How To Create The Best Data Science Team
Data Science is bringing value for almost every business out there, and it is rapidly becoming the primary engine for product innovations and organizational progress. With the right insights, accurate predictions, and agile infrastructure, data-driven organizations will be innovation engines in the future to come and market leaders to look up to.
Unquestionably, we are living in an era of data explosion. Nowadays, more and more companies are setting up their Data Science teams with the idea to leverage the data they have already collected. In that process, they have to think about how big the team should be, what skills future employees need to have, and how to deal with the mix of different educational backgrounds in the same team. Soon, one will find out that setting up a capable Data Science team, it is not that simple as it looked like when the idea has risen.
How to start?
You need to have useful data as the basis for the future Data Science team to work on and all the work has to be use case driven. That is precisely where you have to start.
For different use cases, you might need a vastly different team structure. I know that it might be tempting to create a universal team, but that is really hard and open to many vulnerabilities. In the end, hiring for everything usually ends up hiring for nothing, and that’s not a great way to start your team. If you have trouble with the starting point, think about hiring some external Data Science company or Consultant to help you identify your problem first.
After you identify your problems and set your core use case targets, the next vital topics are connected with the integration of your future team with the current organization and senior stakeholders. They usually underestimate the level of change required to make it work. This connection between your team and the rest of the organization is crucial in delivering excellent and quick results.
Naturally, the larger and long-standing organizations are slower and more reluctant to changes. They’ve been in the comfort zone for too long and even a slight signal of possible changes will encounter huge resistance. On the other hand, just because you are a startup doesn’t mean you can change everything rapidly.
The critical ingredient for successful adoption of the Data Science team is a culture willing to change and learn from the data.
So while you are identifying core problems around which you are going to building a Data Science team, invest some time and effort into the organization's cultural change.
We are also going to mention the famous debate on whether Data Science should be a centralized or a distributed function. This topic has been around since the beginning. One of the possible solutions to this debate would be dual reporting in an agile matrix management structure that will solve the tension between mission-oriented and functional teams. In practice, that means that Data Scientists should be reporting both to the Data Science Manager and Functional Team Lead.
How important is domain knowledge?
Indeed, how important is it? Since yesterday, there was a belief that only PhD individuals can work in the field of Data Science. Well, the domain knowledge is beneficial, no argue about that. On the other hand, I want to stress out that the entire team doesn’t need to have experience in the domain your business is in nor to have PhDs. Of course that the more specific and technical your area, the more likely that you will require people who understand it.
Individuals with different backgrounds can spark great ideas through collaboration and critical thinking.
On the other hand, people who don’t excel in a particular area but instead have a broader spectrum of knowledge are often overlooked. They can be tremendously useful, especially in the early phases of a project - they have experience doing a range of roles and can wear different hats comfortably making things happen easily. Be aware that your requirements will vary over time. Having a person in a group with the right level of understanding across data, business, and technology will help you piece the puzzle together.
Leadership before everything
If you need to hire someone first, it should be one who at least understands how all the disciplines mentioned above fit together. Be aware that that person will be the core to build up the team. Without the one who knows how all of them should work together, the real potential of the Data Science team cannot be reached.
A lot of companies hiring a technical data scientist with the idea to migrate later to a team leadership position. It is essential to mention that the technical leader and a team leader are two completely different roles. If your organization has a great sponsor with a strong understanding and vision on what Data Science should look like and at the same time have the capacity to lead, a group of technical team members can be a useful resource. However, there is a vast list of benefits to hiring someone who understands the value of the business and leadership within a technical context and can dedicate all the efforts into leading the Data Science team.
The second hire should be the one who can supplement the missing skillset, vastly in technical areas, and to fit perfectly in the already established team culture. The same situation is with the third one, the fourth one and so one. Be aware that it will be tensions between team members, and that is normal. When persons and professionals come from different industries, the pressure is something that can make or break the deal. Make sure to create positive tension between team members. If it is leveraged well, you will be able to build innovative solutions that are robust and scalable. Finding business candidates with a good understanding of the technical side and vise versa will help maintain a healthy balance and build a sustainable and long-running team.
One more important topic should be mentioned here.
Nobody wants a low performer with low trust and we all want the high performer of high trust in our team. But pay attention - the high performer of low trust is a toxic leader. Always hire and keep someone with medium or even low results but with high trust instead of first one. Value more trust over performance. Toxicity in your team is bad and it will eventually destroy the team and the whole organization. If you don't know how to find one in your current team, here is the formula:
Go to any team and say who’s the asshole. They will all point to the same person.
Go to any team and say who do you trust more than anybody else, who’s always got your back and will be there for you. They will also all point to the same person.
A natural leader who’s creating an environment for everybody else to succeed. They may not be your most individual highest performer but that person you better hire and keep on your team.
Data Science Team Roles
In a nutshell, the Data Science team should have the following roles with respected responsibilities.
I Business Analyst - analyzes an organization and business processes. Help the team to be more in line with business requirements and to explain any open questions regarding business workflow.
II Data Analysts - monitor processes, evaluate data quality and monitor production model. Having this role in the group allows more senior team members to focus on innovation instead of maintenance.
III Data Engineers - building and maintaining a scalable and technical infrastructure required for modeling, predictions, and analysis. Create and maintain databases, pipelines, and production processes.
IV Data Scientists - owns the modeling process and work on technical development. They take input parameters from the product team leads and work on building tests, models, and evaluating performance.
V Data Science Manager - coordination of the devs, analysts, scientists, and communication with relevant stakeholders. Guides the process and available resources.
Top 3 skills to look for in the future candidates?
There is no Data Science unicorn - a creature who can write code, create modeling algorithms, manage the product roadmap, and communicate results to stakeholders. Data science is comprised of multiple disciplines, and we all have to be aware of that. The team members should have leadership skills, the ability to communicate well and to have outstanding data engineering abilities. When scaling comes in place, you will need to think about storing data properly, sampling at appropriate intervals, and building features that will go into the model.
I would like to extract 3 main skills to look for in the future candidates for your Data Science Team:
- Problem-solving skills and resourcefulness - great data scientists focus on the problem rather than the tools. Understanding, breaking down the problem, and come up with the right methodologies is the actual value that a data scientist should bring. A wide range of experience is particularly challenging to find due to the immaturity of the industry, so resourcefulness should have a huge value. It takes practice, quick failures, and perseverance to learn from mistakes, adapts quickly, and progress in every step of the way.
- Passion for learning and experimenting - when we say science, we meant Experimentation and Learning. Data Scientists are continually experimenting and learning from failures. The field is still pretty young but evolving rapidly. Try to find candidates who are passionate about constant learning, testing new ideas, and the ones who are not scared of mistakes and failures. Fail quickly, learn, adapt, try again. This mentality will be must-have in the future to come.
- Respect for the other disciplines - Data Science consists of many different disciplines, and to be great data scientists, you have to be able to bring those things inside one team. Cross-functional teams are Data Science reality. They are a combination of all the different components that make a solution work. Try to find candidates with experience working with people across disciplines and backgrounds. They are more productive when it comes to delivering solutions. Team members have to understand and appreciate the skills and background from the other side, and that will create that positive tension and synergetic effect inside the team.
Building a Data Science team is hard. Building a Data Science team will bring a lot of value to your business if done correctly. With that been said, as a business leader or a project sponsor, it is essential to know the data you own and how that data can support your vision. When you know that, then you need to find the right people to transfer your dreams into goals and reality. Don’t rush the process. Put in the time to build a strong, stable, and proper team, and that will ensure delivering the wanted results for you and your organization.
Until next time,
Nikola Basta
For more interesting topics follow me on Linkedin
