Scale
From Business Intelligence
About
The scale should accurately depict the data, and visual representations of the data should stand up well to the actual data itself.
- Proportion. The graphical representation should be mathematically representative of the data, and the scale should not change throughout the graphic.
- In a bar chart with $1 and $10 bars, the bar for $10 should be 10 times longer than the bar for $1.
- Graphics involving money over time intervals should always be adjusted for inflation.
- Distortion.The data needs to be accurately represented without any distortion. People interpret colors, shapes, and size before numbers, and additionally make poor precise interpretations based on these factors.
- Adding perspective to a chart distorts data, since it makes it difficult to accurately compare data points that are "far" and "near".
- Circles represent a special problem with proportion, as people have a hard time with accurate assessments of their areas. Differences in values may be more radical than they appear - a circle representing twice the value of a nearby circle is not twice its size.
- Dimensional honesty. 3D elements, although trendy, present several problems. All elements in a graphic should relate to the data itself, and two-dimensional data should have a two-dimensional representation. Interpretation is also an issue by providing several possible places that the edge of the range could be (ie, in a 3D cylinder, is it the top lip or the bottom lip that represents the top range?).
Example
 Bar graph: Descriptive statistics.. Encyclopedia Britannica online. |
This bar graph showing the marital status of 100 people has a number of problems that slow down correct interpretation of the graph. The 3-dimensional bars means that people need to closely follow along the tickmarks to judge the height of the graphic. The angled plane distorts the graphic so that the graph to the right will appear taller than it is compared to the graph on the left. All of these problems add up to a proportion issue - the "widowed" bar does not appear to be a quarter the height of the "married". (This graphic also breaks the design guidelines of carefully using color and avoiding non-data ink.
Relevant Visualizations
Resources
Edward R. Tufte, The Visual Display of Quantitative Information