Sharpen your points with cleaner charts and graphs

Every systems manager has searched for just the right chart to sell a program or
project, to persuade our audience. We've all had to display complex information in a
manner that will make analysis easy and straightforward.

This is not a new problem. William Playfair's 1786 classic on skyrocketing debt was the
first such chart, the start of a 200-year tradition. He created it as a cornerstone of his
polemic against the British government's policy of financing colonial wars through deficit
spending.

The right display can make a compelling argument. It was Dr. John Snow's plotting of
cholera deaths that led to identification of the Broad Street water pump as the culprit in
the 1854 London cholera epidemic. Charles Joseph Minard's famous 1869 chart combined data
maps and time series to depict the devastating losses of Napoleon's Russia campaign of
1812.

So how do you organize a lot of data to make your point this dramatically?

The first step is to eliminate "chartjunk," the term Yale University's Edward
Tufte invented for the content-free decorations added to charts and presentations.
Chartjunk and other visual abominations abound in our spreadsheets, Web pages and
newspaper stories. Instead of conveying information and aiding analysis by the reader,
cute and clever graphics all too often try to impress or distract the reader.

The artist may have more sinister motives for including chartjunk, such as concealment
of unfavorable information or deficiencies of data. Or the designer may assume that all
statistics are boring and require dressing up. Thus the presence of chartjunk may reflect
the belief that the reader is not capable of thoughtful analysis of the information and
must be entertained if the data is to be conveyed. Authors and editors should attack
chartjunk with the vigor they use on superfluous prose.

Chartjunk is the ink applied to a chart that adds nothing to the information content.
Like the signal-to-noise ratio in communications, Tufte uses his "data-ink
ratio." He defines this ratio as the "proportion of a graphic's ink devoted to
the non-redundant display of data-information." If you removed all the ink used that
did not convey information in a chart, the chart would achieve a 1:1 data-ink ratio.

An electroencephalogram is nearly pure data-ink; the multiple lines of readouts could
not be erased without removing data. In contrast, grid lines and tick marks on graph axes
are nearly always uninformative. Their subordination or deletion tends to enhance the
remaining information.

Unfortunately, most graph paper has built-in chartjunk. Dark gridlines obscure data
points plotted on graph paper. Tufte recommends using the reverse side for your plots. If
both sides have printed gridlines, throw the paper away.

Declutter your displays by removing grids, tick marks and boxes. Those that remain
should be light and unobtrusive. Reference curves (connecting datapoints) may or may not
give the data structure and flow. When in doubt, leave it out.

Crosshatching and granular patterns can create distracting optical illusions called
moire patterns. A bar chart with these patterns can shimmer and wiggle before your eyes.
Phantom dots can appear in the white spaces between thick black cell borders. Shading and
lines should be in light colors to focus attention on the data.

Colors can give a presentation visual appeal at the expense of the information and
present conflicting or ambiguous distinctions. Few of us understand the order of colors in
the rainbow, and fewer still can intuitively interpret spectral colors as sequential
gradations. It's no coincidence that the phrase "all the colors of the rainbow"
suggests chaos, not order.

Colors can have widely held associations that are culturally dependent. Red in Western
society is associated with stop lights and stop signs; green implies that you can proceed.
But in the People's Republic of China, red is the progressive color, whereas you must stop
at a green light.

I suspect a subliminal message when one side of a political chart is red or pink,
suggesting that this side is communist. Of course, those who are red/green colorblind have
great difficulty interpreting those colors in a display, as well as on the highway.

All these insights and more await you in Tufte's remarkable little book The Visual
Display of Quantitative Information. He provides many memorable pictures and figures
to illustrate his points. (Yale has a terse homepage for him at http://www.cs.yale.edu/HTML/YALE/CS/Brochure/faculty/tufte.html.
  Perhaps Tufte will write a book on designing Web pages.) If you need to make
your charts clear and influential, you need to read this book.

Walter R. Houser, who has more than two decades of experience in federal
information management, is webmaster for a Cabinet agency. His own Web home page is at http://www.//cpcug.org/user/houser/.