This entry was written by Bryan Pierce of Perceptual Edge.
Recently, I found a glowing press release from a business intelligence vendor named Blink Logic, which announced that their software is now being used by a large financial services company. After reading the press release, I went to their website where I found the following image:
This is a stacked bar graph that is being used to display how Internet Sales have changed from month to month. Each stacked bar combines sales for Mountain Bikes, Road Bikes, and Touring Bikes. Although this might not be the best graph for the job (notice how difficult it is to compare the heights of the red or blue bars), depending on what their client needs to focus on, this graph type might be sufficient.
What bothered me and prompted me to write about this graph appears in the legend in the upper right corner.
In the graph, the colors of the three bar segments are blue, red, and green. In the legend, they’re blue, red, and yellow. Here are swatches of the two colors side by side.
In looking around their website, I found a few other examples of the same problem.
If you look at the original bars closely you can see that the fill color in each of the bars varies slightly. In conjunction with the shadow on the right side of each bar, it’s probably safe to assume that Blink Logic is trying to simulate lighting effects on the bars and as a side effect, the yellow color, which was selected to represent Touring Bikes, is being rendered as green (that is, yellow supposedly darkened by shadows).
Problems of this sort are fairly common by-products of lighting effects. Because of the gratuitous use of lighting effects, for example, I’ve seen pie charts that looked like they contained more slices than they actually did and line graphs where the colors of two or more of the lines were indistinguishable. Even when lighting effects don’t distort the image, they can still slow us down, because they complicate the display, making our eyes try to find and compare gradients of color instead of uniform patches.
The essential point I’m attempting to make is not just about lighting effects, but about all of the glitzy effects and cool features that vendors incorporate into their products and promote to the world. Before a new effect or feature is added, a vendor needs to weigh its true costs versus its benefits. Is it detrimental to the whole? If you add another variable to a bar graph, you complicate it, making it harder to focus on the other data. Often, this is okay, because the benefit of the additional data outweighs the cost of the additional complexity. But what benefit do visual frills add? When you add a color gradient, shadows, or a 3-D effect to your bar graph, you might make it “prettier,” but is it really worth sacrificing understanding in the process?