In most organisations, data science has started its life as a research team– a ragtag bunch of independent researchers not bound by business impact and free to research and work on different areas.
Today, data science teams are the centrepiece of many organisations– a collective of accountable leaders driving business impact at scale.
The major responsibility of a Head of Data Science (HoDS) is to guide data scientists in impactful initiatives. HoDS may not be directly responsible for profit and loss, but has a huge role to play in directing the team towards organisational KRs and successes. In this article, we will look at what’s expected from a HoDS and how organisations can get maximum value out of this role. We will also cover the skills needed for the HoDS and the hiring process for this role.
AIM Daily XO
Join our editors every weekday evening as they steer you through the most significant news of the day, introduce you to fresh perspectives, and provide unexpected moments of joy
The head of data science is a leadership role. The HoDS should create a vision and a roadmap for the data science team in consultation with the company leadership team. The HoDS is expected to set processes and governance for the team. The HoDS should be able to execute the roadmap and drive business impacts with the help of the data science team.
Download our Mobile App
Building a successful data science team is about enabling the right infrastructure to build and deploy efficient models. It is also important to inculcate a culture of impact-driven data science. The data science leader should be able to implement the best experimentation practices and drive attributable impact metrics for the organisation.
Different organisations follow different team structures. Some follow a POD-based approach where different functional teams come together for a specific POD or they could keep a central data science team that caters to different products and businesses.
In the ‘POD’ based model, the HoDS will drive functional excellence overall and set up governance processes for the data science models being developed. The HoDS are expected to keep a tab on the overall work being delivered from the POD. The HoDS may have senior data scientists lead different data science PoDs.
In a centralised structure, the HoDS engages with stakeholders along with various team leads to understand and drive the projects. The HoDS also allocates work to team members and is directly responsible for impacts in each of the projects.
Does this role need to be hands-on?
An organisation can bring in the HoDS at the start of the data science function. In this case, the HoDS can build the team from the ground up and can be hands on to some extent to solve for initial use cases as the team is getting built. However, at this point in time, the company may not be very clear on the data science use cases and investing in a leadership role may not be advisable. It is better to get a few mid-level data scientists onboard first to understand the use cases. The HoDS can then be filled in. If the HoDS is coming in already with a small but operating team, the role does not need to be hands-on. However, the person should be as close to details as possible and should have the immense technical depth to build and guide the team.
Should ML Engineering be aligned to HoDS?
Data Science and ML engineering are two different functions. Data science entails the development of a predictive solution to the problem and ML engineering takes care of the infrastructure development for deploying the model in production. Both can be different departments.
The DS and the ML engineering leaders can establish the right contracts in place for the close working of the teams. In smaller organisations with a relatively small data science team, the data science team can take the responsibility of deploying the models with possibly one or two ML engineers reporting to the HoDS.
Principal scientist vs HoDS: What’s the difference?
At some organisations, the role of principal scientists is not distinguishable from HoDS. But strictly speaking, they are two different roles. A principal scientist may or may not have the responsibility of the team. However, the HoDS is about building and leading teams. The HoDS is also about setting a vision for the organisation from a data science perspective. The principal scientist is typically not expected to have broader responsibilities. However, the responsibility of building deeper and more robust technical solutions falls squarely on a principal data scientist.
The principal scientist should be able to guide and enable the team to come up with SOTA solutions and work closely with engineering teams to architect the deployment solution as well.
The skills needed for a HoDS include technical knowledge and problem-solving; the ability to build and nurture teams; focus on delivery, stakeholders, strategic thinking and executive presence.
Technical knowledge and problem solving
The job calls for technical depth in statistical and ML concepts and experience in handling both structured and unstructured data use cases. Depth and breadth in solving data science problems are critical. The person should be able to understand components of ML engineering and be aware of model deployment issues.
Interviews could focus on real problems the organisation is facing to gauge the candidate’s problem-solving skills. The interviewer should look at the candidate’s past work and the technical problems he/she had to deal with earlier. Few companies have a technical presentation round where the candidates are asked to present their past projects to a panel of experts.
Ability to build and nurture teams
The leader should have managerial experience of at least three to five years. HoDS should have experience in guiding teams– from a technical and career-wise perspective–and should be able to set and build an open and innovative culture in the team.
The interviewers should look for how important the team is to the candidate, how they take hiring decisions and handle underperforming teams. Hiring managers should ensure HoDS candidates have the potential to guide and facilitate the teams’ growth. It’s also advisable to have a team member have a discussion with the candidate during the hiring process.
HoDS should have worked in roles with multiple stakeholders. They should have disrupted some of the existing systems and processes with respect to data science in their capacity. HoDS should be able to empathise with stakeholders and build relationships.
While interviewing for a HoDS role, a dedicated stakeholder round is important. It is crucial that the key stakeholders from product, engineering or business teams like the candidate. The HoDS need to work with other stakeholders as a team and hence stakeholder interview is a crucial part of the hiring decision. Stakeholders can focus on how quickly the candidate can absorb the business context and is able to find a solution to the problems from a data science angle. They can focus on how well the candidate can contribute to KRAs and metrics for the overall organisation. The stakeholders should also look at if the person can prioritise various tasks and communicate clearly.
Focus on delivery
HODS should be able to focus on key priorities and find out the shortest path for the success of the overall organisation and work in a transparent manner.
The hiring manager could focus on the candidate’s thought process on MVP solutions. The kind of transparency and governance process the person can set up should also be an area of focus. Posing scenario-based questions on project roadmap, prioritisation and governance could also help assess the candidate better.
The HoDS must have a clear idea on how they want to shape the data science team. She should think and articulate on the key data science areas that can shape the long term direction of the business. Additionally, the leader should be able to influence the overall OKR of the organisation and create long term valuable data science assets (like patents).
The HoDS needs to establish a rapport with the CXO of the organisations. HoDS is the interface between the data science teams and the executive leaders.
Both strategic thinking and executive presence can be covered by having an executive in the panel. If the hiring manager is the CXO then another CXO can assess the candidate in these areas.
Some focus areas in the interview could be looking at how the person thinks on long term initiatives, how they prioritise initiatives with strategic focus etc. The person can be asked how they would ideally structure the team and look for indicators that signal the big picture thinking. The interview round could also cover cultural aspects. In a nutshell, how well the candidate is aligned with the vision and the culture of the company plays a key role in the hiring decision.
The right data science leader can be a force multiplier in an organisation. The right person could accelerate business impacts through AI. Moreover, the HoDS could pave the way for long term strategic thinking on AI.
This article is written by a member of the AIM Leaders Council. AIM Leaders Council is an invitation-only forum of senior executives in the Data Science and Analytics industry. To check if you are eligible for a membership, please fill out the form here.