How Do You Select And Prioritize Data Science Initiatives? The Best Approach Will 10X The Data Team’s ROI.
A great question came up during the office hours on Friday. “How do you select and prioritize data science initiatives?” This image represents the new problem that data teams face.
In the past, the barrier to value was getting models into production. We’ve solved the feasibility and technical viability problems. Like most of what data scientists do, addressing one challenge reveals two more. Technically feasible doesn’t always translate to valuable. Just because it could be built doesn’t mean it should be.
How does the business decide which models it should put into production?
In low-maturity businesses, the data team is often stuck pitching initiatives to the rest of the company. Prototypes must be built to prove the idea’s viability and show how the solution will fit the business need. The value must be proven up front by the prototype, but that’s not the best way to achieve that goal.
Opportunity discovery frameworks reverse the initiative flow direction so the business brings problems and ideas to the data team. Prototypes are no longer necessary. The business asks the data team to build it instead of questioning whether it’s valuable and possible to be built. The prototype phase can be moved to the end instead of the beginning.
I explain how opportunity discovery gets the data team off the endless prototype hamster wheel in this article. Once these frameworks are in place, the data team is under massive pressure to deliver. I discuss the initiative selection framework and how it relieves that pressure before it becomes crushing. The last piece is prioritization which drives some unexpected changes. I provide advice to manage those changes successfully.