Storytelling in Data Science

Comics are a combined medium of written text and pictorial images used to convey stories, typically in a social context. Initially comics were mostly humorous, but since then have communicated many other sentiments, and even instructive information. Contextual storytelling is analogously important in data science.

There are two breeds of data visualisation. The first is exploratory, intended to facilitate an understanding of the data either by discovering trends or insights, or by developing a hypothesis about the data. The audience is typically a very small group. The second kind of data visualisation is intended to communicate to a wider audience, where the goal is to visually advocate for a hypothesis, or to communicate a large volume of information very concisely. This requires a different skill, and is typically done using different tools.

In this post we focus on the second type of visualisation, because the ability to visualise and communicate data is critical in data science. Even if you do excellent data exploration and apply rigorous statistical techniques, if the results of analyses, hypotheses and other data-driven investigations are poorly presented, they will not convey any convincing messages.

Rationale

The outputs of advanced analytics, which are sometimes complex to understand, action and track, are a case in point. By applying visualisation you can illustrate the ramifications of the derived insights in terms of their business impacts.  Once better understood, the business has a much more open attitude to embark on business initiatives utilizing, and then re-affecting, these outputs.

This is where storytelling is essential – you have to be able to explain what you have found, in terms that business people can relate to. You have to broadcast the lessons learned so that people in the broader business understand their implications.

Visualisation

A key component of storytelling is visual presentation. Visualisation is used more and more to make the data more digestible for non-technical audiences. In comics, the artwork can be more impressive than the story it tries to depict. In data visualisation, however, the art must not interfere with the information that it tries to communicate. Stephen Few posted on his blog: “a data visualisation should only be beautiful when beauty can promote understanding in some way without undermining it in another.” And further: “What qualifies as beautiful for some is not beautiful to others, beyond the basic aesthetics… of good color choices, legible fonts, proper placement and spacing”.

Infographics

Just as we need to avoid “death by PowerPoint” in business presentations, we need to take great care to avoid “death by art” in data visualizations. Edward Tufte, an American statistician and professor of political science, statistics, and computer science, well known for his writings on information design, coined several useful terms such as “chartjunk” to refer to useless, non-informative, or information-obscuring elements, and measures such as “ink-to-data ratio” to argue against using excessive decoration in visual information displays.

In the information dissemination world, the real analogy to a comic is an infographic. An infographic takes a large amount of textual and numerical information and then condenses it into a combination of images and keywords, allowing viewers to quickly grasp the essential insights that the data contains. Interesting examples of infographics can be found at dailyinfographic.com and coolinfographics.com.

Infographics became popular in the early 1990s when news organizations began emphasizing visual communication. People are now spending even less time reading written words. Infographics have now become a widely accepted, even fashionable, way to convey a lot of information very concisely. Infographics can be used to disseminate quite a substantial “story” to a large audience, without requiring them to sit through a presentation. However because infographics typically take up more real estate, and attempt to convey much more information than a single graph, combined with the temptation to make them look sexy and interesting, the artwork and decorations easily get in the way of the information. Chart junk and high ink-to-data ratios easily abound with popular-looking infographics.

Concluding remarks

Whether you write a book, act in a play, paint a picture or draw a cartoon, the story must be told. Whether you write a report, give a presentation, draw a graph or assemble an infographic, the information, findings and recommendations must be communicated to the business. Just make sure sound aesthetics, good design, low ink-to-data ratios, and above all, plain common sense prevails.