Anyone who has ever tried to hire a data scientist probably understands the challenges associated with the exercise. Getting the definition of the role right is only the start. Narrowing down a long list of people who believe they fit the bill to the ones you are interested in speaking with could be a monumental exercise. Evaluating capabilities against the needs of the role is gut-driven and very few objective measures are available since the practice is still developing.
Let’s get some alignment on the definition of the role and the skill/competency requirements right. I define data science as the practice of leveraging data and algorithms to come up with business practices that lead to impact. To be successful in this practice, one needs a good combination of understanding of real business challenges/opportunities, an algorithmic mindset, and an ability to work with tools that allow for data to be interpreted. If one had to translate this philosophy into a job description, the ABC of it would roughly translate as:
Algorithmic thinking: Given any problem, the ability to conceptualize a solution that is not just a one-time solution but is algorithmic and can cover for all boundary conditions, and be valid across several instances of the recurrence of the same problem.
Business savvy: Ability to understand the core business problem and the business context in which it manifests. One must be empathetic to business teams and able to explain analytical solutions in a language that business teams can comprehend.
Coder mindset: The ability and willingness to learn any and all tools required to deal with the data and relevant algorithms and leverage them for the most effective solution.
Finding all these skills in one person is like looking for a unicorn! We experimented on an alternate approach that focussed on the creation of a ‘data sciences team’, instead of trying to create a ‘team of data scientists’. This team had the right mix of people, as well as the right culture in the team to deliver what is required by the organization.
Choose a leader who has worked with organizations like yours for long, and has enough experience working across several functions, so that he/she understands the business well, and has political capital that can be useful for cross-functional projects. Personality traits to look for must include the ability to use influence to resolve conflicts, a lifelong learner with deep humility, and respect for knowledge above all else.
Have one area of focus for each commander – algorithmic thinking or business savviness or coder mindset. Don’t look for more than one of these in the same individual, else you may end up compromising on both. Common personality traits to look for would be the pursuit of excellence in their chosen area, ability to get down and dirty, a trade-off between ‘perfect’ and ‘done’, being constantly curious, and leading through demonstration.
Start with high native intelligence, high motivation for working with an above-par peer group, and a bias for learning unrelated subject areas. Instead of hiring soldiers with a few years under their belt with skills in a few specific tools, we experimented with fresh campus graduates with these traits and they turned out just fine.
Of course, this is a design with the right ingredients, but what ensures that the team is effective? Several aspects come into play, like human attributes needed to be successful, how do we hire, how to set goals, measure performance achievement, rules of how the team plays within itself, how they interact with rest of the organization, etc. But these are conversations for another time. For now, suffice it to say, that your organisation’s data sciences division might be better served by a ‘data science team’ instead of a ‘team of data scientists’.
Tapan Rayaguru is COO at Tredence Inc., an analytics services and solutions company.
(Disclaimer: The views and opinions expressed in this article are those of the author and do not necessarily reflect the views of YourStory.)