Tags
What is Data Visualization?
Stephen Few, a visual business intelligence expert, defines Data Visualization as “the graphical display of abstract information for two purposes: sense-making (also called data analysis) and communication” (Few 2013). This definition conveys two very clear points: information should, if needed, be actionable and consumable. Information communicated clearly with visualizations can help users digest data to make informed decisions.
The history of data visualization dates back to at least the 2nd century AD when people started to arrange data into tables (Few 2013). But it wasn’t until the late 18th century that people began to exploit the potential of graphics for the communication of quantitative data. Scottish economist William Playfair was the first person to use a line moving up and down as it progressed from left to right to show how British debt from colonial wars had grown over time (Figure 1) (Few 2013).
Components of Visualization
Clear, concise visuals, such as annotated maps, graphs, photographs, illustrations, and videos, can often communicate important information more effectively than statistics and numerical tables can communicate important information. While the concept of effective data visualization is quite simple, effective implementation of data visualization is where the complexity lies. For example, data visualization thought-leader Edward Tufte highlights the number of dimensions of data displayed by Google Maps:
In the map in Figure 2, the data has been very subtly sliced into many dimensions. Some dimensions that come to mind include:
· Relative size of streets, rivers, buildings, etc.
· Shape of streets, rivers, buildings, etc.
· Names of streets, rivers, buildings, etc.
· Direction of traffic
· Amount of traffic at the current moment
· Type of road (the signs depict the roadway type)
· Ferry path
· Scale of map (the + / – button)
· Type of location (i.e. the forest icons on parks).
All of this has been achieved with a very minimal design that does not at any point overwhelm the user. It would be very easy to look at this map, find one’s destination, and arrive upon the best set of directions (although Google fortunately provides this for us).
While the above Google Maps example performs its role well (i.e. finding directions), sometimes another visual can better convey mappable data. Figure 3 below uses a detailed speed-time-location display based on archived Intelligent Transportation Systems (ITS) detector data from a private sector data provider (FHWA).
The graph above shows average directional speed by location along the expressway corridor as well as by time-of-day throughout a selected sample day. This can be more powerful than a simple color-coded schematic map displayed in Figure 4.
Such images demonstrate the need to think critically about the data to maximize sense-making and communication. Figure 3 allows users to see where bottlenecks occur on the highway over time, whereas the Figure 4 shows only averages. In addition, it communicates more detailed context to traffic, including the highway exits and mile markers, making it a sensible and communicative visualization.
Why is Data Visualization Important?
As data has grown more complex in shape and meaning over time, digesting information has become more of a challenge. Data visualization has therefore become an important tool to simplify and communicate information to the end-user. Good data visualizations will guide the user to key takeaways and actionable insights and will thoughtfully convey as much important information as possible without obfuscating the underlying message.
A Simple Thought Experiment
To highlight the importance of data visualization, imagine that you have been handed the following smartphone sales dataset:
By scanning the dataset, have you received any benefit? It is possible that you can recognize a pattern by observing the complex set of rows and attributes, such as the introduction of the Windows Phone and phase-out of Windows Mobile. More likely, however, you must transform it in some way before you can extract information. You might build an analytical model that describes the data and/or write down some bullet points showing trends.
In the case of the model, you might be able to accurately describe the data but it is very difficult to interpret. In the case of the bullet points, you may have described a portion of the data, but you have very likely failed to describe the big picture. It should be clear what the end goal to this thought exercise is: a visual can succinctly describe and contextualize the data so that it can be interpreted. The following visual (Figure 6) allows users to interpret and interact with the dataset. (Interactive)
If you have created a good visual to present information, you most likely saved the end-user the time of parsing through the dataset or model results. Effectively, you provided the end-user with a clear presentation of the information so s/he can quickly make an informed decision. While you may have helped the user arrive at a decision quickly, it is imperative to recognize that failure to think carefully about the visual can become problematic.
Problematic Visualizations
Badly designed data visualizations run the risks of poorly expressing, or worse, miscommunicating the intended message. For example, a study from the University of Munich, Germany showed that irrelevant depth cues in bar graphs led to slower decision-making (Fisher, 2000).
Another famous example of graphs utilizing style over substance is the pie chart — although effective for displaying simple proportions, statisticians have pointed out that a pie chart is generally inferior to a bar chart (Cleveland and McGill, 1984). While pie charts have several measures for comparison – length of circumference edges, angles, and area – humans have more trouble distinguishing these measures than distinguishing length as provided by bar charts. As seen in Figure 7 below, the bar chart allows the reader to more easily assess rank and relative size than using the pie chart.
Furthermore, data visualizations can be used to purposely mislead audiences in order to push an agenda (think politics, viral marketing, etc.). This reinforces the belief that usage and understanding of data visualization is very important, as one needs to be able to separate fact from fiction when assessing information.
Conclusion
In summation, data visualization is a very powerful tool that can result in the fast and effective translation of data to information. It can be used to express differences in groupings of data along multiple dimensions, but it can also be misused if one does not properly ensure the quality of their work. Furthermore, as data becomes more complex, a good analyst must keep up in the field of data visualization in order to convert the newer, more sophisticated datasets into actionable information. In the interest of providing examples of new data visualizations, below are a few resources:
· http://www.visual-literacy.org/periodic_table/periodic_table.html
Works Cited
“Chapter 4 – Using Visualization as a Communication and Analysis Tool.” Federal Highway Administration. n. page. Print. <http://www.fhwa.dot.gov/planning/congestion_management_process/cmp_guidebook/chap04.cfm>.
Cleveland, William S., and Robert Mcgill. “Graphical Perception: Theory, Experimentation, and Application to the Development of Graphical Methods.” Journal of the American Statistical Association 79.387 (1984): 531-54. Taylor & Francis Online. Informa UK Limited, 12 Mar. 2012. Web. 27 Mar. 2014.
Fischer, Martin H. “Do Irrelevant Depth Cues Affect the Comprehension of Bar Graphs?” Applied Cognitive Psychology 14.2 (2000): 151-62. Wiley Online Library. John Wiley & Sons, Inc., 1 Mar. 2000. Web. 27 Mar. 2014.
Few, Stephen (2013): Data Visualization for Human Perception. In: Soegaard, Mads and Dam, Rikke Friis (eds.). “The Encyclopedia of Human-Computer Interaction, 2nd Ed.”. Aarhus, Denmark: The Interaction Design Foundation. Available online at http://www.interaction-design.org/encyclopedia/data_visualization_for_human_perception.html