The Harvard Business Review woke up the business world in October 2012 to a new reality in their article- “Data Scientist: The Sexiest Job of the 21st Century“. Yes, you may be having the same reaction that I did – “What?! When did Data Scientists become sexier than those perennially sexy Engineers?” And of course….”What the heck is a data scientist?”
Well, we are now two years into the future from that decree from HBR – maybe it’s time to reevaluate the facts on the ground. Do we now have these jetsonian data scientists with their statistical modeling jetpacks and oracle-inspiring foresight changing how business is done today?
Well… it depends.
#1: If you can find them…
The “sexiness” of the data scientist role also comes from its elusive nature in today’s world.
Let’s face it – the appeal is obvious. It sounds really comforting to business leaders to know that with the data scientist, they will finally get that holy trinity of (1) software programmer, (2) business expert and (3) advanced statistician all in one. That sets the expectations pretty high in terms of the experience, expertise and knowledge that such a person would need to have. The reality is that while you could find people with one or two of these skill sets, finding someone with all three, developed to full potential, is pretty darn hard.
I’m speaking from my personal experience. I lead the business analytics strategy practice at Saama, which is a pure play data and analytics solutions and services firm that heavily invests in finding and hiring data scientists. I can tell you that while these unicorns do exist, they are usually difficult to find and challenging to hire – for multiple reasons that will become clear shortly.
Therein lies the reality that underpins Point of View #1: It takes a village to make a data scientist
We at Saama believe that data science will evolve as a function (rather than a role) in an analytics organization, with a team working together to convert the “data mess” into “insight treasures”. This team includes a combination of business analysts, statisticians, developers, architects, data stewards and change managers, all working in unison as part of a clearly defined organization developing and delivering business insight.
All this while universities figure out how to define and churn out “data scientists” at the same rapid pace that programmers or engineers are churned out today.
#2: If you can figure out how to use them effectively…
While Big Data sounds great in its promise (and data science is the key to that promise), in reality when you look at the vast majority of companies out there today, they have not made the mindset shift needed to really harness this promise. For too long, analysts and managers have been trained to look for “exact” matches, data points that tally clearly and enable transparent drill downs to the roots of the aggregation. A whole industry around Business Intelligence evolved around this premise. Big Data by its very nature is “messy”. It uses statistics and probabilities to replace certainty and exactness. It uses complex models with simple answers to replace simple reports with detailed outputs. All of a sudden, it is not as simple to “look beneath the hood” and convince yourself of accuracy. Unless you have a uniquely gifted, statistically literate workforce (which is really rare for the vast majority of companies), you will need to have a certain level of trust in that data “magician” who provides you insights. That is a tough change in perspective that organizations will have to go through in order to trust and leverage the value of Big Data and Data Science.
And then there’s the translation problem – anyone with an untrained ear who has engaged in a conversation with a data scientist about a business problem knows what I am talking about. Statistics is a real science and its value depends upon careful interpretation. It is really easy to do bad statistics really well. Often, with managers and leaders who have an untrained statistical eye, it is impossible to tell the difference.
Point of View #2: Data science is a new capability that needs a fresh organizational perspective
Leveraging a data scientist to deliver real value needs a fundamental rethink of the value proposition of Big Data and Analytics to the future of the Enterprise. Everything about how we define problems, approach and resource them and interpret their solutions is different in the new paradigm.
Converting data insight (e.g. finding 3 statistically significant coefficients in a multivariate regression model) to business prescription (e.g. targeting 3 segments of customers to generate a 35% improvement in loyalty) to tangible business action (e.g. planning and launching 3 marketing campaigns) is a skill that most organizations will need to get good at… in a hurry.
Then, and only then, would you go out to look for your “sexy” new data scientist. And let me know when you find one!
In upcoming blogs in this series, I will address Saama’s Analytics Solutions Center of Excellence model for what this future data and analytics organization will look like in the modern enterprise. We will need to retool for Big Data – not just our infrastructure and technology capabilities, but also the organizational structure, governance and roles and the process model for how the enterprise establishes, funds and interacts with its data and analytics function to generate real and continuous business value. These are not easy, but nonetheless necessary transformations that enterprises are going through as we speak.
Nekzad Shroff is Practice Director of Saama’s Business Analytics Strategy practice. He is a seasoned strategy and operations management consultant with over 16 years of industry experience. He has a strong foundation in business and IT operations engagements across a diverse set of clients in Life Sciences, Healthcare, Technology, Finance and Insurance industries. He specializes in frameworks for creating innovative business, operations and process strategies for Saama’s clients to align their data and analytics investment to deliver tangible business value.