Data Visualization is Here to Stay, and Here's Why

There are skeptics that question the value of data visualization, and more specifically, the return on investment in developing skills to build effective visualizations. I find this interesting because research has shown that the way data (or a problem) is represented has implications for how people understand data, solve problems, and make decisions (see Zhang, 1997 for a prime example; google “external representations” on scholar for many more).

Data visualization is especially important in work domains where people 1) are dealing with exponentially growing data and increasing complexity of data-processing algorithms, 2) are under time constraints to make sense of data, and 3) are facing deadlines to make decisions from insights provided by data (i.e., a majority of work domains today). In other words, data analysts from various fields (e.g., medical, aviation, national security, etc.) are expected to drink from a firehose of data and efficiently spit out meaningful information that can be used by high-level stakeholders to make time-sensitive decisions.

In contrast, for applications like communicating results in publications, I acknowledge that data visualization may not seem as important. However, if you visually represent your data effectively, it will likely help people consume your data quicker, understand it better, and, as an added benefit, engage with it longer. Studies have found that people engage more with visuals that are useable and aesthetically pleasing (Cawthone & Moere, 2007). Not to mention, visualizing data may reveal connections and/or patterns for your own analysis that may not be immediately evident.

I agree that many folks are trying to find new ways to visualize data, presumably in the pursuit of art rather than usability or application. Many of these new visualizations are superfluous, created with little to no thought or foresight. But sophisticated representations can be created using combinations of basic representations such as bar charts, line graphs, scatterplots, etc. Such integrated representations can reduce the burden on consumers that must navigate page after page of separate graphs to find (with little assistance) important connections or patterns. Furthermore, certain representations on their own have inherent weaknesses. Integrating basic representations can account for these weaknesses.

Building effective integrated representations is not a trivial task. It requires finesse and knowledge of how humans visually perceive different tokens in a representation medium and strengths and weaknesses of different basic representations relative to human perception. Acquiring this knowledge and learning these skills does have some overhead, but as data continues to become increasingly abundant and complex, data visualization will only become a more powerful analysis and communication tool.