So You Just Became a Data Science Manager... Now What?

With the rise of data science there has been the rise of data science managers. So what do you need to keep in mind if you wish to join these data translators that are acting as a conduit between the business and technical data teams? Going from a practitioner to a manager — your job now is to make sure that data resources are being used optimally so how do you go about doing this effectively?

Technical skill does not ensure management competency

Get your head around the fact that these are two completely different skill-sets. The skills required to manage a team of people performing a function such as data science is different from the skills required to do that function. The primary skills that you will be leaning on in a leadership role, are well….. Leadership.

You will be looking to articulate a vision of both what the team under you should be working on as well as what constitutes high quality work for that team, and then executing on that vision.

In doing so, you will be mentoring staff, providing guidance on approach, communicating with stakeholders, taking ownership of the results, project managing, expectation managing, resource managing and a whole host of activities that you likely were not doing in your previous role.

The quicker you recognize that the skills required in a managers position (even those of a team as technical as data science) are different from those of a practitioner, the better prepared you will be to adjust and excel in your new role.

However,do not forget your roots

So while you are getting your head around the fact that management of data scientists is different from being a data scientist, you are probably conforming to the reality of what managers do. And the primary purpose of a manager is to manage resources. Resources being business-speak, mostly for people but also including budgets, partners, processes, technology and any capital that comes into your orbit.

However, guess what one of the most important resources that you have at your behest will be as a data science manager? That’s right — Your data.

In a leadership role, you will need to think more strategically than ever about what your data assets are and how they can be deployed. How can they be augmented, and can other data be sourced or captured that would further enable your team and organisation to meet its goals? Being a former practitioner should give you a direct advantage in this space to other manager’s that are generalists or have come from other domains.

So start thinking about what the strengths and weaknesses are of the data sources in your organisation. How can the weaknesses be mitigated and the strengths magnified? Where does the data come from and how does it flow through your organisation and what decisions do people want to ultimately make with it?

Recognizing data as one of your main resources will allow you to start thinking strategically on how to deploy it, and allow you to make decisions based off its characteristics and ensure that you are being an appropriate caretaker of one of your organization’s most important assets.

Pick where you stay hands-on. But be hands-on

Things being as they are, and data being one of your main resources to manage, I would strongly suggest retaining some level of hands-on involvement.

The easiest way to do this is to have your greatest level of technical involvement be at the beginning and the end of the data analytics / science process.

I usually find myself perusing the data and participating fairly heavily in Exploratory Data Analysis (EDA). Doing the EDA will confirm your understanding of the data landscape that your analysts will be working in, and will help you manage the approach taken.

Likewise, as the ultimate conduit between your stakeholders and the analysts, you will also be wise to retain involvement in the presentation of the final results to ensure that the great work being done by your team is being grasped and appropriately digested by the ultimate consumers of the information. This may entail being well versed in concepts of data visualisation and having a strong ability to tell a story that engages the audience while allowing your message to be consumed. Again this step may see you back on the tools, providing guidance and helping your team craft the final product.

Leadership = inspiration

Data science is an exciting space. Convey the passion that you had as a data scientist into your leadership style. The problems that you are going to be working on are going to be exciting, and the people that you will be working with will be brilliant. So enjoy your work and infuse your team with this energy. Think about all the reasons why you got into data science (the data, the problems, the rapid evolution, the ability to make a difference) and use that vision to create the working environment to assist your team in realizing their own vision based off their own motivations.

Create and define an environment to excel.

Another differentiating area where your previous experience as a practitioner may help you as a manager is managing your analysts' workflow, and more importantly the expectations of that workflow. The data science workflow is unique and a lot of people without a practitioners background may not be aware exactly how iterative it is, and how directly it contrasts to frequently deterministic (at least in reputation) business processes that your stakeholders may be more familiar with.

I have written previously about the need to do a data questions audit before setting off and determining your approach (strategy) and how the ability to manage one’s workflow is one of the most underrated skills in data science. So what can managers do to optimize and protect their teams' workflow?

At a macro level you will need to ensure that your team is spending time on the right tasks and is carving out an appropriate amount of time for higher value adding tasks (as opposed to say, reporting). This is obvious. You will also need to ensure that they are (in the aggregate) spending time commensurate with what each data science stage entails. So for example if you are not putting a large portion of your team’s time into the un-sexy parts of data science like data cleaning, and preparation, you probably are leaving a lot of value on the table (and not doing the sexy parts any justice).

Additionally, you will also need to create an environment where the tools (or the stack) that are being used by your data scientists — best compliments their skills, while being able to appropriately address the business problems that you are looking to solve.

Finally, you will also need to manage expectations so that your stakeholders allow your team the time and space to do quality work, as well as educating your stakeholders on a range of things (where your domain experience will again come in handy) including:

The data science process and the process that your team used to reach a particular outcome (part storytelling, part managing expectations)The limitations of data (given that there is a strong bias towards the enabling, positive aspect of data) the inverse needs to be discussedHuman bias when making decisions and evaluating (statistical) data (i.e. loss aversion, recency bias etc.) and how that can be incorporated into the decision making process

So use the practitioner experience that you have, while at the same time being cognizant that success at analytics does not ensure success in management. Recognize, the asset that is your organisation’s data is an asset under your purview, and that you are it’s caretaker and have been entrusted with deciding on how it will be used.

Finally, be a good boss, more so than any other profession it is likely that future leaders will need to be either digital or data literate — so it’s possible that key future members of your organisation may be in your team. Develop good leaders from your leadership style, and try to develop people so that they can be an asset to your organisation for a long time to come — and take your role as a leader as seriously as that of a manager of data scientists.

This article was originally published here