Data visualization for beginners: a discussion around process

Two weeks ago, I moderated a data visualization panel at General Assembly‘s new SF campus. The panelists were code artist luminaries Rachel Binx, Mike Bostock, Tom Carden, and Scott Murray. There are plenty of tutorials and how-tos that illuminate the technical steps to produce a specific type of viz, but there’s not much out there around knowing what to do in the first place. So, we focused the conversation on the creative process. It was geared towards beginners, partly because the moderator is a novice herself. 🙂

A Conversation with Data Visualization Experts

Our writeup on the session is now up at Source magazine, and General Assembly’s sketch notes artist Jeremy Sypniewski produced a must-see set of visuals around the discussion (an example of which appears immediate above this paragraph, courtesy of General Assembly). I learned quite a bit from hearing these folks talk about their work…one of the particularly illuminating bits for me was a discussion of data viz as tool vs data viz as art, and the different thought processes on either ends of the spectrum.

The panelists emphasized repeatedly that data visualization exists on a spectrum. On one side are the pieces that are purely aesthetic and emotional, and on the other, the focus is purely on conveying the insights found in the data. Tom Carden, a data visualization engineer at Square, asks himself if the goal is to grab attention for a new idea, or to build a tool that will be used on an ongoing basis: “Tools need to be actionable, auditable, and they have to stand up to scrutiny long-term.” Tools should be able to accommodate new data, he said, and should grow with companies in such a way that people aren’t surprised by a difference between this week and last week.

Designs that are about grabbing attention can have a more artistic focus; there is more freedom to try new things. Moving visualizations require more of the audience’s attention, so if you’re using an animation, it has to be worth it. Motion can be an encoding itself; if the same object appears in multiple views, you can use motion to explain how the data shifts from one state to the next. One effective way to use this, the panelists said, is to give people agency – giving the user a slider may draw them in more, and help them better understand what’s going on. It’s also important that the animation be an integral part of the visualization and not tacked on as an afterthought. And of course, some browsers may not display certain effects properly, so it’s good design practice to make sure that most of the value of the graphic is there in static form.

Many thanks to GA for hosting, and to Sha Hwang, one of the organizers of the SF Data Visualization Meetup, who helped organize the event and structure the conversation.