I agree 100% — with Data Science it’s often difficult to do the classic ROI calculation as you mentioned. At the end of the day, I’m attempting to communicate the idea that you need to always be mindful of why you exist, why the project you’re working on exists, and what the end goal is respective to the business challenge at hand. I’ve experienced many teams where the technicians (myself included) get lost in the day-to-day of just doing data science stuff, and aren’t actively working on the business — which we sometimes forget is what we’re hired to do.
As for the diverse team issue: Yes — you’re going to have essential skillsets that you need to get the job done — will be impossible without them (i.e. Data Engineering, Advanced Statistics, Software Engineering, and Implementation, etc.). My main point is: Don’t hire a whole bunch of people with the same background (i.e. PhD or MS in Stats/Bio/Chem/EE who worked in tech, for example). Bring in people who can help your problem but who may have experience from vastly different domains or were brought up in their careers through wildly different circumstances. Research shows that diverse backgrounds and experiences will result in improved projects and problem solving across the board.