Launching and Scaling Data Science Teams: DS Individual Contributor Playbook

If you can’t translate them, you can’t succeed in the field.” A constant challenge between data scientists and business stakeholders is the lack of knowledge overlap: the business stakeholders don’t know what’s technically possible, the data scientists don’t understand the business well enough to solve the problems that matter..You embed yourself in the business to build that knowledge..The next step is acting on it — by proactively seeking out opportunities where a data science solution can make huge difference to the company bottom line..Again, this is not a lesson Kaggle teaches you.Any tech thinkpiece is not complete without an XKCD reference..“In CS [data science], it can be hard to explain the difference between the easy and the virtually impossible.” Think about your company’s challenges from this perspective, and consider it your responsibility to find technically easy solutions to important problems..This point hits home for me personally because I don’t think I learned it my first few years of work..Especially at a gigantic company that already has more columns of data than one person could ever possibly know and a year’s backlog of tickets, it’s hard to think that an idea you came up with could actually be the best use of your time.I finally figured this out during my second year writing recommendation algorithms for Wayfair..We had the idea to show segmented zip code specific sorts — the idea being that even if we’d never seen a person before, the zip code they came in from might be something we could leverage; they probably had preferences more similar to others in their zip code than the general population..Unfortunately, we weren’t logging this data anywhere and had no way to recreate it..This bummed me out, but it didn’t occur to me until one of ETL engineers said “We could just log it if you put in a ticket” that it was even possible to change my data set..A week later, we were logging data, two months later, we rolled out the new sort and it was a big win.The takeaway here is that you need to be well enough in tune with the DBAs/ETL team to know what’s possible to add, aware of what the business problem is, and creative enough/realistic enough to know how logging a new piece of data could allow you to solve something for the first time..Business knowledge will get you further than technical knowledge, but only if you follow through your business knowledge with action.Phrased another way, a consistent complaint from business stakeholders about their data science team is that they’re liable to take the overly academic and scientific approach to a problem.. More details

Leave a Reply