Principles of Data Visual Cues
How to improve information signalling visually.
Background
Data Visualisation has become a keen interest area of mine since I started co-teaching an adult class on Tableau at the Singapore Management University in 2018. Recently, my co-instructor and I have been working on an expanded curriculum, turning the original single module course into 5 connected modules. While our first and original module focused on the principles of visual vocabulary, our 2nd module focuses on visual cues.
As a data analyst / data scientist, data visualisation was never a focus area nor a core competency. Tableau is one of the most popular commercial data visualisation tools globally, and when the university first approached me to develop an adult upskilling class for it, I wanted to approach it differently. The market was already filled with technology vendors teaching the technical sophistry and features of the tool (much like an Excel masterclass), but they lacked business utility intent. What am I solving for with the tool? And so I reached out to a fellow data scientist (KOO Ping Shung), and together we did some research, and created what was to become one of the top adult courses for the university. We focused on what makes a particular data visualisation the right visual representation, and how it can be easily achieved with a modern tool like Tableau that would otherwise be difficult or impossible in Excel. We are currently on version 9 of the curriculum in 2024/2025. (See article #2 — The Problem with Data Visualisation.)
So what about visual cues, and how does it relate to data visualisation? And so I dedicate my 73rd article to developing some principles around this topic.
(I write a weekly series of articles where I call out bad thinking and bad practices in data analytics / data science which you can find here.)
Visual Cues
What exactly are visual cues? Wikipedia defines it as follows: “Visual cues are sensory cues received by the eye in the form of light and processed by the visual system during visual perception.” The Interaction Design Foundation has the following sharper definition: “Visual cues are elements such as arrows, icons or typography that guide users towards certain actions or content in an interface. Designers use them to draw attention, suggest a course of action, or give users feedback.”
In the design domain, there is the famous Gestalt Principle of visualisation that informs the work on visual cues (see digram below). These are particularly useful when it comes to the design of Infographics (e.g. signage).
There are attempts to directly translate those principles into the domain of Data Visualisation (see diagram below). However, I feel that the translation is too abstract and doesn’t carry well. For starters, the domain of data visualisation does not have the same degree of freedom in design (charts are extremely structured) as one would have with infographics. For me, the Gestalt principles are more akin to techniques rather than core principles.
Principles of Visual Cues
Based on my research, as well as my increasing experiences in data visualisation and visual analytics, I’ve landed on 3 key principles when it comes to data visual cues. The principles are to be applied in sequence.
Principle #1: Clarify
Visual cues are sign-posts and guides to seeing the right information signals from a data visualisation (i.e. chart). The principle of clarify is to make sure that the right visual vocabulary is applied. If you had used a ranking chart instead of a deviation chart, then the information signalling / message is entirely wrong. No amount of additional visual cues will solve it. For more on visual vocabulary, you can refer to my article #2 — The Problem with Data Visualisation.
Beyond using the right visual vocabulary, you also need to lay the visual objects out in such a way that is aligned to how information is processed. For example, if your chart requires a legend, then it should be placed just beneath the title of the chart so that you have all the necessary “decryption” at hand to interpret the chart in a single first pass. Too often, we see charts with the legend placed at the bottom (typically bottom right), which then leads to the audience reading the main elements of the chart, not fully understanding, and then referring to the legend at the bottom, and then going back to the main chart elements again. An unnecessary circuitous route. The key here is to leverage the natural path of the eye and place the chart elements along that path in such as sequence as to minimise eye-tracking (i.e. where the eyes jump back and forth) and cognitive processing.
Principle #2: Minimise
The next sequential principle is to de-noise or de-clutter the chart. Strip away all unnecessary elements that do not contain any relevant or supportive information signals. For example, remove unnecessary axis, axis markings, grid lines, colour, boundaries. Keep it clean. This allows the relevant information signal to shine through quickly.
Principle #3: Amplify
The final sequential principle is to reinforce the information signals, to make it stand out further, to make it unambiguous. For example, highlighting with colours or increasing the size of the visual element. Adding labels and call-out boxes are also part of amplification, but I would use these judiciously, as they can quickly lead to visual clutter.
Here’s a good example of how these 3 principles are applied.
Conclusion
Principles are important. I personally feel that if I can’t whittle down a topic into its essential components or wrap it with a small set of defining and operating principles, then I didn’t fully and deeply understand the topic. As I explore the world of data visualisation, I have come to realise that we are only beginning to scratch the surface of this multi-disciplinary domain. As data visualisation tools improve, we have the opportunity to re-define how we might better interact with data visually.
