Data science is a highly multidisciplinary field. And to run data teams profitably, organisations need versatile leaders and managers.
While analytics leaders’ job entails managing resources and a lot more, data scientists deal with technical aspects like coding, modelling, data collection, among others.
We caught up with a few industry professionals to understand if analytics leaders should have an in-depth knowledge of data science. The short answer is: Yes and no. For the long answer, read on.
Yes And No
There are several drawbacks to analytics leaders not being data scientists. Since data science engagements are hypothesis-driven experiments with different ways to arrive at the right solution, good knowledge of data science will enable the leader to guide the team in the right direction.
“The hands-on managers bring subject expertise and have been in those shoes before, which is effective,” said Srikanth Velamakanni, Co-Founder, Group Chief Executive & Executive Vice-Chairman of Fractal Analytics.
“Especially in data science, you observe that people who have done this before have very strong empathy for how it is being done currently, and the junior members also feel that there is a lot of guidance coming in. This helps in avoiding mistakes.
“Also, if you have a manager who can point their team in the right direction, they can really save a lot of time. They can ask them to, for instance, not bother looking at ten different things, but take this particular approach, because they know this is where things are,” he added.
On the other hand, industry professionals also think it is not essential for analytics leaders to have in-depth technical knowledge.
“All three roles are focused and complement each other. CXOs and Analytics Leaders are expected to focus more on the business value, effective program and data governance, not on the technical skillset,” said Prashant Pansare, Founder and CEO of Rubiscape.
“We define the Analytics Leaders persona as the orchestrators of the various stages of the Analytics Program, running multiple projects on a diverse set of technologies producing a harmonious outcome which forms the basis of fact-driven decisions. Tool level knowledge is an advantage, not a critical factor,” said Pansare.
Tradeoffs
Velamakanni asked to consider four quadrants with scores from low to high, with both responsibilities on either axis.
“The high-high quadrant, where managers are high on managerial as well as analytical skills, present the best talent but are very rare to find. And the low-low ones should not be in the system,” said Velamakanni. The other two quadrants also present uncertainties.
“If the leaders are low on management capability and very high on their analytics understanding, they are not going to be good with managing resources. But if they are high on management capability and low on analytics, it could be somewhat dangerous if they get into too much micromanagement.”
If you are high on managerial skills but low on analytics, then you want to let the team be. But if you are high on analytics skills and low on managerial skills, then you want your team to be more of self-starters,” he added.
Essentials
One challenge with pure data scientists is their weaknesses to either look at the bigger problem or explain and visualise the output in simple terms.
“First and foremost, they need to understand the business goals and the environmental factors well – markets, trends, people, and products, which are more than just a technology,” said Kedar Sabne, Co-founder and Director at Rubiscape.
“They should also know the people, process & profitable performance aspects of analytics programs. They should be able to deliver a measurable business value and impact on the key stakeholders.”
The other important factor is the ability to spot the right talent.
“What is happening today is that there is a vast difference between the average talent and game-changing talent. Game-changing talent is at least five times as good as the average talent. So, the ability to spot and pick the right talent needs to be very good,” said Velamakanni.
“Secondly, as a manager, you have to understand how work gets delivered by understanding aspects like time required, conflict resolution, or the quantum of work. They should not come across as too unreasonable or too relaxed and have a good sense of estimation of the effort required.”
The experts also pointed out that the ability to understand diverse data, roles, and mindsets, along with excellent communication, visualisation and business-facing skills, are more important than the core data science skills.
Finally, the analytics leaders and data scientists must be on the same page. To ensure that, the managers should be able to set the team composition correctly.
It is important to ensure the data scientist appreciates the inputs and understands the leader’s value in the engagement. Industry professionals think there is a tendency for leaders only to sit outside and review the progress. In such cases, you might not get the level of cooperation or final outputs as desired.
Upskilling
Finally, the industry professionals suggested several ways leaders could upskill themselves, technically and managerially, to be a better analytics leader.
“One of the most important ways is to read books,” said Velamakanni, “There are a lot of good management books like Peter Drucker, Marshall Goldsmith, or Stephen R Covey to smoothen some of your rough edges. What also helps is having a mentor in people who have done this before to get guidance on challenges that you are facing.”
“On the technical front, one should be enthusiastic about taking up challenges. Another way is to take up teaching as you will learn a lot of things in the process. One could also start writing. They say that writing is a process of discovering how little you know.”
Courses from leading institutions through Coursera, Udemy, etc., are widely available and can help in the initial upskilling phase. Participation in online challenges and forums like Topcoder and Kaggle can further help understand and cement their knowledge and understanding.
