Structuring teams
AA: I often hear a lot of debate internally in organizations on how they should structure their data science capabilities. Some people are huge proponents of centralizing, while others prefer decentralized structures. Obviously, it depends on size, scale, and a number of factors, but in general, do you tend to have a preference on how you guide organizations or structure your own teams, based on reporting lines, skill sets, and so on?
KB: The answer is, it depends. There’s no one right answer to that for different organizations.
On the one hand, I’m a big proponent of decentralization – that is, people should be part of business units and not a separate unit. As a data scientist, I can criticize my own group and say that, often, data scientists like to play in their sandbox. They think, “I’ve got my cool tools, my cool toys, and my cool Python code on my laptop. I like tweaking my models and tweaking the data. Oh, I’...