Fact and Fiction (and How to Tell the Difference) in Data Visualization

This post is a response to “How Might We Visualize Data in More Effective and Inspiring Ways?” Read more of the conversation here.

What’s the difference between saying, “I’m pregnant” and this:

On the one hand, someone who hadn’t seen this visual before might find it unintelligible. The image itself doesn’t say what it’s for, nor how to read it. On the other hand those who know what they are looking at—a pregnancy test—have an immediate, objective status reading. Since they are simply looking at an object, they don’t have to come up with response. Instead they can absorb the information and experience it in their own personal way.

In the budding field of evaluation and impact management, the value comes from communicating the impact—to your team, customers, investors, and others with a stake in what you do. It’s key to make the time to put the results of impact analysis into a form your audience can absorb and use. While it takes a little practice, doing so can be very powerful.

When we introduce impact data to pictures, we tap into a deeply-rooted way humans understand the world, and we can inspire action—which is ultimately the point of measuring impact. But without a factual key or other clear guide that makes the image self-explanatory, data visualization can obscure understanding. All too often such visualizations are used to deceive. This can be hard to see through, partly because people seem to be naturally more skeptical of other people than we are of numbers or images.

One of the most common ways data visualization can be manipulated is by adjusting the scale. For example, CRATX is a publicly-listed fund that invests in debt securities certified under the Community Reinvestment Act, which supports investments in affordable housing and other community development work. How well does CRATX perform?

Here at first glance the returns look huge, until you realize the Y-axis is showing a $0.08 gain.

Another common way data visualization is manipulated is by adjusting the time period. Above, the timeframe was three weeks. Below, CRATX’s percent gain/loss is shown relative to that of the Dow Jones and NASDAQ indices. The time scale is still the same, and CRATX appears to be dwarfed by the two. We are still missing the bigger picture.

Below, if we look at CRATX relative to the Dow Jones Industrial Index and the NASDAQ with the time scale expanded to show June 2007 to October 2009, we finally start to get a good perspective on CRATX relative to the rest of the market. We see that CRATX has been a great, and stable, performer.

The next step would be to identify “peers” of CRATX and plot its relative performance to similarly classed funds.

Things look a lot different over a short period of time than a long period of time. Here are screenshots of the progression by Gapminder (which Engin discussed earlier) of Infant Mortality vs. GDP/Capita, from 1960 to 2000. (Check out the video version.)

Notice how many countries become richer, and that infant mortality decreases significantly, but central and southern Africa are somewhat unchanged. We see strong correlation of infant mortality to income, somewhat regardless of population size.

Also impressive on the Gapminder shots is that the visualization contains four dimensions of data, and five if you press the play button. If all of this data were thrown into a spreadsheet, we could find out the precise number and source, but the patterns would be hard to grok, and the takeaway would not be nearly as memorable.

On a final note, one dimensions of data missing from a lot of visualizations is the personal relativity. With online interaction, you can actually chart where the user stands in relation to the graph you are displaying. Not unlike a “you are here” dot on a map, if you simply enter where you live and your income bracket, you could be personally plotted onto the infant mortality vs. GPD per capita charts.

We hope that keeping these tips in mind will help you understand and create useful data visualizations. Over the next few months we’ll have more examples from SVT’s collaborations with Nonprofitmapping.org, the guys behind Human Translation, Cisco, and we’d love for you to share your examples here.

• What’s the best example of data visualization you know of?
• What’s the worst example of data visualization manipulation you’ve seen?
• What’s your favorite, cost-effective tool to create data visualizations?

Sara Olsen is Founding Partner of SVT Group, an impact advisory and information systems company. Robert Bailey, a 15-year design veteran, recently left Google to co-found agoodproject.com, which will debut later this year.