Our brains interpret visual signals in specific, predictable ways, and modern user interfaces and consumer products are based on extensive study in this area. The way we interact with visual signals was first formally studied and discussed by William Cleveland and Robert McGill (at the time, with AT&T Laboratories) in their seminal paper named Graphical Perception: Theory, Experimentation, and Application to the Development of Graphical Methods, as published in the Journal of the American Statistical Association in September 1984. This paper is readily available via a quick Internet search, and it is worth a quick read.
In addition to Cleveland and McGill, who were statistical scientists (and later professors) focused on researching how we process visual cues, Stephen Few has written extensively on this topic in his books and on his blog, which can be viewed by visiting http://www.perceptualedge.com/.
The authorities on this matter have concluded that the human brain correctly perceives numerical values in a way that is different from how it perceives various visual primitives.
In the descending order of accuracy, visual primitives are as follows:
- Position: This shows how marks relate to an axis, like in a scatter plot.
- The bar or line length: This is why line graphs and bar graphs remain the first and the most popular forms of data visualization.
- Angle: This should be used sparingly, as the human brain actually is not good at differentiating angles within a circle. Pie charts are effective if you wish to communicate measures as long as about five dimensions are included and the slides are labeled with their respective percentages of the whole.
- Area: It's actually fairly difficult for us to differentiate numerical values from each other based on the size of a shape. Therefore, you should use bubble graphs and tree maps sparingly and label them properly when you do use them.
- Saturation and hue: These are the worst primitives that you can use to encode numerical values because not only are a significant number of readers color-blind, but also they are used inconsistently and designed for perception rather than judging.
- Color: This is best used for categorization into groups rather than to measure quantity. However, you can use it effectively in heat maps and highlight tables, as long as you are using it properly and labeling the values.
Your mileage may vary, but consider the hierarchy of visual clue perception carefully. Some conservative data authors will try to stick to scatter plots, line charts, and bar charts (and their derivative chart types) almost exclusively and avoid most other charts such as pie charts, bubble charts, and heat maps.
Leveraging your understanding of how our brains interpret visual clues will help you select the proper chart type. Fortunately, there is a built-in assistant in Tableau Public called Show Me that will help you choose a chart type based on the data elements (measures and dimensions) that you select.