The need to extract meaning from dynamic datasets in a fast and flexible way is becoming increasingly urgent. By making use of the potential of internet browsers, new data visualisation tools and software packages allow us to do so. However, because of the field’s rapid evolution, we risk overlooking some of the basic aspects of data visualisation: your goal(s), striking a balance, finding inspiration, effective annotation, and interaction and animation.
First, the bad news. As data visualisation expert Andy Kirk states, “There is never a single path towards a ‘best’ solution. The inherent creativity and individualism of design work ensures that.” That’s why the following insights serve as ground rules only, and focus mainly on the most important aspects of data visualisation.
Whether it is a bar chart, a pie chart, or a complex interactive visualisation, it’s important to remember why you want to visualise your data in the first place. Frequently revisiting, restructuring and refining your goals will help you decide if visualisation is a good idea, and whether the data you have is sufficient to do so.
Additional contextual data may help you accomplish the task more quickly or more thoroughly — that’s why the buzzword big data was invented. Big data will also help you define the most important variables and values, as well as enable you to know more about the characteristics of your data and your users before making decisions on what is important to display or what colour palette to choose.
In any attempt to visually represent a dataset, you will naturally be making many choices. Essentially, you’ll need to balance effectiveness and expressiveness by conveying information about your data using the appropriate visual marks — e.g. dots, lines, shapes — and assigning visual attributes that efficiently add the intended meaning to your visual components.
Making conscious and goal-driven decisions with regards to that balance will help you create a (not the) best solution. For example, you should not use size variation to express different categorical values, and quantitative differences should not be visualised with hues. The writings of computer science professor Tamara Munzner are great references when it comes to aspects of data visualisation techniques that touch on this balance.
A quick internet search for rankings of perceived accuracy from Cleveland & McGill will help you decide which visual variables and, by extension, which chart types are ideal to present your principal values. Want to dive deeper? Check out the work of Colin Ware, Director of the Data Visualization Lab, or look online for Garner’s whitepaper on integral and separable visual dimensions.
In trying to strike the right balance between these characteristics, you will find yourself at the complex intersection of statistics, graphic design and cognitive psychology. So, staying up-to-date with what happens within these fields and expanding your awareness of relevant concepts will certainly help you to make faster and more effective decisions. Viewing submissions for contests like the Malofiej awards or following blogs like Information is beautiful are also great ways to keep up with the times.
Another, often underrated, aspect of data visualisation is known as annotation. Annotation isn’t just about making efficient use of typography; it also calls for the smart use of titles, introductions, labels, captions, grid lines, axes and units. However, it is just as important to omit unneeded grid lines, labels, etc. — don’t simply stick with the software defaults — an important ground rule that applies to any aspect of data visualisation.
Edward R. Tufte has some useful things to say on this topic, and Noah Iliinsky, UX Architect at Amazon Web Services, and Julie Steele, Director Communications at Silicon Valley Data Science, recommend: “Minimise noise, maximise signal”. Annotation should thus be well-balanced, show what is necessary for correct interpretation, exclude anything that’s unnecessary and blend harmoniously with visual marks and channels. A best practice here is to visually send the annotation layer to the back.
This is a great segue to interaction and animation. Together with bottom-up oriented visualisation tools, interaction and animation offer lots of opportunities to make interpretation more efficient. The ability to manipulate the visualisation in response to user-initiated changes and the capacity to select, zoom, scroll, or reduce a dataset on the fly allows you to go far beyond what is easily constructible within a spreadsheet or standard presentation application. These features will help you present your story in an intuitive way so that it sticks with your audience. And, when approached consciously and intelligently, it should also help you appreciate the beauty and power of the endless possibilities available to craft excellent static visualisations of your data.
I hope that my brief take on data visualisation will help boost your confidence in applying some of these techniques to your current project and/or make it easier to choose between different implementations. In the end, it’s all about picking the option that suits your data without getting overwhelmed by choices and solutions you don’t need. Want to know more about how we can help you with your data visualisation goals?
Don’t hesitate to get in touch!