Yesterday Jorge Camoes asked an interesting question: Can Edward Tufte Do Business Charts?
Tufte is the famous data presentation guru who has brought such concepts as sparklines, data-to-ink ratio, and chartjunk into the common understanding of charting and graphic design. Tufte’s principles revolve around increasing the density of data in a display and reducing clutter and unneeded elements. Sparklines are “word-sized” graphics which condense a series of data into a smallspace, greatly increasing data densities. Chartjunk is anything added to a chart that is extraneous to the actual data itself, such as false 3D effects, shadows and other shading effects, and gratuitous clipart.
Tufte has shared his thoughts in such books as The Visual Display of Quantitative Information, Envisioning Information, Visual Explanations: Images and Quantities, Evidence and Narrative, and Beautiful Evidence
. He also has a popular seminar which he presents in major cities across the country; I attended in Boston back in about 2003.
Many of Tufte’s principles have been incorporated into modern graphical packages. There are at least two commercial packages which add sparkline capabilities to Excel: SparkMaker by Bissantz, and MicroCharts by BonaVista Systems, and an open source Excel utility, Sparklines for Excel by Fabrice Rimlinger. Unfortunately a great deal of chartjunk is also marketed as Business Intelligence and Data Visualization systems. Among these are Dundas, .net Charting, and ChartFX (links not included on purpose), and the list goes on.
Jorge claims that “[u]sing aesthetics to improve function is probably the major contribution of Edward Tufte to the display of quantitative information.” Tufte’s aesthetics incorporates a minimalist approach to graphical design, and while this approach has great merits, it is not supported by scientific evidence. Sometimes Tufte takes his principles to the extreme, as in his boxplot replacement which uses single lines offset b the line width to indicate the quartiles. Other authors, for example Tukey, Bertin, and Cleveland, have provided evidence-backed principles for information display in studies about human perception and cognition.
Tufte also emphasizes a chart being the end product of analysis plus graphic design, where the chart is painstakingly laid out for publishing on a high-quality paper medium. For this, Tufte’s tool of choice is Adobe Illustrator. Tufte doesn’t say much about Excel, other than exhibiting mild surprise that his students are able to apply his principles while using Excel. On the subject of PowerPoint, however, Tufte is a harsh critic and is ready to throw out the baby with the bathwater.
For business purposes, charts have to be effective, but they also have to be dynamic, interactive, and fast. All three of these requirements rule out Illustrator and point to a program like Excel. Excel enables rapid generation of charts, it allows linking of charts to dynamic external data, and with a small (or large, some would say) amount of effort, Excel can produce interactive and even animated displays.
I would hope that readers of this blog have found useful techniques that accomplish these objectives while still maintaining some aesthetic appeal. In addition to my blog, I would point readers to the following for discussions on aesthetic charting techniques:
- Jorge Camoes’ Charts
- Junk Charts
- Me, Myself and BI
- More Information per Pixel!
- Information Ocean
- Pointy Haired Dilbert
- Charts & Graphs Blog
- Perceptual Edge
I’d like to point out two authors who are particularly useful when it comes to business charting. The first is my colleague Mike Alexander, with whom I’ve presented a dashboarding course, and who has written Ten Chart Design Principles, a guest post on this blog. Mike has written a number of books on data analysis using Excel, Access, and Excel pivot tables. I will forgive Mike’s foray into Crystal Xcelsius: heck, it seemed like something worth exploring, I even tried it (but I didn’t inhale). However, Mike’s Excel Dashboards and Reports For Dummies is a comprehensive treatment of Excel’s capabilities as a dashboarding platform, and describes methods for creating displays which are comprehensible, dynamic, and interactive.
The second author is Stephen Few of Perceptual Edge. Stephen has written two books, Show Me the Numbers: Designing Tables and Graphs to Enlighten and Information Dashboard Design: Displaying Data for At-a-Glance Monitoring, in which he incorporates the aesthetic principles of Tufte and the state-of-the-art knowledge in the fields of perception and cognitive science into easy to follow and, most important, easy to implement principles for effective business information display. Stephen’s third book, Now You See It: Simple Visualization Techniques for Quantitative Analysis, uses these principles to describe exploratory graphical data analysis.
Dan says
Yes. Unfortunately for all the praise Tufte gets when I tried to show charts at my Fortune 50 employer that conformed to his aesthetic they were not well received. People that I’ve met in business tend to think Tufte’s principles are crap (yes, that’s the word they’ve used).
So I’m back to following the corporate model (and honestly their charts aren’t that bad). Oh well, the book is pretty on my shelf….
Jacques Warren says
Very interesting question about Tufte’s ability to do business charts.
Reminds me of an example I use when I talk about the relationship between Web design and usability. You want to see the easiest web site to use in the world (OK, besides Google)? Have a look here:
http://www.useit.com/
It’s Web Usability guru Jakob Nielsen’s site. Obviously very easy to use, but obviously…
Jon Peltier says
Dan –
One problem is that many in the Corporate world have never known good charts. They were raised from an early age to think that pie charts were the best, and they’ve seen so many pie charts, they confuse familiarity with effectiveness. The sentiment seems to be, if 2 dimensions are good, 3D must be better. We paid for a program that does gradients, dammit, I want to see gradients.
It’s hard to get people to open their eyes and consider changing their minds. I had a project a number of years ago, subbing for a small contractor that was doing an analysis for a big company. They wanted to automate a series of charts with associated labels and features. Well, you can’t do much with labels, target lines, and the like with 3D charts, so I proposed modifications to the chart formats themselves. I also introduced a color scheme that matched the end user’s corporate logo, instead of the default Excel palette. It was going along well, when finally the committee got together (not just individuals) to review progress. They came back with a question for me: “How much to get all the formatting back the way it was?” So I walked.
As Tufte says, “It’s difficult when big systems are designed because there are a lot of thumbprints left on them, and things just get negotiated away. It’s very hard to have a real impact.”
Jon Peltier says
Jacques –
Nielsen has a lot of good advice on usability, but I don’t think he’s strong on aesthetics. It is possible to combine usable and not-ugly in the same web site, presentation, or publication. Jorge said, “there is nothing wrong with using 3D effects, textures, and all the decoration in the world.” So use them, but just a little. A very little, because a small amount of decoration goes a long way. Just “don’t design technically incorrect charts”, and “don’t hide the patterns” (as Jorge also said).
Gabriela Cerra says
In bussiness, at least, in the operations level the most important things in a graph are
-Timing
-Acurrancy
-Flexibility
I have seen Excel 2003 default graphs doing pretty well during a presentation.
Jon Peltier says
Gabriela –
Excel charts are effective for displaying information when used properly. They can be created quickly, they are as accurate as the data, and, well, the flexibility depends on your creativity.
While they don’t need to be pretty, they also don’t need to be ugly. You could easily enough define a few custom chart types that you can apply in one or two clicks. I don’t use this feature: like a good engineer I tinkered with it until it broke. But I have one or two VBA routines that turn the ugly gray background to white and lighten the various lines. In fact, just doing that only takes a few seconds manually per chart.
Jorge Camoes says
Jon, nice post and great comments, as usual.
Some time ago I designed a sales dashboard that included a colorful bubble chart and other Excel defaults. The final bubble chart was much cleaner: gray circles (bubble area color set to none) and a color-coded “active” bubble. But the users wanted their colors back, and I had to explain why pointless color-coding would harm the display (and the task).
This example illustrates two things: a) a pure minimalist design will always be rejected by the corporate culture; b) it may be accepted if you are able to show its obvious advantages.
I feel that users are emotionally attached to (primary) colors, and that’s the first thing we must break. Once they become aware that the right color coding actually simplifies their work they may be willing to accept 2D charts and even scatterplots….
Jon Peltier says
Jorge –
Primary colors, pie charts, library paste. Unfortunately most people’s charting expertise has not advanced since second grade.
Jacques Warren says
I second Jorge here. I have had that “fight” several times, and as a consultant, I have also learned that client satisfaction is often a fine balance between “Best Practices” and “I Want This”.
On his blog, I asked Stephen Few “The question is why people fall for the bells and whistles? And why great visualization principles are not so self-evident that people would “fall” for them instead?”
His answer: “I’ve found that it’s easy to convince all reasonable people that the silly stuff doesn’t work by showing them examples of good and bad practices, demonstrating how things work, and taking the time to explain why. (…) Given a little instruction, everyone gets it. They might still feel tempted from time to time to gorge themselves on a pie chart or two, but they know they’ll regret in the morning.”
It’s hard to disagree, but I also appreciate the argument that pie charts, for example, have been around for so long, that everybody gets them, whereas we still need to explain bullet graphs since they are not really self-evident.
In my work, I have found, as Jorge puts so well that “a pure minimalist design will always be rejected by corporate culture”. Gee! I still have to create dashboards with Xcelsius 50% of the time…
Anonymous says
@Dan – I wouldn’t say that Tufte’s ideas are crap. I would say that taken to the extreme, you may have a limited success in the corporate sector.
@Jorge – I think the answer is in the middle. Clean charts that tell a story should always be preferred over charts with background images and excessive bling. The power is when you can take a mixture of Tufte, Few and Cleveland instead of taking Tufte to the extreme. One chart I recently presented that had the most impact was a scatter plot.
@Jacques – You can only do so much and recommend what is the right/optimal choice. In the end, if they want a circle cut into triangles, they get a pie chart. Just like if your boss wants the chart with excessive chartjunk, then you have two options. Do it or quit. It’s hard to coach and steer management (especially executives) to certain chart designs unless they are very data centric. Also, Xcelsius can be used to create effective charts, you just need to do a lot of fine-tuning- See Datapig Technologies. Mike has a lot of tutorials.
Jon Peltier says
Anonymous –
It’s possible to do effective things in Xcelsius, but they make it very hard. I did a number of Xcelsius models for a client a while back. I did some cool prototypes in Excel, but they wanted them in PowerPoint and PDF, so I did the Xcelsius near-equivalents.
I ran into issues with Xcelsius’ incomplete treatment of Excel formulas. More important, most of my chart effects rely on combination charts, and these are just undoable in Xcelsuis. Waterfall charts are not possible, because you can’t make a chart series completely invisible (or at least you couldn’t at the time, and I’ve heard that Crystal curtailed all Xcelsius development when they bought it).
Tony Rose says
Sorry Jon – not sure why my comment came through under Anonymous. I posted it.
Thanks,
Tony
Sandro says
Hi Tony
An Xcelsius Present related question:
I like to show two lines on the X-axis, one refering to the quarter (i.e. Q1, Q2, Q3, Q4) the second referring to a value per quarter (i.e. 10, 202, 50, 70).
In Excel I would use carriage return [char(10)] to break the line.
Xcelsius Present returns an error telling it is not compatible with the char function.
Does anyone know how to solve that problem?
Thanks.
Sandro
Graphs says
Nice attractive methods of executing charts has been posted here on the this page. Appreciable efforts has been done.