Product Management for Data Science

A common belief is that you can't apply Product Management principles and techniques to Data Science.

Why not? Here are some of the usual objections:

• Data Science is exploratory, while Product is about shipping

• Data Science is too technical for a Product Manager to oversee

• Product Management is too rigid and focused on rituals

• Product Managers have their hands full with Engineering-related work

While there certainly are nuances to consider when applying Product Management techniques to Data Science work, for many companies it is a viable and perhaps even better approach. This is especially true for Product-Technology companies, and not doing so can lead to:

• Misaligned priorities and suboptimal output

• Work product that doesn’t get used

• Missed product improvement opportunities

• Unrecognized customer insights

A Winning Combination, If You Can Get It Right

Data Science is extremely powerful as long as it is focused on the right problems and delivers tangible output. This doesn’t always happen. However, in many companies Product can be the enabler.

I once led the Product and Analytics groups at a Health Tech firm. While the teams were extremely talented and got along fantastically, they often felt like two different worlds. I felt like I had two jobs and had to constantly switch hats. More importantly, I felt there were tons of missed opportunities:

• Many of the analytics we created didn’t get used

• There was a continual sense we needed more analytics insights in our product

Our biggest breakthrough came when we embedded Data Science into the Product organizational structure. It wasn't perfect, but it helped solve some challenges and created new opportunities.

Since then, I’ve learned even more, including doing individual contributor Data Science consulting work for Product teams (partnering with some wonderful Product leaders).

Here’s what I learned:

Start with Mindset

A strong partnership is build on first recognizing that:

• Product Management can also be exploratory

• Data Science can also be about shipping

Mindset is often the biggest hurdle. Product often errs on the side of specificity (requirements, estimation, breaking work down, etc.) and shipping. Data Science can feel like the opposite. There is a heavy emphasis on Exploratory Data Analysis (EDA). Success is usually not guaranteed. You often go back to the drawing board.

Yet, we often fail to realize that Product Management is also about exploration (Discovery) and iteration (which underpins Agile methodologies). Its ultimate aim is happy customers. Data Science can help here through:

• Better customer insights

• Identification drivers of customer satisfaction or pain points

• Innovative features, including AI and Machine Learning

We also often forget that the ultimate goal of Data Science is to generate business value and insights. Product offers the following benefits:

• More insight into what generates business value

• Better alignment to business goals and priorities

• Processes and accountability to realize those goals faster

For many organizations, this partnership can be an accelerant and performance enhancer for both teams. But to do it right, you have to do the following:

Create the right process and expectations.

If only mindset were enough. You also need the right processes and expectations.

Here are five key steps:

1. Adapt Product Management processes and rituals for Data Science work. Some accommodations for Data Science work is required. Allow for more open-ended requirements and flexible estimates early in your projects, especially in the initial research, EDA, and model training phases. Do this while maintaining transparency and accountability. This doesn't mean you can carry over open-ended tickets month over month.

2. Create a framework for technical oversight of Data Science work. Data Science can be extremely technical. Ensure your Data Scientists have the right support and oversight. Technical Data Science Managers can partner with Product Managers here, similar to how it's done in Engineering teams.

3. Build integration points between Business and IT teams/roles. This is critical. Even if you get everything else right, your results will be suboptimal if you don't figure this out. People often view Data Science as a silo or only focus on a particular component, like building predictive models. Data Science work requires effective partnership and accountability across the Business and IT lines. Product should help facilitate this, but Data Science needs to be equally engaged and accountable.

4. Understand the difference between Data Science work for building features vs. generating business insights. It's important to distinguish between Data Science work for features you'll ship vs. generating business insights. While they are similar, there are some critical differences. For the former, implementation must follow the same process and rigor that's used for shipping any other type of feature. I see a lot of teams get confused here.

5. Hire/train PMs for Data Science work. Data Science continues to be shrouded in mystery at many organizations and is viewed as a black box. This often discourages open communication and transparency. I've seen Product Managers shy away from asking tough questions or topics being dismissed as being too complex. This is dangerous. There is absolutely a way to discuss Data Science in relevant business/product terms. Every Product Manager should understand how Data Science is contributing to his/her product or features and should ensure accountability.

For Product organizations, the opportunity cost of not embedding Data Science into Product is too great.

I often see resistance or skepticism around embedding Data Science into Product. However, if you apply the right mindset, processes, and expectations, it can unlock significant value — for both Data Science and Product, as well as the entire organization.

An effective partnership between Data Science and Product can help generate new insights into customers, enable new features, and enhance the value either team can bring alone to an organization.