How we Really View Data Visualizations

07 Aug 2017 ByBrian Indatavis

Add Comment

There is a lot of research and writing on how to design effective data visualizations. Some researchers measure the speed and accuracy with which people read values from a visualization and exhort us to use bar charts. Others will point to psychological principles such as gestalt laws or preattentive processing ¹ as ways of tricking our brains into seeing the right things. Others such as Tufte exhort us to avoid chart junk and focus on a high data-ink ratio in order to avoid misleading people and best convey information without distraction or confusion.

All of this is good and useful advice, but I feel it ignores some important realities of how people really encounter, view and relate to data visualizations. With that in mind, here are some thoughts on the process of viewing data visualizations and how to incorporate these ideas into how we design them.

Steps

Here’s my intuitive breakdown of how people view a data visualization:

Attention
Curiosity
Gist
Interpretation
Questions & Insights
Memory and Sharing

Attention: Before any of the other steps can happen, before people can read values or gain insights, they first need to notice that your visualization exists.

A steadily growing part of our work is to create tiny animated versions of our graphics for the sole purpose of making you click — Gregor Aisch on Twitter

In this regard, data visualizations are already ahead, as eyetracking studies show people tend to pay a lot of attention to information carrying images². Other ways to attract attention include:

Colours: colours that contrast strongly with their background
Animation: our eyes are automatically drawn to movement
Images: images, scenes, pictograms and other recognizable shapes engage our meaning-seeking minds

Notice that these techniques tend to fly in the face of the usual advice to be restrained and conservative given above. Also notice that going overboard on any of these techniques will lead to kitschy, garish work. Still, don’t be afraid to inject some beauty into your work.

Curiosity: Once your audience has noticed the visualization, their next question will be “Do I care?” We need to help them answer this question in the affirmative as quickly as possible.

Large headlines are a primary tool to convey important messages that draw people in. If your visuals are meant to convey a key message, then clearly stating it can short circuit this whole step. Even if your visualization doesn’t have a predefined message, a bold title describing what might be found with further investigation is still worthwhile. While not examining curiosity, there is research showing that people pay strong attention to titles, especially when they are at the top of the visualization³.

Interactive visualizations can use establishing shots or splash screens to draw people into the story being told by the data. If there are multiple steps or insights to convey, a checklist or other progress tracker can entice people to view the next parts⁴. Just making your visualization aesthetically beautiful will lead people to spend more time trying to understand it⁵, indicating an increased curiosity about its information.

Making the information personal is another great way to help people care about your visualization. Humans tend to be self-centred, so if we can show how they personally relate to the data we’re presenting they are much more likely to want to know more. Interactive visualizations can prompt users for information and create a custom view of the data for each individual. For example, maps can start focused on the user’s city or country, social visualizations can start with the user’s demographics, etc. Static visualizations can still make it easy for people to find themselves in the graphic and highlight relationships between members so they can orient themselves in context.

Gist: Gist is usually where most data visualization advice begins. Here we want to select appropriate graphics and layouts that enable people to quickly form an overall understanding of what the data is saying. Line charts imply sequences; bars imply categories; pies imply proportions, and on and on. Colour can guide viewers to specific data points or to see relationships. All the usual design techniques are at our disposal to convey our data to our viewers.

Interpretation: By now our viewers have already formed a high level conception of what our visualization is about. At this stage they may look closer to verify their interpretation is correct and understand the data at a deeper level. Here axes, labels, annotations and other layers of information come into play. If the visualization is interactive, we may show (or they may explore) multiple levels of detail. The key question being asked and answered is does the data and visualization support their gist of it?

Questions & Insights: If your viewers get this far you’re doing really well. You’ve attracted their attention, piqued their curiosity, told them something interesting and they still want to know more. This is where interactive visualizations excel, allowing users to dive in and explore the data set on their own. A common narrative pattern is to lead people through the data set, highlighting key insights along the way and then allow further free-form exploration at the end (Segal and Heer calls this the martini glass structure⁴). For static visualizations, this may mean looking at additional annotations and content around the visualization or perhaps the surrounding story. Visual structures such as comic books panels, flow charts and other layout techniques can help lead people through layers of analysis in a linear fashion.

Memory & Sharing: A final consideration in our designs is the longer term effects of the work. Will viewers remember what they’ve seen? Will the data affect their opinions? Will they tell anyone else about it? Borking et al has done some interesting work looking at what makes visualizations memorable³. They found that titles that convey the message of the visualization, pictograms or other identifiable imagery and redundancy of encoded information all help people describe the visualization and data later.

Takeaways

When looking at the broader picture of how we view visualizations, there are few key points we can learn from these steps.

The Big Picture: Speed and accuracy, the main concerns of most visualization advice are not the be all end all of effective data visualization design. The big picture matters, and this introduces a host of additional design considerations.

Level of Interest: The viewer’s level of interest needs to be maintained at all times. At first we need to be noticed, then interesting, then informative. Different techniques are applicable depending where we are in this multi-stage process.

Emotions Count: We may think we’re creating objective depictions of data, but that’s not how people view them. Instead they approach them as an emotional medium, as much as an informative one. Beauty, delight, fear and other emotions are primary motivators in whether and how people engage with our work.

Much of the information presented here are general design principles that apply equally to other forms of information and storytelling. The specific techniques used and how we apply them to technical data sets is what warrants extra attention. With care and attention we can help people understand large and complex situations and hopefully provide real value for them.