FotolEdhar - Fotolia
For companies that want to be successful at data science, building their practice should be like putting together a sports team.
Ever since the value of true data science for business became widely acknowledged a few years ago, enterprises have been looking for data scientists. But the role is a mix of computer science, statistics and business knowledge, and organizations have had a hard time finding workers with that mix of skills. Candidates who have all these skills, often referred to as unicorns, mostly work at big Internet companies like Facebook and Google.
So to get around this problem, enterprises should take a team approach to data science. "Successful teams I've built have come from nontraditional backgrounds," said Michael Ferrari, head of data science at The Weather Company, an Atlanta-based provider of weather data and forecasts that is now a division of IBM. "In this kind of environment I think that that's a strength."
There have been numerous assessments of the current shortage of data scientists. For example, McKinsey and Company projects a shortfall of up to 190,000 workers with deep analytical skills in the U.S. by 2018. At the same time, Glassdoor estimates the average salary for data science positions at $113,436. These factors are making the team approach necessary.
Ferrari was hired recently to build a data science team that leverages the company's user-generated location data to improve ad targeting. He's currently in the process of hiring seven or eight new workers, and he's looking for them to come from a variety of backgrounds. Some should have strong data engineering and computer science skills. Others should have deeper statistical skills. Ferrari is also looking for candidates with business and social science backgrounds.
Ferrari said when he gets resumes from people who say they can do everything related to data science for business at an expert level, he assumes there is some exaggeration. Instead of looking for specific boxes to be checked off, he looks for softer skills, like curiosity, which may be difficult to assess during a typical hiring process, but may be more predictive of success within a data science team.
"In the past, I would look for the perfect data scientist," Ferrari said. "In reality, that person doesn't exist."
What's happening in data science now, with businesses looking to spread out specific tasks to multiple roles rather than stacking them up in one position, is similar to IT developments of the past, said P.K. Agarwal, regional dean at Northeastern University's Silicon Valley campus in San Jose, Calif.
He said basic computer skills used to be the domain of specific roles within companies, but now they are standard for almost every type of position. The same thing is happening with data science today. Agarwal predicted that within three to five years there will be more business analysts who specialize in specific fields, like healthcare or supply chain. These people will come mostly from business backgrounds and pick up statistic skills. At the same time, more tech-oriented business workers will learn the basics of more complex systems like Hadoop and machine learning. Either way, he thinks data science will embed itself within basic job functions rather than remain the exclusive domain of a highly trained few.
"There's just such a high demand for data scientists," Agarwal said. "Statistics is a key driver, but people can pick it up."
Still, this doesn't mean that work will disappear for the highest-trained data scientists, the true unicorns. Agarwal said that the analytics world is continuing to evolve so fast, and with it businesses' analytics needs, that there will always be plenty of demand for them. Even now as the number of people with basic analytics skills increases and automated software picks up more data preparation slack, artificial intelligence and cognitive computing systems are coming on the scene, demanding high levels of expertise at machine learning.
"The evolution of machine learning and artificial intelligence is just going so fast," Agarwal said.
Data science teams require a mix of different skills
Analytics teams don't necessarily need data scientists
Data science jobs remain difficult to fill for some businesses
What is DSaaS?