Robert Kosara, a visual analysis researcher at Tableau Software, argues that if you want to learn how to tell stories with data, look at the Web. If you examine the right projects, you will find a classroom full of useful exhibits that use data to tell stories and examine questions—and provide lessons for both decision-makers and analytics professionals.
In a video that accompanies this article, Kosara reviews the lessons from four data visualization projects. These presentations include two from The New York Times (about political parties’ views on a jobs report in September 2012, and discussions about carbon emissions in the U.S. and China at the time of the Copenhagen climate conference).
In the third exhibit, Kosara discusses a chart-filled blog post by Elon Musk of Tesla Motors responding to a Times test drive of his company’s electric car. The fourth visualization is a Washington Post examination of National Rifle Association contributions to candidates for Congress. All use data to tell a story. And each shows how interpretations of data can vary depending on one’s point of view.
Data storytelling is still fairly new, Kosara says, and it pays to study the work of others to glean ideas of what techniques work well and to analyze why. “Looking at what the media are doing in particular when it comes to visual representation and the narrative of the building of the story, can be quite effective and especially because these are things that are available,” he says.
Below are summaries of the data visualizations with website links to the originals, accompanied by excerpts from Kosara’s video presentation.
How Political Parties in the U.S. Viewed a Jobs Report
The visualization: “One Report, Diverging Perspectives,” New York Times, October 5, 2012.
Context: In October 2012, one month before the presidential election, the U.S. reported that employers added 114,000 jobs in September. The jobless rate dipped to 7.8 percent. Coming one month before the presidential election, the news sparked debate about the Obama Administration’s economic policies.
What the visualization shows: The visualization presents three panels: a middle view where the September jobs numbers are put in context with monthly job creation and the unemployment rate. On the left is a view of “how a Democrat might see things” and on the right there is a tab to show “how a Republican might see things.”
Kosara’s take: “It shows something that I find quite common in data reporting or when people talk about data is that there are different ideas about the same data,” he says. “The New York Times here makes a very interesting point about how the same numbers, in this case this is the unemployment numbers that came out late last year, how they can be seen differently, the same numbers can be seen differently, by Democrats or Republicans.”
He adds: “If you were to present this data somewhere in a presentation to a decision-maker, you might either want to try and show both sides, and try and not to take sides, which is what this example is showing. Or if you want to be clear about which side you think the interpretation should go. Then you need to make a point why one side is the right one. And of course that should be something done in business but not in the reporting.”
Different Ways of Counting Carbon Emissions
The visualization: “Copenhagen: Emissions, Treaties and Impacts,” New York Times, December 5, 2009.
Context: Before the international climate conference in Copenhagen in 2009, policymakers discussed stemming the growth of carbon dioxide emissions and how to balance that goal with economic growth in developing countries.
What the visualization shows: This presentation charts data about carbon emissions growth in the U.S., Europe, China and India, and how they can be calculated—by geography, on a per capita basis, or per dollar of GDP. If you examine emissions as a function of GDP, China’s growing economy produces less carbon dioxide per dollar. If you examine emissions by total metric tons, however, China is projected to produce far more pollutants than the U.S. (Note: Kosara’s comments focus on one section of the visualization; the Times project also looks at the Kyoto Protocol climate treaty and projected effects of climate change.)
Kosara’s take: “These are different views, again of the same data as in the previous example. And they are interesting because they are shown quite nicely and the structure walks you through these different views and kind of gives you a sense of why these different views exist and how they impact what the decision would be going forward.
“So this is an interesting template, an interesting idea, for how to present information when you are looking at decision making and picking the path forward.”
A Disputed Test Drive of an Electric Car
The use of data: Charts presented in a blog post, “A Most Peculiar Test Drive,” by Tesla Motors Chairman Elon Musk on the company’s blog, Feb. 13, 2013.
Context: In “Stalled Out on Tesla’s Electric Highway,” on Feb. 8, New York Times reporter John M. Broder published an account of a test drive he took of the Tesla Model S, which ended with his having to hire a tow truck when the car ran out of battery charge. Five days later, Musk published his blog using data from sensors on the car to refute the Times story. Among many reader comments and online discussion about the story, Broder posted his point-by-point response to Musk’s blog. The newspaper’s public editor also weighed in, writing that while Broder “left himself open to valid criticism by taking what seem to be casual and imprecise notes” about his trip, he took the test drive in good faith.
What the visualization shows: Musk’s piece is a criticism of The Times article that includes annotated charts that show what sensors on the Model S recorded, measuring conditions such as speed and distance, cabin temperature, battery charge levels and estimated range based on battery charge, among other data.
Kosara’s take: The charts using data are effective because of the way they are annotated and part of a coherent argument, Kosara says. “It’s interesting that the way they are making this point here is that they are saying well, there are these points that were made in the story, but we have actual data to show that some of these are not actually true, or at least we can [say that the reporter] was just not taking proper notes about what he was doing.”
This approach created a lot of sympathy for Tesla, Kosara notes. “But of course it is also a bit dangerous because there are different interpretations of the same data and if you give the journalists a bit more benefit of the doubt that not all of these numbers were exactly correct, you can see that some of the patterns [the reporter] describes are still visible in the data.”
He concludes: “This is a nice example of using data in a very public way to make a very strong point, and really provides the evidence behind that point.”
The Gun Lobby’s Influence
The visualization: “How the NRA exerts influence on Congress,” The Washington Post, Jan. 15, 2013.
Context: With the gun control debate on the national agenda in the wake of the December 14, 2012 shootings at an elementary school in Newtown, Conn., The Washington Post examined campaign finance data to create a display of candidates, including incumbent members of Congress, who received and did not receive contributions from the National Rifle Association (NRA).
What the visualization shows: The presentation shows dots to represent each candidate for office and then moves and changes the size of those dots depending on whether the candidate won election, how much each received from the NRA, their party affiliation, whether they are in the House or Senate, and how the NRA rates the voting records on gun-related issues. A user can follow a particular lawmaker through the various views.
Kosara’s take: “This is an interesting way of walking through a fairly complex transformation of data to look at different ways of slicing and dicing the data. Looking the parties, looking at winners and losers, looking at the Senate versus the House and so on.”
Kosara says this presentation has lessons for business visualizations. “That’s also a common thing that you would do in a lot of business cases, where you want to present data, not just as one view, as one set of numbers. But there are different ways of looking at data that are not necessarily even about different interpretations, but just different ways of breaking the data down into smaller pieces and then trying to figure out which of those are actually interesting, which of those are useful, to make decisions and so on. You very often need to do that,” Kosara says.
He adds: “But to actually turn that into a reasonably cohesive story is very difficult and so this is a good example of how this can be done.”
Michael Goldberg is editor of Data Informed. Email him at firstname.lastname@example.org.
Editor’s note: The original version of this story referred to Robert Kosara giving a talk at an event on marketing analytics and customer engagement. That event is now a series of webinars, and information about those speakers is available here.