My grandfather always said, “To become a good captain, you should be a good sailor first”. Once I was able to comprehend that simple message, it became evident that traditional wisdom is something that we, as humans, should always keep with us in everything we do. Another valuable lesson I learned from my grandfather was one of “practice makes the master”, which I have always understood to mean that it takes a lot of practice before you can master anything.
Becoming a captain requires spending a significant amount of time studying a lot of different things. Yet, even when you invest years into studying, it’s not possible to become a good captain if you don’t have any prior experience in a boat. The same is true in the data world.
To become a good data manager, you should be first a good data scientist. Yes, a data scientist—not a project manager, not a master in business administration, and not a ‘master-spreadsheet-data-analyst.’ Those who manage teams should possess great experience managing human expectations as well as possess great problem-solving and project management skills, among others.
Today, there is a significant and immediate need for data scientists worldwide. However, I have observed that new recently hired team members are being managed by individuals with no experience in data science. And, on more than one occasion, I have encountered project managers and MBA professionals without any data science experience managing data scientists. In some cases, individuals that come from any organization that has been escalating positions end up owning and managing a data science team without experience. This is the perfect recipe for failure.
When inexperienced managers are preparing to hire a team of data scientists, they may think that going to a 40-60 hour training about Big Data is enough to understand the basics. Many managers may think, “At the end of the day, it’s all about people management, resource management, and timeline deliverables.” Still, this aggravates the situation.
In this scenario, when a data scientist, with at least 2 years of experience, is hired and finds out that their manager has no data science experience, they usually start searching for another job. This situation makes the data labor market even more complicated. We all know about the high rotation of data professionals.
So, here are some recommendations to successfully manage a data team:
Managers Should Have Real Data Experience
While it may seem obvious, if an origination has no staffers with people management and project management, they cannot expect an on-staff data scientist to immediately transform into the perfect manager. Building that perfect manager may require committing to a few things, such as grooming someone from the IT department or someone with database management/architecture experience to fill the role. After all, those individuals understand several crucial factors, such as working with others and following best practices. Yet, those individuals must be ready to:
- Work with large quantities of data
- Manage messy data
- Manage access to data
- Work with programming languages
- Digital infrastructure
Bring Data Experience First
For newly minted organizations, things may be a little more complicated, and many managers may want to start with junior data analysts/data scientists, typically due to budgetary restrictions. If you are about to start, bring an experienced data scientist first who can pave the way and understands what needs to be done in the way that it has to be done. Later on, you can bring juniors to support the main projects and gain experience.
Diversity is Key
When building a successful data team, don’t focus only on data scientists. Make sure that multiple roles are involved—either new hires or combined with existing organization resources. They should also be specialized in multiple fields, such as database management, business intelligence, data engineering, machine learning, etc. In addition, the more diverse the team members, the better the outcome of the team will be.
Forget About Waterfall. Embrace Agile.
This is unfortunately one of the biggest issues that I detect in many teams and managers: they all want to set up timelines and specific milestones, as many managers have been following this method historically. Data projects are not that simple and not easy to split into very specific scopes with sub-projects and sub-tasks following a very strict timeline to deliver something very specific. Data projects are less structured in their execution and can move in any direction very easily. Maintain a global target in mind and make sure that there is flexibility in delivering a solution.
