Graphical perception – the visual encoding of data on graphs – is an important consideration in data exploration and presentation visualization. In their seminal work, “Graphical Perception: Theory, Experimentation and Application to the Development of Graphical Methods,” William Cleveland and Robert McGill lay the foundational theory for setting guidelines on graph construction.
Their graphical perception research enables the data scientist to maximize the likelihood of value transfer for the incurred study cost (AKA Data Monetization). As we encode the information (relevant data) into graphics, the viewer has to decode the data and interpret the results. This asymmetric and error prone knowledge transformation. Fortunately, Cleveland and McCill have identified several best practices that reduce the likelihood of viewer misperception.
Five Graphical Perception Best Practices:
- Use common scales when possible – hard to compare across scales, especially offset
- Use positions comparison on identical scales when possible
- Limit the use of length comparisons – proportions are difficult to interpret
- Limit pie charts – angular and curvature comparisons are hard to interpret
- Do not use 3-D charts or shading
Elementary Perceptual Task