Demystifying the Role of Today’s Data Scientist

By Mark Stone, Contributor

Earlier this year, a Glassdoor study identified the data scientist role as the best job in America, offering the highest median salary of all careers.

In today’s business climate, however, mastering machine learning, AI, and perfecting algorithms may not be enough. Data scientists are being called upon to more effectively communicate their findings to customers, their teams, and the C-suite. In other words, the art of storytelling is becoming an essential skill for this subset of scientists, and continued education courses are even offered to address the gaps.

Unfortunately, many companies are failing to achieve a clear communication channel between data scientists and executives, causing friction for both parties — and throughout the organization. Equally concerning, a Stack Overflow study, based on surveys with 64,000 developers, found that next to machine learning specialists, more data scientists are looking for a new job compared to other professionals.

As executive leadership considers what it means to invest in data scientists, it’s first important to identify what, exactly, a data scientist does. Moreover, if the work is so lucrative, many business leaders are eager to know: How can they help shape today’s loosely-defined data scientist role so that it provides the most value to the organization in this time of massive data collection?

A Data Scientist, in Broad Strokes

Jennifer Hobbs, senior data scientist for STATS, a sports data company, explained that some companies use the data scientist job title as a catch-all. In reality, the roles are more granular, such as data analyst, data engineer, or machine language engineer.

“Data science is an overly broad term that can mean very different things,” said Hobbs. “One of the things I tell people looking for jobs as data scientists is that you need to really read the job description and see what tools they’re looking for — it can vary widely.”

Still, many companies expect data scientists to have a range of similar skill sets. According to Hobbs, the data scientist’s responsibility is to use data to make sense of wide sets of information, to make recommendations, and build models to identify and predict business outcomes and behavior. “Somebody comes to me with a problem, “she explained, “and I can translate it into something that can be solved with math and build a model around it.”

It’s in this model-building element where storytelling comes into play. However, telling a story without connecting it to a desired outcome, she emphasized, won’t cut it. “Storytelling around analytics, to most people, isn’t going to be interesting — they want to know why it’s important,” she said. “It’s crucial that you communicate the importance and the value of the models you’re building.”

For Hobbs, better stories require better data and analysis. Then, it’s a matter of painting a picture. Like Hobbs, Rory Armes, CEO of Eight Solutions, works with and hires data scientists, but he is skeptical about their storytelling abilities. “The guy running our data science group is just brilliant — he is smart, very articulate, and clearly describes to me what data science is and what’s happening,” he described. “But he loses me in about five minutes.”

To combat this misfiring in communication, Armes suggested data scientists bring graphic visuals alive with easily digestible information. When this fails, he recommended adding a layer between the data scientist and the C-suite. At Eight Solutions, leaders also rely on software to bridge that gap.

Armes described a tool that acts like a data connector, taking in data from a source and decoding it into a form that is more readable for non-data scientists. “It’s like a Google Translator for raw data,” he said. “It gives us value quicker and is a good translator between the stats people and the C-Suite.”

And while the field is still in its infancy, there is reason to believe investment in data scientists will only continue to grow. “We realize [data scientists] are a different animal with very specific skill sets,” Armes said. “Many won’t have a true data science background, but you ultimately have to hire programmers to get the data models created.” By broadening his search criteria to include programmers (especially those from reputable companies) he’s been able to hire more quality candidates.

For Hobbs, smarter investment begins with developing more accurate job descriptions that are in line with data scientists’ daily responsibilities. “Employees will start to figure out what they like,” she said. “So, you’ll see more specific listings; people will find their niche that way.” With this recipe, she predicted, there will be smoother implementation and satisfaction for both the employer and data scientist employee.

But First, Data Strategy

Hobbs recommends companies avoid the mistake of thinking data scientists are the “cherry on top.” Instead, companies need to expand their recruiting efforts to focus on functions that span their data strategy, including data management, deployment, development and operations (DevOps), permissioning, monitoring, and interface building.

“You need to be clear about your data model building component — how you distribute it, to whom you distribute it, and the types of products you want to put out,” she said. In her estimation, about 80 percent of data science projects fail because of the high degree of difficulty in aligning data strategy to business strategy. A Gartner study pegs the failure rate at closer to 85 percent.

“The key is to think about production first,” Hobbs said. Companies should be asking questions, such as, when does this go live and how will we deploy it? Where will the data come from? What unexpected challenges might pop up? How will users interact with it? How will we monitor model performance? And, what resources do we need to build this?

Hobbs singled out Netflix as a company focused on building tools that enable their data scientists to work more efficiently: “I think the companies that are successful are all about integrating data science throughout their entire technology stack.”

Yet data science is only one part of the big data picture. Data engineers, programmers, and machine learning engineers are also in high demand, and the whole team — from the top down — must be clear on how organizations can incorporate data science across departments and throughout its company culture.

To retain the best people and enrich the work environment, Hobbs recommended businesses help data scientists develop unique and specialized career paths beyond the organization. Yet, for Hobbs, it doesn’t matter how good a company’s data science team is or how strong their algorithms are if their findings not being shared with the world.

“If you can’t deploy [the story of the data] into the world,” she said, “it’s not making you any money.”