Earlier this month, we attended “Finding and Telling Stories Using Data Visualization,” a presentation by Kaiser Fung, of JunkCharts fame. We were lucky to see Kaiser during his visit to San Diego, and enjoyed meeting the man behind the blog. Plus, he shared some key insights for creating “good” dataviz. Want to know the dataviz tips?

dataviz seminar
Done right, visualizations are more impactful. However, done wrong, visualizations can make data even more confusing!

Bad dataviz

As many as 99% of the submissions Kaiser gets to his site are what his readers call “crappy charts.” But rather than making fun, he wants us to agree on what is good and bad, with specific terms and language to communicate with each other. Bad data visualizations:

  • Mislead
  • Misinform
  • Confuse
  • Disorient
    Kaiser shared four key ways in which dataviz gets “bad.”

  • *Splitting information. *Whether it is across web or print pages doesn’t matter. Forcing your users to flip back and forth to get the whole picture is bad dataviz.

  • Redundancy makes bad dataviz. Common examples of redundancy include:

    1. Repeating scales
    2. Unnecessary grid lines
    3. Extra value boxes
  • Bring your own calculator” visualizations. Requiring users to pay close attention or to hover over every part of a visualization to get the whole story is not helpful.

  • Pull out your timer” viz. The purpose of a chart or infographic should be clear. Close examination or an excessive amount of time should not be required to figure this out!

“Bad” dataviz example

Kaiser led his session with a chart about what Americans call sweetened carbonated beverages. It turns out, there are significant geographic patterns in the answer to this question.

soda chart from Junk Charts

The original chart is on the left, and while it was well received at the time, can be considered “bad dataviz.” Reasons include:

  • The dots are big and overlap
  • What is the answer in all that white space?
  • Lack of clarity with overlapping dots of different colors
    Later in the seminar he revisited this chart, to discuss how it was improved in later years. As you can see, it addresses the “bad” dataviz points above. This led off his discussion of good dataviz.

Good dataviz

The language Kaiser wants to unite us around for good data visualization include:

  • Precise
  • Obvious in meaning
  • Interesting
  • Accurate

Dataviz rules of thumb

Make it thick

Kaiser’s first rule of thumb comes from Edward Tufte. His concept of “making it thick” refers to the ratio of data to ink. What is conveying data? Hopefully it is data points and not unnecessary grid lines or cartoon figures.

Efficient data visualizations

Making dataviz efficient means using visual elements that work on their own. That means using charts that incorporate scales. Or avoiding pie charts that are not obvious.

Easy to read

Kaiser wants to make it easy for users. He recommends charts that show insights and how they relate the bigger story, not overly nitty-gritty visualizations. He warns this rule of thumb can be the hardest to achieve in the age of big data.

Make it scream

Interestingly, Kaiser offered the opportunity to break some dataviz rules of thumbs with this item. His example showed small multiples where one particular chart should really have a different scale. Normally, this would be a no-no. But keeping the scale the same or plotting a tiny country on a world map is “burying the lede.”

Speak directly

Don’t let interesting _looking _charts or imagery get in the way of telling the story and making the insights obvious.

Knowledge in the head

“Knowledge in the head” is the term Kaiser used for the sort of mental baggage we bring to viewing a chart. For example, red and blue might seem like innocuous colors, but they can represent political parties to some users.

Interactive

During the seminar, we got the distinct impression interactive is almost always better. Good thing we include interactive options in our library! However, there were some clear exceptions to this rule of thumb, including:

  • When it creates performance overhead, slowing down the delivery of a chart
  • When it makes it take users longer to figure out what the chart is about or distracts
  • When it is requirement to use the visual cues of the chart

What to remember in your dataviz

Toward the end of his presentation, Kaiser summed up dataviz using the following three questions:

Q- what is the practical question?
Is it obvious, interesting?

*D- What does the data say? *
Does it answer the question? Is it accurate?

V- What does the chart say?
Are the visual elements contradicting the data? Is it the right chart for the data? Are rules of thumb followed?

Just V is not enough. Truly good dataviz requires a secondary level of thought and critical eye, that also addresses Q and D.

Attend a dataviz seminar

Kaiser’s seminar was great. And we would not want to reveal all his tricks of the trade, so be sure to follow him on Twitter to find out when he will be visiting your neck of the woods.