What does your ‘Minimally Viable Data Scientist’ look like?
The creation of data-driven value requires the right skills, and that means bringing together an analytic community of data professionals. But before we run off hunting unicorns, there are a number of things we need to be aware of:
1) There is no agreed definition of Data Science
Whilst “data science” as a term is typically attributed to Professor Jeff Wu (who coined the term as a suggested rebranding for “statistics”) the modern use of the word stems the marketing hype surrounding the big data movement. At Mango, we define Data Science as “the proactive use of data and advanced analytics to drive better decision making”, but every data scientist you speak to will have their own favourite definition. Note: if you want to read more about our definition, read this.
Given that there is no single agreed definition of “data science”, there is also no agreed definition of “data scientist”. So, within reason, “data science” can mean different things across different organisations. Understanding what data science means to you, and what the business believe it will deliver (i.e. why you’re investing in it in the first place), is a great starting point to understand the skills you’ll need in your team.
2) Data Science requires a mix of skillsets
When you start looking for definitions of “data scientist”, you’ll quickly run into a variety of venn diagrams. This is because there are a range of skills required to actually deliver data science outputs, particularly in a complex, commercial organisation. If we consider the Mango definition (above), then we need to be able to:
• Manipulate data sources of varying shapes and sizes
• Model the data using a broad range of analytic approaches
• Engage effectively with stakeholders across the business to ensure the “change” lands
• Create production-grade code to deliver the insight into the hands of decision makers.
At Mango, we believe a Data Science function needs a mixture of 6 different skillsets to succeed. These are represented on our Data Science Radar:
3) Unicorns. Don’t. Exist
When we start thinking about these combinations of skills, it is important to note that (sadly) … Data Science Unicorns do not exist. Having interviewed “data scientists” for ~20 years now, I’m yet to meet one. When I say “unicorn” here, I mean someone with the full set of data science skills on the radar.
4) Data Science is a Team Sport
Because of this mixture of data science skills (which isn’t necessarily found in a single person), together with the potential “always on” requirement of analytics in a data-driven company, Data Science is seen very much as a “Team Sport”. In other words, whilst we may not be able to find unicorns, we can build a team that (together) have the skills needed to deliver data science.
Building a great Data Science Team
Given the above, building a great data science team is about:
• Understanding the skills you’ll need in your team to deliver on your objectives
• Knowing what your ‘Minimally Viable Data Scientist’ looks like
• Hire “spiky people” (people with key strengths that you can bring together)
So … what is a ‘Minimally Viable Data Scientist’?
Your ‘Minimally Viable Data Scientist’ (MVDS) is a theoretical data scientist who has the minimal skills required to operate as part of your team.
For example, at Mango our “MVDS” has to have strong “Programmer” and “Technologist” skills, since our data science work is built on good programming foundations in R and Python and developed using software development approaches. As a business, we’re also Consultants, so soft skills are of importance. Beyond that, we need at least a solid grounding in modelling, visualisation and data wrangling. So our “MVDS” looks like this:
If we’re screening data scientist candidates, understanding whether they meet this minimal threshold allows us to understand whether they will be able to operate within the team. However, we’re then looking for “spiky people”.
Hiring “Spiky People”
Using the Data Science Radar during the recruitment process allows us to quickly understand whether a candidate meets the “MVDS” threshold, but also allows us to understand their skills across these 6 axes. Next we’re looking for “spikes” in the chart that represent particular specialisms for an individual. For example, consider the following 3 (theoretical) candidates:
Each of these candidates pass the “MVDS” threshold and have specialisms in at least 1 area (i.e. spikes). What we’re looking for here when we’re interviewing is how these specialisms complement the rest of the team. The “MVDS” approach means we know these 3 individuals can operate well within our team, but then it’s a case of looking at skill gaps, building these capabilities and understanding how these specialisms will impact the communal skillsets we already have.
When hiring for your data science team, it’s important to understand what you’re hoping to achieve and therefore what skills you’re going to need in the analytic community as a whole. By having a strong understanding of your ‘minimal’ skillset, you can clarify the threshold beyond which people could operate within your team – this allows you to focus on adding people with specialisms you need to succeed.
Why not join our webinar for a guide to Building the Ultimate Data Science Team.