What makes a chart good?

Good data visualization tells a story, but not in a narrative way. For example, says Scott Berinato, Senior Editor at Harvard Business Review, and luncheon plenary speaker at #TechTransfer17, if there’s a fire in your hotel room and you run to read the escape route, you’ll get more information faster from a floorplan with big red arrows than you would out of a paragraph of text telling you “Turn right, walk twelve feet, and then turn left and proceed down the stairs.”

“Visualization is 80 percent of what the mind does,” Berinato said. At the beginning of his presentation, Berinato laid out the answers to the question why viz matters. (Read about it here.)

Then -- after a brief break for a fire alarm in the hotel (an uncanny illustration of Berinato’s opening point) -- he returned to the stage to answer the big question:

What is a good chart?

The short answer is: it depends on the context. What are you trying to say, to whom, and where? Examples of bad charts are those that include superfluous information, require the person delivering it to explain the color-coding scheme, or are too small or quirkily-rotated to read accurately.

“Pretty isn’t enough,” Berinato said. “Focus on the idea, not the structure. High contextual awareness is important: what am i trying to say, to whom, and where? Design execution is important. But decisions about color, shape and type of chart follow naturally after you decide what you need to communicate.”

There are a lot of great tools out there--like Tableau and Plotly-- but the better the tool, the more training it takes.

“To get started, all you need is a browser and Google,” Berinato said. Ninety-five percent of chart drafting should happen with a pencil and paper -- not your computer, Berinato said. “Talk, sketch, prototype” over and over again until you arrive at the chart that distills the central idea you are trying to convey. Then figure out how to create it on your computer.

Where is my unicorn?

Berinato emphasized that the charting process isn’t a one-person job.

“People keep looking for the unicorn -- the ‘data viz person’ who can do it all, but it doesn’t exist,” Berinato said.

In the history of data visualization, going back more than a hundred years, designing good charts used to be expensive, hard to do, and required a team of specialists -- subject matter analysts, draftspeople, graphic analysts, and so on.

Now, thanks to the democratization of data visualization through the development of programs such as Chart Wizard in the 1990s, anyone can make a chart. That means there’s a lot more data visualization out there, which is both a good and bad thing.

“When it’s easy, convenience trumps skill, art, and thinking,” Berinato said. “Just because we can [make a chart] doesn’t mean we should. Just because it’s easy doesn’t make it good.”

The individual data viz guru is as elusive as the unicorn. No single person can replace the team approach -- a design thinker, a data wrangler, and a subject matter expert -- and the “talk, sketch, prototype” process necessary to convert a meaningless chart into an actionable visual.

“Make the team,” Berinato said. “Be agile. Sit together. Learn about each other. And use the talk-sketch-prototype methodology for getting to good charts.”