How to Build a Data Science Team, a Member at a Time


There simply aren’t enough good data scientists to go around. Instead of looking for one person, you can form a diverse data science team.

It’s official: Data scientist is the best job in America for two years in a row. And it’s easy to see why: These positions are in high demand, with a forecasted need for 2.9 million analytics jobs by 2018.

That’s good news for anyone looking to enter college and gain the skills they need to get a job in this field. But it’s tough for businesses: There’s a lack of talent to tap into when filling these positions. Job listings stay open longer than the market average, and poaching from other businesses of top talent is a real threat — like just last fall, when Salesforce took many of its 175 data scientists from LinkedIn.

Does the perfect data scientist exist?

Why is competition so fierce and what can companies do about it? The fact of the matter is, there just aren’t enough good data scientists to go around. The reason it’s hard to find good data scientists is because companies expect these individuals to be experts in many disciplines at once.

[ Related: Data Scientists vs. Data Engineers and Data Analysts ]

Instead of finding the one person who has a full breadth of complex skills, companies can instead get great internal data science assets by forming diverse teams that bring their own strengths to the table. This is how the rest of a company is set up, each team with its own specialization — whether it be sales, marketing or data analysis. Then these groups come together to create a sum greater than its parts and drive business change.

The fact is diverse teams thrive because their heterogeneity breeds an environment where the team doesn’t see success as a given thing. And that, in turn, leads to a team that strives harder for success than homogenous ones. According to a Harvard Business Review article, adding an outsider to a team that doesn’t know the correct answer to a problem doubles the chance of them arriving at the right solution. Even though the team perceives the work to be harder, the outcome is better, and likely these teams work harder because they feel the problem needs the attention.

All that said, a data science team does need to be steeped in many different areas of expertise that cannot be sidestepped when creating this team.

First, these professionals need to have a firm grasp of coding, and they need to stay up to date, so their skills are relevant to the languages that fits the needs of the business. Yesterday that might have been C++, but with the emergence of artificial intelligence and machine learning, a knowledge of Python, R, or Scala might now be more desirable.

[ Related: What Does a Data Scientist Really Do? ]

Next on the skills list is a strong background in probability and statistics. While initial predictive analytics investments may involve boxed machine learning solutions, eventually this team is going to have to make some custom changes to its algorithms. Having people on the team with a strong understanding of statistical modeling will ensure that when the data and problem at hand requires more complex solutions, someone will be able to step in and customize the algorithm for the problem at hand.

In addition, a data science team needs members with the acumen of a scientist to understand and craft an experimental design. These rigorous experiments add validity to any research project, determining the effectiveness of theorems and their probabilistic equivalents. The gold standard in every deductive reasoning discipline from physics to psychology, this skill will allow your team to control and test for whatever business factors are most important at the time. Then, at the end, they can determine that the team’s findings are valid and should be taken as gospel when making a business decision.

Add in domain expertise

On top of all the above, they need someone who has a deep understanding of business and the industry — rivaling someone with an MBA. As data scientists come to more prominence for businesses, their skill sets are displacing the role of the business analyst. Instead of being buried back in a server room, data scientists are now sitting at the table with department heads and the C-suite, providing business critical decisions. They need to have the business acumen — and feel confident in their abilities in the company’s domain — to navigate this aspect of the job.

Getting these skills to all lined up in a single person can feel like an impossible task, trying to hire a unicorn, if you will. However, businesses can stack the cards in their favor by creating a diverse team that has distributed skills in all the necessary areas. Businesses should aim to find a team of people whose skills balance each other out and let collaboration reign supreme in planning out their approach to analytics.

Arvin Hsu

About Arvin Hsu

Arvin Hsu is the senior director of data science and machine learning for GoodData . He has over 15 years of experience in the field of data science and data modeling, including six years building machine learning based data products with both enterprise companies like Disney and startups. He's passionate about the innovations being created at the intersection of big data, machine learning and enterprise data.

Leave a Reply

Your email address will not be published. Required fields are marked *