Data Science and Product Management are like Chocolate and Peanut Butter

“Hey, you got chocolate in my peanut butter!”

“You got peanut butter in my chocolate!”

“Delicious!”

So goes the old commercial for Reese’s Peanut Butter Cups, which makes the point that some things are just better when they’re together. I think that the same can be said for data scientists and product managers – not only do we benefit from working closely together, but effective data scientists serve as the product managers of their inventions.

Data Scientists position their work in the market

When we think about the quintessential data science project, we think about using data to build a brilliant model that addresses a foundational need for the business that was previously unaddressed. Maybe its an algorithm that is too complex for most humans to execute, or maybe the dimensionality of the problem is too high to handle with conventional methods – data scientists find ways to address these problems and create ongoing value.

This goal is incredibly similar to the “Positioning” component of product management, as described by the Pragmatic Marketing Framework: “Describe the product by its ability to solve market problems”.

It’s relatively intuitive that no company would produce a product without identifying the problem that they hope to address. Similarly, no effective data scientist should build a model that does not address a need within the company.

Data Scientists elicit and gather requirements from diverse stakeholders

Saith the Pragmatic Marketing Framework about gathering requirements and managing stakeholders:

“Articulate and prioritize personas and their problems so that appropriate products can be built”

“Manage proactive communications with relevant stakeholders from strategy through execution”

The relevance of these principles to data science should be obvious – we do not build models for ourselves; rather, our models should address actual problems that actual people have and we should work with this people closely. We should strive to understand their roles in the organization and the problems that make their lives hard. In turn, we should create solutions that address those problems in order to make their lives easy.

As we design and build our solutions, we should work closely with our stakeholders at every step in the path – because we do not do their jobs, we may not necessarily know the smallest nuances of their responsibilities. Instead, it makes more sense to create a feedback loop – come up with an idea, prototype it, show it off, and let the stakeholder try it out. If you’ve missed something that they need, they’ll be sure to let you know.

Data Scientists build their own analytics roadmap

Revolutions don’t happen overnight – instead, product managers take an incremental and cyclical approach to delivering the value of their products. You might be familiar with this approach in the technology world as the Agile methodology. When it comes to products, Pragmatic Marketing thinks of the product roadmap as “illustrating the vision and key phases of deliverables for the product. The roadmap is a plan, not a commitment”.

As a data scientist, it is vital to understand that incrementalism is the best option for delivering value. Rather than building a model to perfection and releasing it only when ready, instead align periodic releases with enhancements in the delivery of value. Remember that for most problems, no current solution exists, so any solution is better than the status quo. By delivering something quickly, you gain the trust of your stakeholders, and follow-on enhancements to your solution make that bond even stronger.

Finally, by thinking of your work as a roadmap, you enable yourself to find synergies that improve your future work by not having to reinvent the wheel. Say you’ve built a way to produce high quality forecasts for one specific problem. You can then generalize that tool to build more forecasts that apply to other problems, or more detailed/granular forecasts for that specific problem.

Become a better Data Scientist by becoming a better Product Manager

Given the immense value of approaching modeling problems as a product manager might approach product development, it only makes sense that this is a set of skills worth training. Think of your models as your products, and position them for success. Here are a few resources for product management methodology that I’ve found to be increasingly applicable for Data Scientists:

The Pragmatic Marketing Framework is widely regarded as the best general-purpose product management framework and provides numerous courses, trainings, and certifications that are broadly recognized in the industry.

This is Product Management and Rocketship.FM are podcasts that explore the different dimensions of product management through interviews with leaders in the area at major companies, such as Google and Facebook.

Finally, one of the best ways to learn about product management is to sit down and talk to the ones closest to you – building a partnership with your company’s product team is a great way to learn about how your organization does product management and gain exposure to the product management mentality.

I think you’ll find that product managers come from diverse backgrounds, think in a number of different ways, and have many different opinions about the best ways to approach problems.