In praise of a good graph

I see hundreds of examples of bad data graphics every month. TV stations track audiences with no points of reference. Lines in charts emerge from the left and flow to the right, leading nowhere in particular. Bar charts, like silhouettes of...

I see hundreds of examples of bad data graphics every month. TV stations track audiences with no points of reference. Lines in charts emerge from the left and flow to the right, leading nowhere in particular. Bar charts, like silhouettes of cityscapes, produce mind-numbing results. Chart-crap is one of the downsides of being in the media planning and research business. (I’m anticipating the audience-trending graph for Temptation Island, colour coded by age break, signifying nothing.)

That’s why I’m a sucker for a good graph. And I found a really excellent example at the bottom of the front page of the National Post’s Jan. 9, 2001 issue. The graph consisted of a simple black line on a blue background, entitled ‘The R-Word Index.’

The R-Word Index quantified the number of articles published by the National Post and The Globe and Mail that incorporated the word ‘recession’ over the last 12 years. The graph plotted R-Words over time. I could quickly see the extent to which R-Word usage ebbed and flowed relative to our country’s economic health. The R-Word line moved up to its peak position in 1990 just as the recession worsened. The line moved downwards in 1997 as the economic health of the country improved. There was an up tick in 1998 and a pronounced drop in 1999-2000.

Of particular interest, and the reason why the graph was timely, was the distinct upswing detected in the R-Word Index graph as we move into 2001. According to the R-Word Index, we can soon expect a recession.

The article that accompanied the graph made reference to a similar analysis conducted by The Economist, and quoted from The Economist’s Jan. 6, 2001 issue: ‘…an increase of this scale has always signaled the start of a recession.’

The best data graph triggers thought. As a result of seeing the R-Word Index chart, I began to think about correlation between R-Stories and recessions.

The R-Story count seems to precede the recession and so I detected predictive value. If there is predictive value, is causality at work here? In other words, is it possible that recessions might be caused by the upswing in the press coverage of the possibility of recession? Might the media, in the act of reporting, actually be creating the news? Is this an example of the act of observation causing an alteration in the experiment?

For those of you who share my fascination with the presentation of data, I direct your attention to a book entitled The Visual Display of Quantitative Information by Edward R. Tufte (1983, Graphics Press). The author notes that up to two trillion images of statistical graphics are printed every year around the world and that most of this output can be improved. In addition to providing examples of the best and worst in visual data display, the book provides practical principles of design theory.

Tufte displays, on page 41 of his book, what he considers to be the best statistical graphic ever drawn – Minard’s depiction of Napoleon’s march on Moscow in the winter of 1812. A broad gray line, representing 422,000 soldiers, moves eastward across the map. The thickness of the line denotes the army’s size and by the time the army reaches Moscow, which is identified by a time scale, the line is one-quarter of its original thickness. Once the army turns westward, the line turns black and proceeds to shrink to a thread-like state as only 10,000 soldiers struggle back across the Polish border. Five variables are plotted – army size, location, direction of army movement, temperature and date – but the impact of this historical disaster on the observer is immediate. It is a graph that produces thought and generates an emotional response.

The art and science of statistical display has a relatively short history.

Thomas Hankins of the University of Washington delivered a lecture a few years ago concerning the history of graphs. He found that graphs first appeared around 1770 and became common only around 1820. William Playfair’s graphs of the British economy were the best known of the early efforts.

European wars stimulated graphic design in the late 1700s. Fortresses had become complex and dirt had to be moved the shortest possible distance, yet still provide ample protection. Hence, the 3D effect on two-dimensional paper made its entrance.

Railroad development in the mid-1800s provided another set of mathematical challenges that were addressed by advances in data graphics. Minard, of Napoleon’s march fame, created two design concepts – graphical tables and thematic maps – in order to visualize the movement of rail traffic through France.

Another graphical advancement called ‘nomography’ was developed in 1899 and had wide application throughout the 1900s – from First World War gunners (who used it to direct anti-aircraft fire) to scientists (who used it to describe the movement of blood within the body, for example).

Each of these graphic breakthroughs was revolutionary in a way. New design concepts were created by imaginative people to deal with new kinds of data sets. The design matched the data and that’s what made the graphic result so perfect.

And therein lies the best advice that can be given to all the TV, magazine, newspaper, radio media sellers and researchers in town.

To quote Tufte: ‘Graphical decoration (non-data ink), which prospers in technical publications as well as in commercial and media graphics, comes cheaper than the hard work required to produce intriguing numbers and secure evidence.’

In an ideal world, the graph for Temptation Island’s audience results will plot weekly audience size using weekly Gravol pill sales. In this way, true thought and knowledge will be the outcome.

Rob Young is a founding partner and senior vice-president, planning and research at Toronto-based Harrison, Young, Pesonen & Newell, one of Canada’s largest media management operations. Send your comments via e-mail to