How to find the right data scientists
With rapid advances in artificial intelligence and machine learning, demand for hiring data scientists has never been higher.
Dubbed the "sexiest job of the 21st century" by Harvard Business Review, data science is one of the hottest jobs in existence today. For good reasons. It's high paying, known for great work-life balance, and in constant demand across organizations. In fact, last August, LinkedIn reported there were more than 151,000 unfilled data scientist positions in the United States, with particularly acute shortages in New York, San Francisco, and Los Angeles.
Marketing departments are looking for data scientists to deliver more differentiated and memorable customer experiences. Finance teams covet their ability to provide deeper understanding of risk and markets. Operations officers want them to help streamline processes and improve efficiency. And the U.S. intelligence community is after them to help organize and interpret data in ways that enable better strategic decision-making.
That high demand means companies struggle to find enough data scientists to fill their open job requisitions. It also makes it harder to match individuals with the right skill set and experiences to address their specific requirements.
These guidelines can help companies find the best and brightest data scientists on the market.
Understand the differences among data scientists
One of the biggest mistakes in seeking data scientists is that the organization fails to clearly identify what the term means to them in the context of a particular job. The role has changed dramatically over the past several decades.
Differences sometimes create confusion in the hiring process.
"The 'data scientist' title means different things to different people," says Meta S. Brown, business analytics consultant and author of "Data Mining for Dummies." "People often want to hire data scientists but can't articulate good reasons for doing so. It's not uncommon, for example, for job descriptions to include lists of skills that don't exist in any single human being. You have to be realistic. Don't expect any one person to fill every business analytics need."
Conversely, the hiring manager doesn't always need every available skill. "Searches may be inappropriately influenced by individual preferences of current staff," Brown adds. "The wants aren't realistically matched to what the business needs."
To avoid this trap, spend time clarifying why you need a data scientist (or if you really do). Then focus on the business needs and skills that can address them. Only then is it a matter of finding a data scientist with the right qualifications.
Unfortunately, there aren't any universally agreed-upon data scientist categories, so matching a particular type of professional to a specific need can be problematic. But data scientists do focus on certain types of deliverables, and that's a fair place to begin searching.
Yael Garten, director of Siri data science and engineering at Apple, writes that one type of data scientist creates output for humans to consume in the form of product strategy recommendations. She calls these "decision scientists." Another group creates output for machines to consume, such as models, training data, and algorithms. Garten calls these "modeling scientists."
More specifically, Miryung Kim, an associate professor in the computer science department at the University of California, Los Angeles, polled nearly 800 software industry data scientists and identified several common categories, including:
- Data preparers: Individuals who spend the bulk of their time querying and preparing data for analysis. They are more likely to work with SQL than machine learning algorithms.
- Data shapers: Professionals who have the skills of preparers but also delve into machine learning. As such, they are less likely to work with SQL or structured data.
- Data analyzers: Data scientists who apply classical statistics, math, and data manipulations to spot trends that could advise business and marketing strategies.
- Platform builders: Employees who build platforms and instrument code for the purpose of collecting data. They are more likely to work with distributed systems, like Hadoop.
- Data evangelists: Data scientists who spend considerable time engaging with others, such as business leads and product management, about how they can put good data to use.
- Insight actors: Data science types who spend their time acting on and disseminating data-driven insights.
- Moonlighters: Any engineer, marketer, or product manager who dabbles in data science when the need or inspiration arises.
- Polymaths: Jacks-of-all-trades who spend their time doing various data-oriented tasks, such as building platforms to gather data, analyzing it, or acting upon it.
An additional category to know about is the citizen data scientist. Gartner, which coined the term, defines a citizen data scientist as someone who creates or generates models that use advanced diagnostic analytics or predictive and prescriptive capabilities but whose primary job function is outside the field of statistics and analytics. They are power users who can perform both simple and moderately sophisticated analytical tasks that would previously have required more expertise. View citizen data scientists as a potential complement to on-staff expert data scientists rather than someone you'd hire to address broad data science requirements.
"It is important when trying to find the right candidate to know which category of data scientist you really need because one size does not fit all," Kim says. "If you end up with the wrong type of data scientist, you could find your efforts to use and analyze data will be dead on arrival."
Explore nontraditional sources
With so much competition, your company may need to look beyond the obvious recruitment sites to find qualified data scientists.
One place to start is with your job requisition. As with security professionals, the best candidates often follow a nontraditional path. "Organizations grappling with the skills gap and the shortage of quality data scientists should consider nontraditional sourcing, focusing on skills and potential rather than degrees," says Jared Goralnick, head of the LinkedIn Recruiter product. "When we say nontraditional talent, we mean candidates without a four-year degree, elite credentials, or spotless work histories." Explore alternative talent pools such as accelerated training programs or boot camps, local community colleges, and organizations. If your company cares about investments in data science, send someone to make the rounds at key industry events, including the Strata Data Conferences, TDWI Chicago Conference, and Forrester Data Strategies and Insights Forum.
Andrew Flowers, an economist at Indeed.com, notes that companies may also want to look beyond universities, which are "behind the curve" on training students for jobs in data science. He suggests also considering hiring individuals directly from respected for-profit training programs, such as DataCamp, Springboard, and General Assembly. "These companies are not accredited universities," he says, "but they can adapt more quickly to train individuals in the latest data science skills."
Author Brown agrees, comparing the current search for data scientists to the early days of the U.S. space program.
"Programmers were vitally important to meeting the goals of the space program, but there were no experienced programmers to hire. Schools did not teach required skills, and almost nobody had work experience in computer programming," Brown says. "People with no programming experience quickly learned to write software that safely got people into outer space and back using only computers that had a tiny fraction of the power of the humblest modern smartphone. So recruiters should never think it's impossible to find the talent they need. They can help develop it."
In a competitive labor market, companies have to get more creative to find the right data scientists.
Citadel, an investment company, answered the challenge by sponsoring a series of datathon competitions in universities across the nation. The Data Open competitions invited thousands of students to work in teams on large and complex datasets and then present their findings to judges. Winning teams took home $25,000 cash prizes and opportunities to interview at Citadel. The company's recruiting team, meantime, created a unique opportunity to see the skills and potential of future data scientists in action—and maybe even hire a few of them.
Companies can also work with local schools and universities to forge or support data science disciplines. That benefits students in the community as well as their own employees. At least two dozen schools already offer data science master's programs, and the number is likely to keep growing.
Retrain existing staff
While it's always sensible to scout for fresh talent, sometimes the best candidates already work for you. Even if they are new to this form of data analytics, current employees who are familiar with the business and its goals have a leg up on other candidates
Consider training promising employees to help them grow into exceptional data scientists. "Our research shows that the most common previous job titles that current data scientists held five years ago were research assistant and researcher," Goralnick says. "Looking for people currently employed as researchers could be a great place to source that kind of potential."
In addition, organizations can tap into the growing movement to create data scientist-focused industry certifications, which is another way to help train employees. Earlier this year, for instance, The Open Group, a vendor-neutral technology standards and certifications body, announced a new data scientist certification program. Individuals can certify their skills and experience directly through the program, or organizations can adopt The Open Group Standards and build an internal certification process, the consortium says. Individuals are reviewed by peers to ensure they meet the necessary standards and will sit under one of three skill levels and experience.
Make no mistake about it: Finding qualified data scientists is not easy. But with solid planning and an understanding of exactly what you need from your data scientist hires—along with a willingness to explore multiple traditional and nontraditional candidate sources, creativity, and openness to retraining existing staff—companies can find or create the right data scientists to meet their needs.
Hiring data scientists: Lessons for leaders
- Clearly identify the type of data scientist you need. Different areas of need may require different sets of skills and knowledge.
- Don't discount skill-specific training programs.
- Make use of available resources, both with interested staff and standards-driven education.
This article/content was written by the individual writer identified and does not necessarily reflect the view of Hewlett Packard Enterprise Company.