< Articles & News

5 Tips For Building A Great Data Science Team

January 18, 2015   |   5:26 PM
Teradata Articles

Fulfilling your company’s dreams of capitalizing on data science is a daunting task. If the business press is to be believed, your next multi-billion dollar commercial success hinges on finding that one person who can divine the hidden opportunity lurking in your data.

Well, the myth of the data scientist is true. They exist. They have been documented at Google, Facebook, LinkedIn, and Amazon among other places. Data scientists are rare, but there is hope. Here are some tips for assembling a great data science team.

Tip #1: Stop hunting unicorns

There is a path to data science success that does not involve seeking that one perfect person. Rather than hunting one person with development, mathematics, statistics, and business domain expertise, assemble a team of people with strong foundational skills and a drive for success. Provide them a framework and environment within which they can work together and complement one another.

Begin building your team with data engineers, project managers, domain experts, machine learning experts, and data modelers. Fill some roles from within your ranks; by incorporating existing talent, you’ll have some team members with business and domain knowledge. At Think Big, we rely heavily on hiring top notch people with foundational skills and teach them big data technologies and how to collaborate on high functioning cross-functional teams.

Of course, if you find a great data scientist, by all means hire that person. But make sure you’ve got a team player.

Tip #2: If you build it…

Data science requires more than brilliant minds; you need a solid data infrastructure.

Nothing frustrates a data science effort more than having to wait for tools to become available. Spend some time getting infrastructure in place so that your team can start having an impact right away. Get your Hadoop cluster up and running at the same time you’re recruiting your data science team.

Tip #3: Have a compass

There’s a pervasive belief that data science works by hiring smart people, turning them loose on data, and waiting for great things to happen. The challenge to your data science team is not to boldly wade into your data and find something interesting. Not so; efforts must be aligned with business goals. Form a hypothesis such as “we believe that by looking across multiple channels we can better understand our customers, improve their satisfaction and increase their lifetime value.” If you have established clear objectives and deliverables, it is very clear whether your data science team is delivering results.

This raises a cautionary point. Data scientists may have their favorite tools and techniques. Make sure that they are not so tied to a particular toolset or machine-learning algorithm that they lack flexibility to work on your mission, not their own research interests. A related risk is getting caught up working on secondary or tertiary problems that are interesting, but likely to have little impact on your main goal.

Tip #4: Have a timetable

It is vital in a corporate data science program to translate ideas into real world results. Unlike pure research where publication is the benchmark for success, business demands iteration and delivery.  It’s important to build data science teams with people who are not only very smart, but who excel at getting things done. Assign a project manager to your data science team to track project efforts and deliver at a steady pace.

Tip #5: Learn to spot success

One of the first signs that a data science operation is on the right path is when there is clear collaboration among the business analysts, data engineers, and data scientists. Look for them to be working cross-functionally on projects. If they are busy and their capacity is starting to be strained, that is an indication that good ideas are being tested from multiple angles.

To have the most impact, we find that building strong teams is a much better approach than hunting unicorns. The truth is that data science is a big field, and a cross-functional team is better prepared to handle real world challenges and goals. Hiring smart people who like learning and collaborating with others on interesting problems is the best way we’ve found to create great data science teams.