«

»

Graphic Design in Data Science

Share

Graphic design in data scienceThis post is about an interesting debate around the relevance of graphic design skills in the data scientist’s portfolio. Should there be a graphical designer in the data science team? Should graphic design be taught in the data science curriculum?

The human brain is highly adapted to discerning visual patterns – sometimes even when there aren’t any. It responds more effectively to visual representations than to textual data, especially tabular data. With the vast amounts of data now being analysed, visualisations are even more important to make sense of the data and to communicate the message(s) contained therein.

Data Visualisation

Visualisations are appearing as graphs, maps and other graphical elements in the reports and dashboards produced by all the major BI tools. In fact, some tools focus predominantly on data visualisation.

There are two breeds of data visualisations, which differ in their audience and application:

  • Exploratory data visualisations (as named by John Tukey) are intended to facilitate the data analyst’s understanding of the data. These may consist of scatter plot matrices and histograms, where labels and colours are minimal by default. Their goal is to help find and develop a hypothesis about the data, and their audience is typically rather small. (Some products like Tableau and Qlikview focus primarily on exploratory data visualisations.)
  • Communicative visualisations are intended to appeal to a wider audience, where the goal is to visually convince them of a hypothesis. (R can do static visualisations, and tools like Processing and Flare are use to create rich interactive visualisations). However, to me there is a huge difference when a visualisation is used to augment the verbal explanation of a hypothesis to an audience vs. when it must communicate its message “by itself”.

A data scientist must have the ability to create visualisations that convey the stories about the insights discovered in the data. Even with good quality data and rigorous statistical techniques, if the results of an analysis are poorly presented, they will neither convince the right people nor convey the right message. However, the ability to create “self communicating” visual narratives often requires a separate skill — often using separate tools (more about that under infographics below).

Graphic Design in Data Visualisation tools

Data visualisation tools like Tableau have some essential aspects of graphical design beat practices already built-in. The developers of these tools have taken great care to follow the recommendations by data visualisation specialists like Stephen Few. If you use the correct configurations, you can create some really effective data visualisations – where the message that the data conveys clearly stands out, without being cluttered by unnecessary ink and other designed widgets. Tableau has a clever wizard-like function, which recommends the appropriate graph types for the data and analysis under consideration. Of course you can still force it to use the wrong graph type, or configure it to produce really bad visualisations.  So although you don’t need to be a graphical designer per se, you still need to know how to apply the correct settings and configurations to apply good design principles.

Infographics

If we are not using a data visualisation or data exploration tool to interactively communicate about the data, we may need to frame the story differently, either in terms of presentation slides, message conveying images or using a clever infographic.

Not even the experts agree what exactly constitutes an infographic. However, they do agree on a common goal – to present and communicate complex information quickly and clearly. The objective is to improve cognition by utilising graphics to enhance the human visual system’s ability to see patterns and trends. In some cases though, graphical elements such as text and typography are used liberally to blatantly convince the audience what they should be seeing.

Decoration is the one single design element that divides the schools of thought about infographics:

  • Business intelligence expert Stephen Few sums up his disdain for the ornamental aspect of infographics: “When visualizations are used primarily for artistic purposes, they are not what we call data visualizations or infographics, which are terms that have been in use for a long time with particular meanings.”
  • David McCandless, on the other hand, has popularised artistic visualisations and introduced data as a storytelling category to a wider audience. He describes his work: “I love taking all kinds of information – data, numbers, ideas, knowledge – and making them into images. When you visualize information in this way, you can start to see the patterns and connections that matter.”

If visualisation has to serve as a representation that amplifies the cognition of data, one can measure both the efficacy (how easily comprehensible is the data) and the veracity (how truthful is the explication of the data) of a given visualisation. Decoration, for the most part though, introduces visual noise into a design, thereby compromising both measures.

However, the important point for me is the aspect of storytelling in infographics. I see a data visualisation as a factual representation of the trends, insights, etc., contained in the data, but an infographic as a “self communicating” narrative of the same. Alberto Cairo, infographics professor and author of The Functional Art, put it like this: “So you’ve amassed terabytes of data. Now, tell a story.” As another example, journalist Reg Chua described a particular very powerful visualisation: “It’s not simply a dump of data, but one designed – intended – to persuade.” The concept of persuasion regularly comes up as a recurring theme in discussions about the purpose of visualisations, especially infographics.

This is where the services of a graphical designer may be required. I don’t know many data analysts or full blown data scientists who have the artistic skills to put together a persuasive narrative as in an infographic. Well, not one that would be classified as artistic or pleasing to the eye anyway.

Graphic Design

Graphic design is a creative process, used to convey a specific message to a targeted audience. It includes a number of artistic and professional disciplines that focus on visual communication and presentation. Words, symbols and images are created and combined using shape, colour, imagery, typography, visual arts and page layout techniques to form a visual representation of the message.

Graphic design is positioned as an interdisciplinary, problem-solving activity which combines visual sensitivity with skill and knowledge in communications, technology and business. The interesting term for me in that statement is “problem solving”. How much does that overlap with the problem solving skills required for exploratory or hypothetical data analysis? I guess not all that much… However, it also implies that the graphic designer needs an understanding of the client’s products or services, goals, competitors and the target audience. It is therefore not a purely artistic process. For that reason, graphic design is sometimes referred to as Visual Communication or Communication Design – this shows how important the communicative aspect is. In some organisations graphic designers are even called communication designers. A communication designer has similar skills as a graphic designer, but is more concerned with designs that convey specific messages visually, for publicity, broadcast, interactive or environmental communication.

In the context of data science, graphic design is the combined science and art of visual communication. A good communication designer will choose the right tools and formats to convey a message effectively.

The problem is that most graphic designers – no matter how “clever” their designs are – at the crux of it, do normally not work with and do not understand data. So it takes very clear communication to convey the message that we need communicated to them first. The up side of this is that once understood, they usually know very well how to convey that same message in a non-technical fun and interactive fashion to a non-technical audience.

Concluding remarks

With the data volumes and the wide range of tools now at our disposal to communicate with and persuade people, we can much more effectively choose from a wide range of presentation tools and formats. The well-told tale, complete with great colour and anecdotes, backed up by rigorous data analysis, and supported by great multimedia elements, may well continue to be the gold standard we aspire to; but we also need to work on how best to harness our reporting and presenting capabilities so we can create other types of data-driven narratives that touch and persuade people.

It seems that if we mostly do interactive data explorations to find, illustrate or prove hypotheses, we can use a data exploration tool, and apply the best graphical design principles within the limits set by the tool. If we need to convey the message in the form of a story to a wider audience, we can utilise the skilled services of a sharp communication designer to frame the message, for example, as an infographic.

However, data science is not always that clear cut. Some days you are exploring new data, sometimes you are postulating or proving hypotheses and other days you are convincing or persuading different audiences through slides or storytelling. Some days you even reverse the process by first representing the data (through design) and then subsequently refining it (through exploration).

So do you now need to go out and contract the services of that communications designer? Cost-effectively? I will leave the justification of that business case over to you!

Leave a Reply