There is some debate as to whether exploratory data analysis should be done before you get to slicing, dicing, and chopping (or, in the data world, before you dive into the dirty work of data preparation) or after. More and more, data workers are answering: both. Exploratory data analysis should be viewed as an innately cyclical process. Preparing your data will likely prompt new questions that necessitate more data exploration, and so forth. It’s important to not only adopt new processes that allow you to quickly explore, prepare, and repeat, but also new technologies that aid agility.
Adopting this thinking around exploratory data analysis will also prompt new thinking around who should do this work. Historically, statistical software has been used for exploratory data analysis. While statistics are a skill that many data scientists or analysts have in their back pocket, it’s often doesn’t scale across the organization. And given that the purpose of exploratory data analysis is to uncover key insights that guide downstream analytics projects, those who know the data best need to be involved with exploratory data analysis. They will have the best perspective on the data. Visually-driven tools that allow anyone to both explore and prepare data ensure that the best eyes are always on the data and that everyone in the organization can speak the same language.