Building a team, not hiring individuals. Where to start, what to look for and how to decide.
And here I am. Back at the beginning. Having to build a new data science team, from scratch. Again. But this time it feels different. Strangely, the number of mistakes I’ve made or witnessed in the past is giving me confidence.
Building a team is not easy. You can’t simply order skills and personalities on Amazon. Individuals are exactly that: individual. With unique skills, experience and behaviours, which they bring into the dynamism of team environment.
If this isn’t hard enough, there is an extra complexity for data science teams. Data science is a nascent industry. Businesses are learning how to make the most out of it, new technologies and methods are emerging daily and data scientists are discussing myriads of data roles to demonstrate the complexity of the field.
Structure follows strategy When building a data team, it is — more than ever — important to have a clear vision and well-defined data strategy, aligned with the overall business needs. Before you call a recruitment agent or email the head-hunters, ask yourself: What is the raison d’être of the team? What do you want the team to achieve? Will you be building a data-powered products or augmenting decision making? Where will the team sit in the organisation? What data is currently available? What technologies should be used? What are the responsibilities of the team?
Dream big Now, with these considerations in mind, start to form a mental picture of your dream team. Focus on the skills, knowledge, experience, and expertise you think you need. Crucially, don’t limit your thinking to individual positions. Instead, keep your thinking broad — focus on the problems your team needs to solve, rather than on the actual people who will solve them. And take full advantage of this being a green-field situation. It’s rare. No legacies. No tie-ins. Dream big with your feet planted firmly on strategic ground.
While your business’ strategic needs will be very specific, I believe the following trio of traits must exist within every data team, noting that they rarely appear fully formed in one person:
thirst for knowledge: seek someone who has studied every method and is constantly looking for new trends
data enthusiasm: find that person who seriously wants to prove the value of data
engineering mindset and strong coding skills: make sure you have someone who deeply understands what it means to create production-ready solutions
Incorporate these traits into your job specs and advertising, making sure that you don’t fall into the trap of confusing the overall team vision with the specifics of the experience and skills required. Be clear on the types of problems your data scientists and engineers need to solve. And explain the technology available and the data to be used. This approach will make the hiring processes considerably more efficient.
Tomorrow thinking today You have your dream team mapped out — now it’s time to make it real. Approach the hiring process with an open mind. You are building a team, not hiring individuals. So, think more like a sports team coach, remembering that individual brilliance rarely wins you the world cup. Consider candidates in clusters, anticipating overlaps and gaps. Seek to identify opportunities to complement and supplement, rather than a one-size fits all approach.
Then, with CVs in hand, I recommend following Warren Buffett’s sage advice in the interview process. He says:
“You’re looking for three things, generally, in a person. Intelligence, energy, and integrity. And if they don’t have the last one, don’t even bother with the first two.”
So, what does this mean in the context of data science?
Data science is a craft. You gain experience and mastery through projects. Lots of projects. Where the more difficult the projects you have under your belt, the better. Unfortunately, this is not always visible from a CV. I’ve seen fresh graduates with tons of experience from their school and pet projects. And I’ve also seen people with 10-year corporate careers who — remarkably — have none.
Use your interviews to assess real track record, not clocked time. Seek craftswomen and men — as intelligence in data is a by-product of rolling up your sleeves and getting splinters. For me, energy means two things in this context. Firstly, data is a dynamically evolving field. You need people who are proactively keeping up. This is akin to the ‘thirst for knowledge’ trait. Secondly — and arguably more importantly — data science is meant to solve business problems. Ergo, hire people who are genuinely interested in business mechanics, problem solving and want to make a difference. Plus, it’s energy that lifts people up when projects stumble. And it’s energy that propels people to keep picking themselves up, time after time, iteration after iteration.
When it comes to integrity, look for people who won’t make the world of data more ambiguous, complex and unpredictable than it already is.
As Tony Dungy says, integrity is “the choice between what’s convenient and what’s right.” Data scientists need to be true to their morals when it comes to the use of data, validation of assumptions and making analytical choices. I would also add that, for me, integrity is about clear communications and congruence — which includes not adding to the tsunami of ‘consulting speak’ and jargon which drowns our industry. Baffling our business colleagues with acronyms and technical colloquialisms isn’t helpful. After all, the better people outside of data science understand our world, the faster we can drive meaningful results.
And, my final recommendation for your interview process, it to hire for today, but think of tomorrow. Go back to your data strategy throughout the recruitment process to test that you are identifying people that will grow with you and the team.
The enigma variations Now you have your core team assembled — the fun begins! You have hired great musicians, now it’s time to transform them into an orchestra. Get to know each other. Find your rhythm and ways of working. Learn each other’s strengths and blind spots. Give everyone the opportunity to shine and actively encourage the team to build on each other’s successes so that your solutions evolve. After all, this is how you find the best, workable answers to your ever-evolving business problems. This is how you deliver — repeatedly.
It is important for the business and the team’s morale (and your ego) to land some early successes. The lessons and momentum from these initial projects is the glue that will hold your team together. Only then can you bring in reinforcements and grow and develop the team further.
So, my plan is laid out: i) structure follows strategy; ii) dream big; iii) tomorrow thinking today, and iv) enigma variations.
I’m looking forward to being inspired by what my fellow data scientists, data engineers, ML engineers and others achieve, humbled by all their knowledge and skills, and excited to get to know them and work together on amazing projects.
As ever, I’m indefinitely grateful to Chelsea Wilkinson for patiently shaping my thoughts into a publishable format and finding analogies for my key messages.
Comments