How to hire the Executive Data Scientist(CDS) and Scale Data Science Across the Enterprise.
In the age of data engineering and data science, many firms are trying to hire their Chief Data Scientist or Chief Analytics Officier but continue to struggle with how much engineering and science skills those leaders need versus leadership skills.
In my consulting and coaching practice, I am hearing some general themes in terms of challenges which demonstrate somewhat jumpy thinking in terms of what are the essential skills of the Chief Data Scientist: Some skills you can teach and others come from deep experience.
Firstly, decide if you are looking to hire an executive leading data science or an individual contributor python expert. These are very different skills, and there are no Chimeras? Has anyone seen a Unicorn recently? I don’t think so. I subscribe to the complimentary and pod building theory of hiring which says the Chief Data Scientists needs to be surrounded by a variety of team members, not the least of which is the data engineer to ensure proper deployment of the ML solutions and to ensure IT resources are dedicated to projects. Also, one question which may help you in hiring is asking yourself and knowing what are the connected skills to data science. Before DS was Predictive Analytics(PA) before PA was quantitative analysis(which skill sets are new and which ones are overlapped would be a good heat map or visual/infographic to create). More on this later.
If you are currently experiencing some failures with Machine Learning Projects or AI projects particularly those who are outsourced to vendors and consulting firms and now you are searching for a new Data Science Leader, I wanted to caution you on some trends. Let’s say you have a data science manager and data engineer in place, but you are blaming them as they continuously default to recommending machine learning solutions, but you as the COO or Managing Director or President don’t know whether or not the best solutions for the current business problem is ML. I want you to ask yourself the following Question:
Do you have an Executive Data Scientist in charge or an Individual Contributor(IC) data scientist influencing or running your function? If it is the latter, you may have over-indexed on the science and engineers aspects of their skill set and not on the executive translator component of the role. The Executive Translator skill comes from years of operating data science leadership experience. (If you need help with designing the job description or the interview questions, please get in touch with us, as we don’t recommend behavioral based interviewing alone for this type of hire/role.)
Ask yourself does the person or team recommending DS or ML solutions lead with the technology or ask questions like:
What business problem are we trying to solve with the ML or AI solution? Are we trying to solve a customer problem, and operations issues or a risk issue?
Is ML the best solution versus perhaps other forms of predictive or prescriptive analytics such as Regression or some other type of analytics?.
Machine learning and automation are appropriate for specific business problems but not all. This is where perhaps the rapid separation of Data Science from Business Analytics will come back to haunt us in the sense that we need broader multi-disciplinary skills to make sure that the problem is solved with the best solution. I am a big fan of thinking about Analytics, DS and AI as being under the same umbrella as we don’t want to end up in a situation which happened in marketing where we had a digital marketing and overall marketing starting to separate at the expense of the customer. In the end, they mostly stayed together.
Collaboration and configuring the right DATA SCIENCE GOVERNANCE AND INTERACTION MODEL will be essential going forward. Creating dominant Business Teams combined with Analytics, Data science, and AI, Data Engineering team(ADSAIDE) in the right configuration will be the key going forward, but we may want to stop to consider the skills sets and how they are configured to complement one another for maximum impact. We are currently fragmenting these skills in pockets across the enterprise and are seeing may suboptimal solutions.
If we want to scale Data Science, we recommend the following approach. These will continue throughout a series of blog posts, perhaps with coauthors as well.
#1 EXECUTIVE DATA SCIENTISTS or EXECUTIVE ANALYTICS LEADERS need leadership skills first., wait you mean that they don’t need to code in Python or R?
– While it is undoubtedly helpful to have a leader who either has consumed the deliverables from R and Python and know enough to challenge the team or has a coding background in R or Python combined with a knowledge of Cloud-Based Technologies, I would say since these executives may not code, this is not more important than leadership traits because you can teach R and Python(Get in touch with me if you are looking for a list of resources for this education), but you can’t teach quickly:
- Intellectual Curiosity
- Vociferous Learning
- Executive Communication.
- Translator skills. Translating complex business problems and requirements into analytics and data science speak.
– Executive Data Scientist and AI leaders with good character and Ethics are vital. You can’t have these leaders pushing their agenda for DS/AI/ML at the expense of the firm(Remember the CRM debacle?). Some of the projects they will advocate having a hundred million dollar price tags associated with them. Leaders who say things like “ Let’s make the business case to understand the returns and longer-term impact of these investments” are the ones who you should be hiring. Also, a leader who has a healthy balance and views analytics and Data Science as a hierarchy of data mining I find make the best recommendations. A Data Science Leader who says something like “God forbid we don’t use machine learning to solve all problems” is the one who is demonstrating balanced and multi-disciplinary thinking, and that is the one I want to hire. Why? Because they know when to recommend supervised versus unsupervised learning and they know when regular regression might be good enough. Also such as an executive who understands all of the statistical tools across a variety of use cases gets the gold star over one methodology. Also a CDS who likes to whiteboard out the data science process and ensures a path to deploy the algorithm is a winner in that they ensure successful outcomes.
Side Bar: Overindex on Communication and Collaboration, Look for a good grounding in data science and engineering skills but balance both hard skills and soft skills., Avoid in any leadership role IC data scientist masquerading as executives. This is particularly tricky as python has only come into its own in the last five years and there is the shiny object syndrome which can be hypnotizing I admit. What these tools can do is, in fact, impressive, but you need to sit down and write a clear job description that reflects what you need now and in one year especially if we are talking about the CDS role. Hire leaders who are grounded in data science fundamentals, but also understand the broader analytics and IT ecosystems. Multi-disciplinary abilities are key!. They also need to understand data management concepts like Data Lakes and how data quality matters. Can ML work if the data is crap? Nope. Check out their LinkedIn profile have they continued to educate and transform themselves on data science topics? Are they involved with the industry, education institutions, certificate programs and more?
We would love to hear your thoughts and please know we acknowledge and enjoy all the advances in technology and in-memory and open source tools, but we want to know what constitutes a leader in Data Science in your mind? Looking forward to an interesting dialogue on this. There are many POV’s I am sure.
Dr. Tony Branda
Founder, Analytics Hall of Fame.