one of the strengths of our program.”
The program spans academic disciplines including courses on ethics, high performance computing and communications, as examples. Miller said health care organizations like Prevea have a critical need for individuals with cross-discipline skills and vision.
“Being able to understand what your data is telling you is critical,” Miller said. “Whether it’s related to cost cutting, standardizing care, improving quality or root cause analysis, your data can tell you a lot. However, partnering with other organizations with similar data can be a huge opportunity, allowing you to compare across organizations, states, countries or even the world. Clues can be hidden in your data and exposed when comparing it to others.”
Facing the data challenges
Marco Vriens, La Crosse-based chief research officer for Ipsos MarketQuest and author of The Insights Advantage, said organizations like those his company serves face a wide variety of data challenges, including “what data to collect and analyze and how to analyze it, and then getting the firm to act on the data and insights it provides.”
The latter, he said, is often the biggest challenge.
One often-cited story in the world of data science involves the social media site LinkedIn, which in 2006 was growing but still struggling to maximize its growth. One recent addition to the team, a Stanford physics graduate named Jonathan Goldman, found patterns in LinkedIn’s data that he believed could dramatically accelerate its growth. However, The Harvard Business Review noted, “LinkedIn’s engineering team, caught up in the challenges of scaling up the site, seemed uninterested. Some colleagues were openly dismissive of Goldman’s ideas.”
Fortunately, the company’s top leadership gave Goldman the go-ahead to experiment with the data and Goldman ultimately launched LinkedIn’s “People You May Know” segment, which ignited LinkedIn’s growth trajectory skyward.
Miller noted that “healthcare is an industry that tends to change carefully, so it tends to be a bit late to the party in some areas, including analytics. Healthcare creates a massive amount of data and we have just recently learned new ways we can use that data to improve the care our patients receive. Whether it’s looking at how care was provided and the resulting outcomes or how we can increase our capacity to see more patients while still providing the same high level of care, we need people that can see correlations and help our colleagues see them.”
The limits of automation
Much of the data management world touts the emerging technologies – both hardware and software – that help organizations manage their data, and some even suggest this evolution lessen the need for skilled data scientists. Miller doubts that will happen any time soon.
“This sounds like the standard ‘Will the computer or technology make your job obsolete?’ question. It’s difficult to answer in this case,” Miller said. “Looking at some of the cognitive learning capabilities out there, (computers) can already do a lot to identify the hidden links between causes and effects. However, currently we need to tell it what problem we are trying to solve. A good data science analyst will dynamically identify possible uses of the data and then be able to build visualizations that tell the story.”
“I think the process will continue to be automated, but like with data analysis, it is very difficult to fully automate that,” Vriens said. “There are so many subjective choices that need to be made along the way that are context and business problem dependent.”
The promise delivered?
Used properly, Vriens said, data science holds the potential for companies to achieve “fast insights and real-time decision making,” but many remain reluctant.
Biskner said many businesses are reluctant to invest much in data science because they haven’t seen enough evidence that there’s a payoff.
“The technology part is easy,” Biskner said. “We’ve been collecting data for years, but what’s been missing for some people are the successes. People need to see those successes – the use cases that show what can be done.”
“I think we get closer to this all the time,” Miller said. “Like anything, most people stand on the shoulders of people before them and few make their own path and stand on new ground. As more new ground is uncovered, others can follow their lead and better understand how big data can be used to solve big issues.
“Making use of the data takes real focus and partnerships between the CIO/IT department and operations to understand the needs and produce valuable and actionable data. It takes vision to see data as a solution in industries like ours. If needed, take small steps so people can more easily follow you and build up a following before taking those bigger leaps.”
Rick Berg is a freelance writer and editor based in Green Bay.