Building a world-class Problem Solving team

In my role as Head of Data Science and Analytics, the single most important problem that I am trying to solve is: What does a world-class team look like? How should you build and more importantly, sustain such a team?

It is always best to start with a few basic axioms (I call them so because these will need to be time-invariant):

1. It is all about the Mindset:  This is about building a culture of Problem Solving. Statistical Modeling has been around, we are barely scratching the surface of AI – tools and methods will come and go. The goal should be able: how do we solve problems in the best possible manner to deliver the most impactful solutions?
2. The Best teams are ‘fluid’: where team members bring multi-disciplinary skills and more importantly, a learning and curious mindset. One common factor that I do insist is a strong interest in thinking@scale – Data Science in the Enterprise needs a strong engineering mindset (not to be confused with the Data Engineering as a specific capability – more on this later)
3. Building a team is a Work In Process: The day you feel that you have solved this problem, stasis will set in – and that is but a step away from complete irrelevance. I think of Managers (including myself) as Coaches: whose job is to constantly re-build teams around individual strengths and shape a game strategy that they want to execute in a match.

Next comes the brass tacks question – what are the main capabilities that a great Data Science team should have? I think of these as 3 core areas of competence – which are not mutually exclusive:

1. Product Owner/Engagement Owner: In my mind, this is the single most important role in any team. Not only will this person to hold the vision, but also play the critical role of shaping thinking with business stakeholders – and the best of them will be thought leaders who will help define the journey of converting Data and Analytics into Business outcomes. This role would come with a strong Consulting mindset and ideally, with a background in Applied Math (Data Scientist) or Engineering (Product Development)
2. Data Scientist: The Applied Mathematician. I would look for someone with strong communication skills and a problem-solving mindset: who can take a business problem, break it down to component questions and for each one of them – convert it into analytical question, which in turn leads into a hypothesis. Round that up with communication: the ability to explain analytical model outputs in simple terms (e.g. storytelling with charts). In fact, the knack for communicating is going to be a key differentiator.
3. Product Engineer: This is more than a Data Engineer. The industry has done itself a huge disservice by equating data engineering with SQL and ETL/ELT. We are now well and truly in an environment where a successful outcome requires execution at scale. That goes way beyond the ability to pull together datasets for an analysis or building a dashboard. It requires a product engineering mindset – one that approaches every use-case with the following lenses:
   1. How do you build a scalable solution? This influences not just say, how modular and efficient the codes are (esp. for the data ingestion) but also a product mindset on integrating, say a model output, into a digitized workflow.
   2. How do you build a user-centric solution? It is no longer good enough to build a great statistical model. The real ask is to enable users make better decisions in the context of their environment. And that requires an approach that starts from a foundation of design thinking.

And if it looks like building a team of unicorns, here are a few more things to consider: 

1. These roles are not mutually exclusive – i.e. the best team members will have two or more of the skillsets. The most valuable ones being the ones who are at the intersection of three. It is not enough to be looking for a unicorn – we are looking for a unicorn that can fly and comes with a halo! The good news is that as the industry matures, we see an increasing pool with these capabilities
2. The real value comes not just from these individuals – building a great operating cadence is just as important:
   1. Work in Pods: Self-regulating, small teams working in an agile environment. Ideally, each Pod should have a business partner; and in some cases, it is fine to have Product Owner/Engagement Manager to own multiple pods.
   2. Outcome driven: Hold teams to outcomes, not effort. Outcomes must be incremental (one guideline is to run 8-week Epics with 2-week sprints) and validate the business outcome at the end of every 8-week cycle.

The fascinating part is that we are seeing these roles and operating structures evolve as we speak. Data Science and Analytics is rapidly changing – one of the few industries that truly sits at the cross-section of multiple competencies. We continue to learn on a daily basis – and the big challenge as a leader is to shape and execute these changes wtihin the framework of organization dynamics that normally take months, if not years to change (think HR policies, organization design etc.)