Building the Data Science team

Search "types of data scientist" and there can be hundreds of millions of results, but most of them begin with articles titled, "The 7 types…" or "Demystified…" or "Four Types…". It's incredibly important to know what kind(s) of data scientists are required to solve problems or answer questions.

But who else is necessary to make data science successful? Are they "traditional" data scientists? Do they need to be on the team?

When it's time to begin hiring data scientists, there are a lot of things to take into consideration, but most of those considerations are questions you've already answered because you're building the data science team the right way.

Building a data science team depends on what types of problems are to be addressed or the questions to be answered.

In this post I'll outline the roles of the non-traditional data scientist, those people who help data science teams be successful. Data science teams are, in my experience, most effective with specialists. Those specialists may excel in a variety of areas that have nothing to do with machine learning, predictive modeling, or statistics. The related skills are often overlooked or left to other parts of the organization, where data science has little to no control, insight, or opportunity to partner.

While data scientists don't tend to be responsible for an organization's data and data structures, they must have an intimate understanding of both. In building a data science team one of the first team members could be described as "knowing where the bodies are buried". This person could be from the data services or database admin team; they are totally committed to knowing what data is captured, where it's stored, how it's organized, its strengths and weaknesses, and how it applies to solving data science problems. Engaging this person or people on the team often includes a response to a question about data this way, "That data? We have that. But did you also think about…?" I think about this role as the data concierge.

Something to chew on: Do you want your CFO explaining the data science team's model output to the rest of the executive team? Do you believe an account person can differentiate the value of data science at a pitch?

Data science should be one of the most communication-driven teams in an organization. Does your data science team produce milestone reports on project progress? Are they presenting findings and recommendations to executives or stakeholders? Do those presentations require PowerPoint or another presentation software package? Who "pitches" data science solutions to prospects? Communications expertise is critical to successful integration of data science into an organization. Data science needs a story teller. That role can be and is often played by the head of data science, but the peripheral presentation development skills can be executed by a data scientist on the team, someone who can turn the often-complex output into an attractive and simple story.

Why can't the data science team just borrow the DBAs and Corporate Communication teams for data and communications? They can. But the specialization of data science work and the complexities of converting it to clear and actionable next steps can require a more intimate understanding of the work. The language of data science can be overwhelming--is it predictive modeling? Machine learning? Deep learning? AI?--so finding that linchpin data scientist with the capabilities to convert and/or simply explain outputs and develop stories is a powerful team addition.

Defining the role of the data scientist should be based more on the problems at hand. You might not want an image recognition, neural net specialist data scientist developing your classification models to determine the probability that existing clients will buy your new product. Or, you might want a natural language processing expert in lieu of an expert in prediction to determine sentiment, theme, or emotion associated with open-text website feedback. While data scientists can be extremely flexible they're becoming more specialized. This is why it's critical for organizations to understand the breadth and depth of the problems to solve, and just as importantly where those problems are in the organization.

As a final tease: do you consider geospatial data and leveraging that data as data science? Are geospatial experts data scientists? Do they have a place in a digital world? Would you include geospatial experts on your data science team?

I'll leave the "types of data scientist" to the millions of search results and finish with this: Building data science without a data concierge or story teller is one way to cripple the integration of data science into an organization. If you want to know more about the multiple roles of a data science team reach out to me at sam.johnson@bluejacketsol.com. And, if you want to share your success stories on the types of teams you've had experience with or built, I'd appreciate hearing more.