Five Signs of an Effective Data Science Manager
In this article, we will go beyond the theoretical realm of what a data science manager does and focus more on how to become an “effective” data science manager.
Productivity vector created by pikisuperstar
The times have changed rapidly and these days job titles and responsibilities do not really depend on the number of years of experience. You can lead the data science teams at any age as long as you understand the business objectives and align them with the team’s expectations and interests.
I have listed down below the characteristics of an effective data science manager from the two lenses of business and team leadership.
We all must have heard of the push and pull model at some point in our professional life. One analogy to this push-pull concept is whether you keep waiting for somebody to push the work to you.
Or, you are opportunistic enough to understand the business problems and pull the high-impact work on your plate.
Are you a pioneer in helping the business understand how and where machine learning solutions can help vs where it would be a mere aesthetic?
Business vs ML Metrics
Your ability to make your stakeholders ML-aware will take you farther in this career. It shows that you are able to think long-term and are able to bridge the gap between churning models with high accuracy vs building the models that add tangible business value. Let us all be honest that the machine learning model building process is time and resource expensive and needs a clear understanding of the opportunity size and impact of ML models.
While presenting your proposal to the leaders, you should clearly lay down the assumptions, pros, and cons of using ML for the use case, the vulnerabilities of the model, etc. It takes off a huge load from the business executives and brings synergies and trust to the data science team.
Meeting the Team’s Expectations
So you have convinced the management with the value proposition of your solution but a data science manager is also tasked with gauging the pulse of the team. There is a perception that a data science team gets to work on and with new state-of-the-art cutting-edge models all day long. But the reality also includes building automation workflows for repeat tasks, working on a proof of concept for 4 months only to see it scraped off, model monitoring, and maintenance, explaining model predictions, etc. to name a few.
The essence is you need to identify the align the interests of your team - some may be from a core research background and can think through innovative solutions while others might be interested in being a full-stack data scientist aka all-rounder.
Always remember that the work that is appearing to be a burden for one of your team members (could be because they have done it long enough and aspire to challenge themselves with something new) might be a good opportunity to learn and upskill for someone else.
Its all about finding the coherence that comes with the often underrated skill of knowing your team well. Do not maintain a mechanical aka transactional relationship with your team where you are only communicating the tasks and not incorporating their expectations from work.
Keeping up the Team’s Morale
Truth be told - you can not get a team to work for a long time on a task they are not interested in. But you still have to finish the task. A good data science manager will note the interests of all the team members and should be able to truthfully explain the reality. But what's the reality?
That the management understands and plans an individual’s development and growth charter where business objectives converge with the team’s goals and interests. But presently we need to finish a business deliverable and whenever the next opportunity arises, the team’s alignment will be worked out as raised by them in one-on-ones.
You need to be utterly faithful to your commitments with the team as you are working with brilliant minds in the industry. You can only onboard them by making them understand that the successful completion of current tasks will take them a step closer to their areas of interest.
Note that we are not living in an ideal world, so chances are high that you might not be able to meet the expectations of the entire team at once. You are working under the constraints that are implicitly called out loud and clear - business first. So, what will you do then?
That’s where the concept of emphatic leadership comes from. Your awareness of the team’s expectations and willingness to help them accomplish those is a testament to your management skills already. The efforts in the right direction are well-recognized by the team. Those who are yet to receive their share of the pie also understand that given the right time and opportunity, the manager will position and enable them to achieve their foreground
What do You Promote: Competition or Individual Growth?
Personally, I am not a fan of promoting competition within the team and wrapping this motivation under the name of healthy competition. I rather choose to respect every individual for their own uniqueness and tap onto their potential.
What extra would I get by asking a fish to climb the tree?
Yes, everyone should be agile enough to learn and deliver fast. But this deliverable can very well come from their area of interest, rather than forcing them to walk in someone else’s footsteps. A good data science manager holds himself accountable for the team’s learning curve and promotes some buffer time to explore the work of their own interest. Oh, but you might think that your area of interest would not necessarily find a business outcome.
That's where the good part of management shines through - you motivate them to spend 15-20% of their time exploring the new algorithm, learning a new skill like a programming language, etc. Essentially, giving them the wings to think outside the boundaries of the pure-play business context.
While we all wish for a rulebook that can set all managers to success from the first day, unfortunately, there isn’t one.
Everyone has their own management style and should stick to what comes effortlessly to them. You can not fake yourself long when managing teams. So, it's best to be your genuine self and do what an effective manager does - delivering the business outcomes while keeping the team motivated.
Vidhi Chugh is an award-winning AI/ML innovation leader and an AI Ethicist. She works at the intersection of data science, product, and research to deliver business value and insights. She is an advocate for data-centric science and a leading expert in data governance with a vision to build trustworthy AI solutions.