In data science applications, a key determinant of success is how the analytical problem is framed, even before any data sources or algorithms are selected. This talk discusses a framework helpful where the goal of analytics is to help organizations use data to make better decisions. Application of the framework begins by asking three key questions related to decision-making, and uses the answers to these questions to guide selection of data sources, algorithms, data visualizations as well as how the organization will use the analytic results:
- What is the decision being improved by the use of analytics?
- Who is deciding?
- What is the value of an improved decision?
We’ve found that often analytics project sponsors cannot articulate the answers to these questions, at least at the outset of the project, and that a critical role for us as consultants is to help clients refine the answers, thereby better understanding the problems they are trying to solve refine the answers. Sometimes answering these questions yield results that may be surprising to data scientists, such as that the most technically accurate model may not be the best for a given project or that adding big data to a project may be counterproductive. This talk expands on these questions and illustrates with examples taken from consulting practice. (Note: Some of this talk previews concepts to be covered in the course INFO 290: Managing Analytics Projects, to be taught at the iSchool in the fall of 2015.)
