Given the size and scope of the census data, we realized early on that trying to fit all the data within a single visualization design would be a recipe for disaster. Where we could, we wanted to boil down the data into the simplest forms that could support a range of analyses. As we were designing for a general audience, we settled on the approach of creating a collection of visualizations that present selected slices of the data. In essence, we wanted to make our visualizations as simple as possible while remaining useful, but no simpler.

Our design philosophy thus required that we figure out which data dimensions would be of greatest interest and which visualization designs and interaction techniques would best support active exploration of those dimensions. To do this, we began simultaneously exploring the data itself and the space of visualization designs.

Before crafting an interface to help others explore data, I wanted to ensure that the data was interesting enough for others to even bother. I used a number of methods to conduct my exploration, including SQL queries, Excel, and visualization systems. The most useful tool was Tableau, a database visualization system. Using Tableau, one can map database fields to visual encodings in a drag-and-drop fashion; the application then queries the database and visualizes the result. (Full disclosure: I have worked as a consultant for Tableau Software.) We were thus able to prototype a number of different approaches (for ...

Get Beautiful Data now with O’Reilly online learning.

O’Reilly members experience live online training, plus books, videos, and digital content from 200+ publishers.