Saturday School: Designing Great Data Products

Welcome to Saturday School! This morning I’m reading a short discussion by Howard, Zwemer, and Loukides about how to create data products that are clear, concise, and that meet the needs of the people you work with.

By data products they mean tools that allow people to not only obtain forecasting information that they need, but also give them a way to take action based on the insights gleaned from the analysis.

They suggest a 4 step method for doing this they call the Drivetrain Approach. The steps are intuitive: define a clear objective for the product; identify the inputs (levers) that you can control; figure out what data you need to control the lever of your choice; and finally; build predictive models based on the objective, levers (also uncontrolled variables), and data.

They present several very explicit examples, and walk through them like case studies. This pamphlet was free from O’Reilly. I enjoyed the essay–for someone at my skill level its nice to think about the possibilities of data driven optimization without having to wade through all the details (although at some point I’d like to be able to wade through all of them). I also really enjoyed thinking about how to use products like these in healthcare, specifically quality and performance improvement.