Want to find the next great data science guru? Look for flexibility.
As demand for data scientists far outstrips the pool of qualified applicants, employers must look elsewhere – to individuals with no experience but high potential. The question: what defines “potential”?
Now, an emerging body of work is helping us understand the right combination of knowledge, skills and personality traits needed to become an outstanding data scientist. The first-ever, scientifically developed competency model and personality assessment for data scientists is changing how we recruit for potential in this high-growth area.
The key finding: personality and “intangible” attributes play a major role in the success of a data scientist – greater than most realize.
In talent-starved markets, employers turn to competency models and talent assessments that identify those who could be trained up to fill open positions – even if they lack directly relevant experience. Yet in the still-young field of data science, scientifically valid models did not exist: there wasn’t a large enough population of data scientists to generate a valid sample.
To overcome the lack of both proven candidates and a way to identify unproven, “high potential” talent, we decided to create a data scientist assessment tool. We used our own team of 500+ data scientists – among the world’s largest – as our sample. And, we partnered with Hogan Assessment Systems to extend their job analysis and competency model research into the area of data science.
First, we began with the “classic approach” – asking our data scientists to rate the importance of 400 skills, knowledge, abilities and personality traits. We found clear trends in four areas:
- Technical – Key competencies include statistical modeling and machine learning.
- Data science consulting – Key competencies include business acumen, communications, and teamwork.
- Cognitive – Key competencies include critical thinking and inductive and deductive reasoning.
- Personality – Key competencies include inquisitiveness, perseverance and ambiguity tolerance.
In the absence of a supply of technically skilled candidates, personality attributes play an even greater role in identifying those with potential. Yet it can be difficult to determine these traits, even through unstructured interviews. So, we also developed what may be the first statistically valid personality assessment for data scientists.
By combining the two efforts – a detailed survey of our own data science experts and a new personality assessment – we honed in on what it takes to thrive as a data scientist.
For example: the single most important attribute, our work revealed, is flexibility in overcoming setbacks, the willingness to abandon one idea and try a new approach.
Often, data science is a series of dead ends before – at last – the way forward is identified. It requires a unique set of personality attributes to succeed in such an environment. Technical skills can be developed over time: the ability to be flexible – and patient, persistent – cannot.
The lessons we learned along the way – and immediate impact we have seen – helped us see the data science recruiting process in a more informed, effective way. And while they reflect our own experience, these lessons apply to any organization looking to expand its ranks of data scientists.
- First, it’s not enough to “know” what it what it takes to be a successful data scientist: you have to be able to measure those traits, to quantify them. It’s far too difficult – and high-stakes – to go with an ill-defined or one dimensional definition of data scientist attributes.
- Second, the tangible accomplishments that define the textbook data scientist may not reflect the real reasons for their success. Put another way, our assessment went beyond education, training and a track record to ask, “Why?”
- Third, it can be done. In the absence of a large pool of fully trained, credentialed candidates, we found that one can identify current employees who have the ability to learn “the art of data science,” and will excel at it, given the chance. It also helps pinpoint the right training for current data scientists, and place a premium on the key skills as part of your appraisal process.
- Finally, the process lent confidence. Measuring intangibles and multiple success criteria – for a critical hiring need – requires a high level of statistical confidence. Hiring data scientists using a poor model: that would be ironic (and painful). We used multiple methods to explore the topic, and arrived at the same outcomes. The result: validity.
For organizations facing a data scientist talent crunch, our experience offers two conclusions. It tells us that fundamental attributes play a huge role in the success of a data scientist. And: that it is possible to overcome the limitations of an air-tight talent market. The “talent” is there: it just needs to be found in a different way.
Looking ahead, it’s clear: demand for data scientists will not let up. Building out the data science function is essential to future success. Yet in a challenging, often frustrating environment, this approach offers a way forward, a healthy dose of confidence and optimism.
Angela Zutavern is a vice president in Booz Allen Hamilton’s Strategic Innovation group, focusing on data science and complex analytics for government and commercial clients. She has developed many of the strategies that are now helping hundreds of business and government organizations take advantage of analytics to make better decisions and gain a competitive edge. Ms. Zutavern advises clients on all aspects of applying data science, including implementing leading edge data science technologies and techniques, creating analytics strategies and cultures and getting the highest return on analytics investments.
Jamie Lopez, Ph.D., a Booz Allen Senior Associate and industrial- organizational psychologist aligned to the Strategic Innovations Group (SIG), provides talent solutions to his client base across the commercial and federal sector. He co-leads Booz Allen’s TalentInsightTM Group focusing on talent strategy, employment selection, workforce analytics, training, and development solutions across burgeoning technical career fields including Data Science, Cyber, and Predictive Intelligence. Dr. Lopez advises his data science clients on implementing comprehensive talent management solutions inclusive of job analysis, competency modelling, statistical validation studies, selection measures, and high- fidelity work simulations.
(image credit: Pink Sherbert Photography)
Interesting. We take a similar approach evaluating candidates in three categories. We find communication skills to be crucial too. Iike how you have them bucketed into your consulting category. You can visit this article to learn our three categories and to read our list of 20 data scientist interview questions: http://www.sas.com/en_us/insights/articles/analytics/data-scientist-interview-questions.html