Best practices for building a data science dream team

Register now

In a world that is drowning in data, organizations of all kinds are looking for a lifeline— someone to make sense of it all. That’s why the job outlook is bright for people who have data science skills, with computer and information technology occupations expected to add more than 546,000 jobs by 2028.

But an evolving environment for data scientists means they must fine-tune their skills to make solid contributions to a data science team. Equally important, and also evolving, are the strategies managers need to recruit and develop people to staff those teams. Stellar tech skills are not enough. An effective data science team must include people with overlapping skills, all with a deep understanding of a project’s overall vision.

An expanding market

Thanks to global trends such as the internet of things—leading to “smart cities” and digital transformation in several industries—the analysis and use of data now permeate almost every aspect of life. Even more traditional workplaces such as manufacturing, insurance, education and health care are delving into data to sell their products, serve their customers and meet corporate goals.

My own experience in the oil and gas industry is a good example. In the last five years, we’ve increasingly used data science tools in every aspect of business, from pinpointing underground oil sources to updating engineering plans, monitoring pipeline flow, predicting maintenance needs and forecasting market dynamics.

In the broader world, an increasing reliance on data affects the ads people see on social media, the way their health insurer contacts them and the time their street lights come on. Not every city is equally “smart,” of course, but it’s an aspiration, even in developing countries.

A data science team may include people in a dozen specialty areas, such as computer programmers, data or applications architects, analysts, statisticians and several types of engineers. However, the team leadership can be broken down into three basic roles:

  • Business analyst. This person’s interaction with the end user of the data will guide the entire process. Among other things, they must ask the right questions. What is the ultimate goal of the data collection and analysis? Is it realistic? How will data be used and/or monetized?
  • Data scientist. This person works with others to build a workable model for the project. As patterns emerge from data analytics, the model should help clear a path to the end user’s goals—for example, suggesting how to adjust prices, motivate behavior changes, improve customer service or make a process more efficient.


  • Data visualization engineer. No one wants to make an important decision based on thousands of numbers crammed onto a spreadsheet. Data visualization engineers work with others to make the numbers understandable, using tools like Tableau, Chartio and Google Analytics. Visuals such as charts, imagery, graphs and animations help non-tech people make sense of it all.

Building skills, building teams

Building a data science team is different from building other types of business teams. The unique challenges include incorporating emerging technologies as well as the varied skills required by the various players. Some tips for effective team building are:

  • Hire a good analytics manager. The three roles mentioned need coordination, which means someone to drive the execution of the project. A common misconception is that the roles can work in silos once a problem is defined, and then simply pass on their work to the next silo. A good manager validates the business cases and return on investment, shields the team from poorly conceived plans, and acts an interface between the data science team and the rest of the organization.
  • Hire from within where applicable. In large organizations, knowledge about the company, its systems and strategies are very valuable. Even people without extensive data science experience should be considered as possible business and technical assets, because of this shared vision. With a little mentoring and cross training, they could fill key roles in the team.
  • Build a self-sufficient team. The team needs to have autonomy in taking use cases from any department within a company and providing end-to-end solutions. This independence helps eliminate hidden agendas and biases and keep the data science team on course toward the project goals.

It takes more than technical acumen to develop a successful data science team. It is important to pull in people who understand the vision as well as leadership to keep everyone on track. Data thus becomes more than just numbers, formulas and spreadsheets, but a valuable tool to succeed in a global economy.

For reprint and licensing requests for this article, click here.