How do you know you're using the right chart to make your point?

In his recently released book, Good Charts, Scott Berinato, a self-described "dataviz geek" and a senior editor at Harvard Business Review, provides an overview of how visualization tools work. Specifically, he lays out a system for how anyone can create better charts to impress and persuade others.

The book is full of exercises you and your team can conduct to pick the best charts for your projects and purposes. One of its most helpful sections explores the science of how we react when we first look at a chart. Where do our eyes go? What are our initial thoughts? Berinato distills the laws of first sight into five principles:

1. We don't go in order. When you're reading a sentence in English, you process the words from left to right. With visualizations, there are no such rules. Your eyes will go where they go. If there's a title at the top of the chart, you might not notice it or process it for minutes. 

The only thing you can count on is variability. What people first notice in a chart varies depending on the chart type and the person looking. 

2. We see first what stands out. So if there's something in your chart you want everyone to notice, you better make it stand out. What does "stand out" mean, in a visual context? It means you should use the peaks, valleys, intersections, dominant colors, and outliers of your visuals to steer viewers to the right first sight.

But you can't use all of them at once. For example, what are the first three things you notice here?

 inline image

Most viewers will first notice the blue line, the steep gray mountain, and the "outage" line. Does that mean that this is a good chart? 

It depends on what the presenter was hoping to convey. If the presenter hoped to show the connection among an outage, customer-service calls, and customer-service performance, than this is, indeed, a good chart. 

But what if the presenter's primary concern was simply to illustrate how customer-service levels were slipping? And what if the presenter wanted to make the point that the outage--while unfortunate--had little to do with it? If so, then perhaps this chart would be a better choice:

 inline image

The addition of the "service restored" marker shows that customer-service performance continued to trend downward, even after service was restored. This chart also makes it clear that customer-service levels began slipping before the outage.

And did you notice the new title of the chart? Probably not at first, since, as we discussed, your eyes are likely first to feast upon the visuals. But when you get to the new title--Declining Call Center Performance--what you see is summation of the chart's message. 

So another lesson here is to rethink your use of titles. "Instead of thinking of a title as an intro, the title should be a confirming cue," says Berinato. "It should be a tap on the shoulder that says, 'What you think you see here is true.'"

3. We see only a few things at once. In short, your most prominent visual cue has to make your main point. The chart below, created by Catalin Ciobanu for Carlson Wagonlit Travel, illustrates how to do this:

 inline image

Carlson Wagonlit Travel's clients had assumed, like many of us do, that stress rose with frequency of travel. Ciobanu's chart showed that stress could either increase or decrease; and if you traveled infrequently, there was wild variability in your stress level. The chart led to a discussion about why this was. And the title--"Who Suffers Most From Travel Stress?"--personalized the problem, reinforcing that each dot in the chart represented a real person, not a percentage. 

4. We seek and make connections. Your viewers will try to form a story out of whatever they see. In other words, your chart is creating a narrative. For example, the narrative of the second call-center chart above is a simple story: "The outage, while unfortunate, has nothing to with our declining levels of customer service."

One of the most famous narratives in the history of charts comes from Florence Nightingale's diagrams of British casualties during the Crimean War:

 inline image

The story in Nightingale's chart is that disease (represented by the blue-gray color), above all else, is what killed soldiers--more than wounds (the red-pink color) or other causes (the black-brown color). This chart is one reason Nightingale is credited with improving sanitation in hospitals. "She did her job greatly," says Berinato. "She used visual simplicity to tell a really profound story."

5. We rely on conventions and metaphors. Every choice you make has to consider the tacit assumptions and feelings of the audience. For example, in the U.S., viewers are likely to associate the color red with something negative, and the color green with something positive. They're likely to expect time, if you're graphing it on the X axis, to move forward from right to left. If you defy or ignore conventions like this, you're taking a major risk. 

There are nuances in play here too. If your chart includes both light blue and dark blue, the viewers are going to believe that those colors are representing related entities. So you need to think not only about the colors you choose, but your color groupings. "In general you should embrace, not fight, deeply ingrained conventions and metaphors when creating visuals," Berinato writes. "Flouting them creates confusion, uncertainty, and frustration, which will weaken or eliminate a chart's effectiveness."