Q&A: Make Your Data Tell a Story
Researcher, lecturer, and graphical presentation specialist Jonathan Koomey discusses how to improve visual presentations with a few simple rules.
- By Linda L. Briggs
- August 21, 2012
Despite all the graphical presentation tools available for business intelligence, poor visual presentations are the rule rather than the exception. Today's BI software encourages the same bad practices that have bedeviled the field for years, says Jonathan Koomey, author of the book Turning Numbers into Knowledge: Mastering the Art of Problem Solving. With a few exceptions, he says, vendors simply aren't pushing the state of the art in data visualization. "It's long past time for the insights of [Edward] Tufte and [Stephen] Few to make their way into all of the most widely used business intelligence tools," Koomey maintains.
In this interview, he discusses how to improve visual presentations with a few simple rules that reflect current knowledge about graphical displays. A researcher, author, lecturer, and entrepreneur, Koomey's work focuses on critical thinking skills, environmental effects of IT, and climate solutions. He is the author or coauthor of nine books and as a consulting professor at Stanford University he worked as a researcher at Lawrence Berkeley National Laboratory for over 20 years. Koomey was a visiting professor at the University of California, Berkeley in 2011, where he is a research affiliate with the Energy and Resources group.
BITW: With more tools at our disposal for analyzing, charting, and displaying data, are visual presentations getting better?
Jonathan Koomey: That's a tough question. Let me first narrow the scope to "visual display of quantitative information" (which also happens to be the title of Edward Tufte's first and most famous book). I can't really speak knowledgeably about presentations that include video or other fancy stuff, so I'll focus on what I know.
Anecdotally, I have noticed few improvements in the general state of graphical display. I still see people using the default graphs in Excel, for example, even though those continue to be problematic. What Tufte calls "chart junk" is still more the rule than the exception, and abominations (such as bar charts with a superfluous third dimension that conveys no information) continue to be widely used.
My friend Stephen Few (author of Show Me the Numbers and Now You See It) recently gave me his view on progress in this area. There are some vendors, such as Tableau and Spotfire, that have studied graphical display and are helping users to do it more effectively, but many more still allow (and even encourage) the same appalling practices that have bedeviled this field for years. The difference is that companies pushing the state of the art understand what Steve calls "the science of data visualization." The others don't. The skills needed to build a big data warehouse aren't the same as those needed for effective display of quantitative information, but too many vendors act as if they are, and don't yet incorporate into their products what we now know about doing it right.
The key to improving the general practice of graphical display is for the vendors to retool their software to reflect the latest knowledge in this area. Once that happens, things should improve quickly, but I've been surprised by how long it has taken for the industry to take these ideas seriously. Tufte published his book Visual Display of Quantitative Information in 1981, and Show Me the Numbers came out in 2004. It's long past time for the insights of Tufte and Few to make their way into all of the most widely used business intelligence tools.
It seems that with so many tools at their disposal, people have lost touch with the real purpose of graphing data. Remind us -- what is the primary purpose?
Well-designed graphs yield insight. As Edward Tufte says, the best graphical displays "give visual access to the subtle and the difficult" and enable "the revelation of the complex."
Graphs should tell stories because that's how people learn best, so decide what stories you want to tell and choose graphs that support them. Your goal is to teach your listeners what you've learned, and the only way to do that reliably is keep a relentless focus on just a few key ideas.
A common mistake is to make graphs to display technical virtuosity, but that's a confusion of means and ends. Stay away from fancy graphing tricks and focus on the few key stories that emerge from the data. All else is a distraction.
Is poor understanding of the genesis of the data -- where it came from, and hence its quality -- part of the problem, or does it all come back to presentation?
Good presentations begin with credible data and analysis. Your data and its limitations constrain the number and type of stories you can honestly tell, so understanding data provenance is critical, but so are careful thinking about analysis methods, version control, and quality checking. The integrity of your storytelling is dependent on the consistency and accuracy of your data, so I would never say that "it all comes back to presentation."
What are a few simple rules regarding the presentation of information visually?
Here's how I boil it down:
1. Focus on a few key stories that will resonate with your audience, and design graphs that help you tell those stories.
2. Avoid what Tufte calls "chart junk," which is ornamentation that does not show data, such as clip art.
3. Review every bit of ink on your chart and make sure it conveys data -- it if doesn't, modify or delete it. A common example is heavy gridlines, which almost invariably can be made a light grey, thus creating reference lines without distracting from the data itself.
4. Never use 3D graphs unless they actually illuminate (rather than obscure) the data. Avoid bar, line, or pie charts with the extraneous third dimension -- they immediately peg you as a graphical display amateur.
5. Leave enough time before your presentation or report deadline for comments, editing, and revisions.
Finally, study the work of the masters to improve your technique and revisit these lessons every few years.
Who do you find tends to benefit most from the advice in your book, Turning Numbers into Knowledge?
I wrote the book to train young researchers back when I worked at Lawrence Berkeley National Laboratory. I gave them the book to save me time. I'd say, "Here, read this, then let's talk." It was one way to take smart young analysts and make them productive relatively quickly. Universities almost never teach students the craft skills needed to be effective analysts, but that's what this book does.
The book is written in an accessible way, but those who find it most useful are advanced high school students, university students, graduate students, and consultants at the beginning of their careers. Others have written me later in their careers to say that the book is a useful refresher even for them, in part because it points toward other resources.
One of my students at Stanford wrote me an e-mail recently about how my insistence on good table and graph design changed her professional life. Because I drilled (and graded) the students on the quality of their graphs and tables, at least some internalized the key lessons. This student has become the "go to" person within her organization on Microsoft Excel analysis and graph design. She's known for clarity of presentation in all her analysis, and she credits my relentless push for better tables and graphs (in part) for her recent success.
Are there additional resources you can point people to for better presentations?
The two best books on the science of graphical display are classics: Tufte's Visual Display of Quantitative Information and Few's Show Me the Numbers. Tufte's book is a great place to start and to learn basic principles, but some of the examples are far afield from the world of business. Few's book gives advice that will be immediately useful for a business analyst "in the trenches". For more general information about improving your communication style and content, I'm a big fan of Garr Reynolds' book Presentation Zen.