Build, Lead and Develop a High-Performing Data Science Team Within the Commercial Function


I recently caught up with some friends in Singapore, with whom I shared ideas on how to lead, build and develop a high-performing data science team in the organization. I will share my opinions in this article.

With the rapid explosion of data generation, it is becoming more important to use data science to generate insights in almost every organization. However, I have also heard and read articles that portray data science as hype or a bubble. This is understandable, especially when we put an academic powerhouse into a commercial environment. Doing good science requires a huge amount of effort and research, while a good business must always execute in a timely manner. So, hiring a group of data scientists in a commercial environment will naturally conflict. If they do not prove themselves valuable to the business, they will be purged or made irrelevant.

Data Science team need to balance between doing good science and delivering practical value they bring on the table

Following are some areas to consider when building a data science team.


Leadership

It is extremely important to have a strong data science leader who mitigates the risk of being purged. Leadership in data science isn’t just leadership in using some of the latest algorithms to marinate numbers from your database. Data science leadership is in a different league. On top of having a deep appreciation of the concepts behind the math or algorithm, data science leaders must be aware of how this information or these insights can drive a new business operating model. Furthermore, data science leaders must also be aware of diverse skills and possess the soft skills necessary to pull together these resources to make a large-scale project successful.

Yes, that sounds pretty insane, but allow me to explain further. As people move up to take on important/leadership roles in an organization, they graduate from being developers to managers to leaders. One of the most important tasks that a leader must do well is make decisions – know how to play, where to play and what to play. Imagine a great leader with a good personality, soft skills and charisma in the big oil and gas industry moving on to take the CEO role in a big company in the fashion industry. No matter how charismatic the person is, or how many leadership qualities the person has, he or she will have a very difficult time making good decisions amidst the many proposals from the trusted panel of advisors (although a mature organization would have already developed playbooks or decision-making mechanisms to mitigate the risk of a bad decision).

Without good appreciation of data science domain, you find it extremely difficult to make good decision

Data science is an intersection between IT, commercials and “science”. Therefore, it is a good idea to appreciate each of these areas. Otherwise, the leader might not be able to differentiate a good proposal from an average or poor one. If you are forming a brand-new team, with no playbook in place, you really need a strong candidate. For a commercial organization, a good data science leader is one with good technical training and experience running a business or P&L (e.g., a Ph.D./Master student who was running a business during his or her candidature, or a business executive who went back to school and received graduate degrees in data science). Commercial acumen or business sense is extremely important for the data science team to stay relevant in the commercial function. In the early stage, the leader would likely need to be very involved in working with the team on a small number of projects to deliver value before slowly moving toward managing the people, nurturing the culture, managing the stakeholder and partner ecosystem, and making strategic decisions.

Data Science is a team sport. But a very strong leader would very significantly increase the odds of success, especially when you are building this team in large commercial function without a playbook or baseline

Identify Capabilities That Are Relevant to the Organization

If you are having a migraine, do you visit a colon surgeon or a specialist? Similarly, a patient will not be comfortable getting a neurosurgeon to do a bypass surgery.

Data science is wide and deep. Do you need an NLP expert? Image analysis capability? Or just a generalist to treat the organization’s common cough, flu or fever? Map out those capabilities and focus on getting those people in your team. Within the commercial organization, I recommend focusing on only 2–3 capabilities at the start and peg business decisions that can be optimised via those 2–3 capabilities rather than the other way around. The rationale for this suggestion is that the team must quickly deliver “actionable insights” that have a real impact on the business.

Note: Sometimes it is a good idea to get cross-discipline data science capabilities for diversity. That decision will be important when you are interviewing and searching for data scientists. The capabilities map will guide your search process.


Finding and Attracting Talent

Talent in one company might not be talent in another. Define what characteristics of a talent are important to your organization

Talent attracts talent. The first step in building a talented data science team is to hire a senior data scientist or leader. Finding a good leader is really important! During interviews, a good data science candidate will be able to quickly assess the prospect of the new team in terms of fit. I encourage you to share the current projects and the strategic direction that the data team is heading. The more specific and clear you are, the better. Of course, balance this with the sensitivity of the information you share.

The cost of finding a wrong candidate as the pioneering team is extremely painful. So take time to find your candidate

Take time to find and evaluate the candidate. Once you have found a good candidate, go all out to secure the candidate.

You might want to attend data science meet-up sessions in your country and as many networking sessions as possible to connect with prospective data scientists. Partnering with universities and hosting some programs with them would help you spot early talent as well.

If you are looking for the best practices, I am sorry to disappoint you by saying that most best practices don’t work well. You want the best practice? Go all out to find a good leader and attract these talents


Ongoing – Developmental Mentorship Session

Employees are hired and receive payment in exchange for their skillsets and the services they render to the company. Beyond the “transactions”, it is the employer’s responsibility to train and develop employees and make them more “employable”. Yes, you read that right – make them more employable and valuable in the market. If they stay on your team, they will be great assets. If they move on to bigger roles, they will be contributing their talent and elevating the “standards” of data science. Either way, this is a good thing. And, somehow, you will realise that you never “lose out” on good candidates.

Train your data scientist and make them more employable and valuable to the market. Trust me, your team will never lose out on talent

A good mentorship plays a critical role in keeping a data scientist engaged in the team. I think your data scientists would greatly appreciate programs or sessions that sharpen their technical skills and enhance their business acuity. For technical skills development, we must provide both in-class and out-of-class learning sessions for teams. I would like to expand a little more on business acuity here.

Business acuity

Business acuity refers to the ability to quickly grasp a business’s situation so as to maximise the business outcome that it is driving. Some people have the ability to do this naturally, while others will take time and struggle a little. Business acuity is difficult to teach because there is no right or wrong way to approach all business decisions. In fact, sometimes the “right” answer can be irrational. Unlike data science, in which a business decision is optimised via a certain set of known parameters, the parameters of a business decision are often much more than what we can imagine.

In addition to the need to understand the domain and the business of an organization before working out the models, I have designed business acuity sessions and mentorships for data scientists. I observe that data scientists can deliver on their projects more successfully when they possess “business acuity”.


Conclusion

Building and developing a successful data science team is not easy. It definitely requires a lot of work. There are no fixed rules, playbook or best practices to share because every company is unique in many areas. You will probably need to be “street smart” enough to know what works and what doesn’t. I hope the discussion above will help you. Feel free to reach out if you want to delve deeper into this topic.

7 views0 comments