Data Visualization Principles - Technology Information - Column

ENT, Oct 6, 1999 by Robert Craig

Effective data analysis and understanding can be helped or hindered through the use of data visualization technology. A good graphical presentation of data can make it easy to understand available information. Conversely, a poor graphical presentation can make it impossible to comprehend the data.

In this column, I'll discuss some data visualization principles based on the work of Edward Tufte of Yale University. Given the increasing use of graphics, especially on the Web, this information can be helpful when designing business graphics, presentations or a user interfaces.

Tufte has three excellent books on data visualization. The Visual Display of Quantitative Information is about number images and examines statistical charts, graphs and tables. Envisioning Information analyzes maps of data. Visual Explanations focuses on depicting cause and effect.

One fundamental principle is that a display should be focused on the data. While this seems like an obvious concept, it is one that is often ignored. Designers often clutter charts with icons or other elements that they think are cool. But these have no value in terms of clarifying the viewer's understanding of the data. Also, the order of data can have a significant impact on a viewer's ability to understand it.

When you are designing a graph, chart or user interface, resist the temptation to add extraneous graphical elements that don't clarify or expose the deeper meaning of the data. Also, ensure that the data is presented in a meaningful order.

One example that Tufte relates is how the scientists at Morton Thiokol prepared graphs to illustrate the relationship between launch pad ambient temperature and O-ring erosion after the Challenger disaster. Their effort was futile for two reasons. One is that the chart had little rocket icons on it, cluttering the display of data. Second, the data was ordered in terms of launch sequence, rather than by temperature.

Clear, precise thinking about a problem will lead to clear, precise understanding about how to present the data. In this case, the rocket scientists should have created a simple scattergram with two axis: one for the degree of O-ring damage and the other for the launch pad ambient temperature. When ordered and displayed this way, it is obvious that there was a direct relationship between O-ring damage and a cold launch temperature.

Another principle to consider is that a good graph is multidimensional. An example chart cited by Tufte shows the fate of Napoleon's army as it entered and exited Russia in the winter of 1812. The chart has a band that illustrates the size of the army. When Napoleon entered Russia with 400,000 men, the band is wide; the width was then reduced to a thin line illustrating the 10,000 men who made it back to Poland. The graph also illustrates the army's direction, location over time and the daily temperature. A good graph will visibly demonstrate the relationship between cause and effect. In this instance, the relationship between the number of surviving soldiers in the army and the temperature -- which fell to -30[degrees] C -- is striking.

Use elements like color, direction and text to illustrate the richness and complexity of data. Once you learn to do this, you'll never be content with creating a two-dimensional pie or bar chart again. --Robed Craig is vice president of marketing at WebXi Inc. (Burlington, Mass.), and a former director at the Hurwitz Group Inc.