Welcome to Part 2 of our series on data visualization! Follow along as we journey through the world of turning data into easy to understand images. (Read Part 1 of the Series)
Statistics have a reputation for being misleading – “There are three types of lies: Lies, damned lies, and statistics.” It is no different with data visualization. Poorly designed charts and graphs can confuse or mislead even informed audiences, and are easy to create due to inattention, internal politics, or simple lack of know-how from the creator.
To avoid building poor visualizations, the first concept in RG+A’s data visualization framework is to “Be true to the data.” This idea can be broken into two more specific statements: “Tell the truth,” and “Don’t mislead.”
Telling the truth is straightforward. In our case, we are hired to do quality research and present the findings accurately; as vendors, RG+A’s reputation, brand, and business are based on the trust that we will do so. If we put data into a graphic that reflects the unaltered reality of the information gathered, then we can say we are telling the truth.
However, it is entirely possible to show accurate data in a misleading way – even unintentionally. To avoid this, RG+A has created guidelines that prevent the creation of true-but-misleading graphics:
- When displaying a comparison, use the same scale
- Don’t show count for Product A and percent for Product B, choose one
- Don’t use a 40%-100% scale for Product A and a 0% – 60% scale for Product B,
- When displaying a comparison, use the same base (or clearly specify the differences and rationale)
- Don’t show Product A share amongst females but show Product B share among females and males
- Start charts at the origin if possible
- Help the eye compare data – use appropriate visuals for the data
- Ensure visuals accurately represent the comparison shown
The first three guidelines are straightforward. As the image above illustrates, any comparison must be between analogous groups. That is, not only must we compare relevant sample populations and equivalent data, but we must display the data in a comparable way. We do not show counts in one chart and percentages in another. We ensure that the scales and units are equivalent. Finally, we start the chart at the origin and always use a consistent upper limit.
To help understand the nuances of the final two guidelines, we turn to a piece entitled, Graphical perception: Theory, experiments, and application to the development of graphical methods. The key findings inform the application of rules 4 and 5:
- The human eye is great at comparing distances and heights, and understanding linear graphics
- It’s NOT so great at judging or comparing angles, curves, area, or “3D” graphics
This work implies that clustered bar charts and dot plots should be the standard for displaying most data, with pie charts, donut charts, and stacked bar charts reserved for limited and specific occasions (i.e., when you have fewer than 5 categories).
Representing data accurately is the most important part of any visualization. Once an accurate graphic that is both true to the data and not misleading, we think about revising it for ease of interpretation.
What tips do you have for making data visualization impactful? What challenges have you encountered?
Visit this post on LinkedIn and become part of the dialogue.
See you next week for Part 3 of RG+A’s data visualization series: Supporting ease of interpretation