Data Science Director Kelsey Kincaid describes her journey building a data science practice
In 2012 Thomas H. Davenport and DJ Patil named “Data Scientist” the sexiest job of the 21st century. These same authors followed up with another article in 2022 that pondered whether this was still true. They concluded that “…the job of data scientist will only continue to grow in its importance in the business landscape.”
They note that the number of postings for data scientists on Indeed has more than doubled between 2013-19, and the US Bureau of Labor Statistics predicts that data science will grow more than nearly any other field this decade. Businesses almost unilaterally want data science, but the high demand, varying definitions of what data science means, and the ever-evolving nature of the field make it difficult to deliver.
Data Science is one of the fastest growing teams at Evolytics due to client demand. Our experience building out this team of award-winning data scientists, who work on cutting-edge consulting projects, gives us a unique perspective on how to build a successful data science practice. It is never as simple as bulk hiring and can fail without a thoughtful approach. Here is our best advice on building out a data science team at your organization.
Evaluate the Current State of Data in Your Business
Data Science cannot exist in a data vacuum. Before building a data science discipline, you should have a foundation in data management, data engineering, and business intelligence. Your business should already prioritize making strategic decisions based on data. Without this, any data science discipline will immediately fail. Good data science requires a lot of data, a team of people that know data forward and backward, and leadership that is excited to move the needle with data. At Evolytics, we use the Data Lifecycle to visualize the key steps and disciplines required to develop a data-driven organization.
- Align: How do we define success? How will we measure it? What’s our overall data strategy to get there?
- Acquire: How are we getting the data? This includes call tracking, tag management, CDPs, etc.
- Aggregate: This stage is typically led by the data engineering team who cleans and transforms the data for analysis.
- Access: How are decision-makers seeing the data? This is typically accomplished through dashboard builds, but also includes ad hoc data pulls.
- Analyze: Once we have the data, how do we use it to inform our business decisions? Data Science really begins here.
- Act: What decisions can we make based on the data to optimize our business? Most frequently, this is experimentation, personalization, and machine learning.
Search for Internal Talent
Since you already have a culture of analytics, you almost certainly have talent that is ready and interested in data science techniques. Evaluate the skills and interests of your current team, and identify individuals that can move the needle. Don’t focus only on hard skills like programming, but also evaluate business knowledge, communication skills, and cross-team relationships. Without these skills, scaling data science will be an uphill battle. Technical skills are also more quickly teachable than softer skills that are built with time and experience.
Supplement with External Resources
Once you internally identify a few key individuals to drive your new data science discipline forward, you must also search for external talent. I firmly believe that you cannot succeed by hiring 100% externally, but I also firmly believe that you cannot succeed with 100% internal talent. The discipline is most likely to succeed by pairing institutional knowledge with industry experience.
You should also identify gaps in your internal team. A successful discipline will include a good mix of those with technical skills, strategic vision, and the softer skills mentioned above. It’s impossible to hire and maintain a team of unicorns, so it’s likely your team will consist of members with a variety of specialties: statisticians, engineers, hackers, business analysts, and experts in other relevant areas.
Once you understand where your team does not index highly, that’s when you evaluate what to recruit from outside your organization. Fill these gaps by hiring for specific roles, engage Evolytics, or both. I recommend both. What you gain through external hiring provides embedded resources that build relationships and sell the vision of data science with other stakeholders in your organization. What you gain from engaging with Evolytics is a diverse range of talent capable of filling gaps on your team as they come up, experts with visibility into the latest best practices, and a broad range of industry knowledge.
We have seen our clients find repeated success by combining internal business resources, new hires, and partnering with our Evolytics data science team.
Build a Common Language & Ensure Continuing Education
Your statisticians, engineers, hackers, and business analysts are likely specialists in something, but will constantly need to broaden their knowledge and skill set. The goal is to develop a team of T-shaped data scientists (otherwise known as generalized specialists) who demonstrate unique expertise and are well-versed in other areas. A leader should help establish which skills and knowledge are critical for each person to possess, then put a system in place to fill gaps. This breadth will give each person a solid foundation while also revealing how each person’s depth adds value to the team.
Another way to reveal the value each person adds to the team is to establish a common language. You can do this using one of the many tools available to help understand the unique strengths and personalities of your employees. We use StrengthsFinder at Evolytics to help build dynamic teams that complement one another. This system allows us to understand one another quickly when making assessments such as, “This project is ambiguous and will require a lot of Ideation, but will need an Achiever to take that plan and put it into action.” Whether you choose Gallup StrengthsFinder or something else, we strongly believe that teams succeed best when they have systems like these in place that highlight each person’s superpower.
Another key component of team building is developing best practices for the work the team produces. Data Science is best when performed as a team sport, so peer review is critical. Code reviews, pair programming, and validating statistical rigor are all critical pieces of the peer review process. Brainstorming sessions should also be part of the day-to-day, as dynamic teams benefit from multiple perspectives. Another key to success is providing your team with opportunities to stay up to date with industry trends, advances in programming, and the goals of your business. The analytics industry constantly evolves, and you cannot risk falling behind. Leverage the unique talents of each team member to host continuing education sessions that elevate your team.
You’re now ready to hit the ground running. Your team should take a multipronged approach to execute the company’s vision. They will need to evaluate your technical architecture to ensure it can support data science while also demonstrating quick wins. This will set you up for scale while showing your executive team the tangible results of their investment in data science. Ask your team to identify a few relatively simple analyses they can do to start generating new insights right away. I recommend starting with market basket analyses or segmentation. Decision trees are also a powerful way to reveal complex insights to your leadership in easy-to-understand ways. As the team rakes in the wins, they should document technical gaps and work with your IT organization and/or external partners to evaluate ways to close them. Zero in on your Data Science team’s goals and requirements, and use this to choose the tech stack that is best for your organization—both financially and functionally.
While your team and business grows, you should revisit these tasks on a frequent basis. For the Evolytics Data Science Team, that means using the Data Lifecycle, not a Data Ladder. Lastly, it’s critical that you constantly advance your organization’s use of data and keep developing a stronger team.
A great article on the topic I found valuable:
“Companies are struggling to hire true data scientists — the ones trained and experienced enough to work on complex and difficult problems that might have never been solved before. And hiring them becomes far more difficult if the company isn’t the biggest brand or the biggest name. Finding and retaining IT workers, in general, has been difficult for a while now. It’s exponentially harder when it comes to data scientists.”