Figure 1a and 1b
Data Visualization Techniques
Aim high by sketching first
by Kaiser Fung
When people hear that I analyze Vimeo’s data, they always remark on how different the Vimeo experience is compared to YouTube, the king of online video. Vimeo spells beauty and artistry, while YouTube feeds on viral clips (think Harlem shake and ice buckets), cat videos and gags. Most users who upload videos to Vimeo are professionals who let the polish and purpose of their art shine through. By contrast, any smartphone user has access to an advanced videocam these days.
It is time for professionalism to come to data visualization. Excel has made it simple for anyone to make data graphics, just like the smartphone has turned everyone into filmmakers. As a quick visit to YouTube demonstrates, not many amateur clips hold our attention; the ones that do have received long hours of editing and treatment. Similarly, great data graphics require a professional attitude – and dedication.
One key aspect of professionalism is sketching. Behind any great chart is a trail of abandoned sketches. As a visual medium, charts cannot be designed in the head. Many an idea sounds great in one’s mind, but cracks appear when one inspects a sketch. Participants in my dataviz workshops quickly see the limit to written critique when the designer explains why a proposed edit would not pan out.
For example, several students did not like the unconventional direction of the axes in Figure 1a, and suggested a switch to standard coordinates with zero at the center; when this proposal was sketched, they discovered that reversing the axes would cause key data to clump together, making the revised chart worse, as shown in Figure 1b.
Sketching allows ideas to be worked out visually. It exposes bad concepts and validates good ones. It provides fodder for creativity. It sometimes highlights problems with the data and suggests ways to overcome them.
A few years back, I posted on Junk Charts a critique of Figure 2a, together with an alternative visualization, shown in Figure 2b. (For this article, I fixed a small issue with the ordering of the categories. The blog post was titled “Getting the Basics Right Is Half the Battle.”) The original chart was published in The Wall Street Journal, and it used data from a new report on the download speeds of various broadband service providers in the US.
The problems with the original chart are easily discerned, but the alternative design congealed only after I made a bunch of sketches using JMP software’s Graph Builder. (If you see incomplete and even inappropriate labels, gridlines and the like, it’s only because those are of lesser concern during sketching.)
For a single series of continuous data with text labels, the starting point is a column chart with the numbers running up. When the text labels are long, a bar chart affords room for the words to be displayed horizontally. Ordering the service providers by the ratio of excess performance, as in the original chart, makes sense at first sight, but this design muddles the distinction between types of technologies. Assigning colors helps while resorting the data; first by technology and then by provider is superior.
A critical moment arrives with the recognition that neither bars nor columns accentuate the message (Figure 3). A reference line is added at the 100-percent level to separate those service providers that outperformed from those that over-promised in delivering speeds. For this data, the zeroth level doesn’t matter, and yet a bar or column chart requires starting the axis at zero. I experiment with a dot plot (Figure 4), which has the extra advantage of spacing out the data. After looking at grouped and ungrouped versions, I select the former. To add context, I indicate the technology-level average ratios using vertical line segments.
The final graphic effectively brings out the key messages of the data. Broadband providers using DSL technology generally over-promised and underdelivered in transmission speeds. Cable providers on average exceeded their promised speeds, although individual performance varied substantially.
In thinking about your data graphics, the metaphor of Vimeo versus YouTube is telling. Professional filmmakers and artists have formed strong communities on Vimeo, and their presence is written all over the website. As a designer of data visualization, you should aim high. As a rule, we are rewarded handsomely for time spent sketching and refining ideas. Modern software provides flexible, interactive tools that make sketching painless. It took mere minutes to generate the various sketches shown here. I highly recommend using sketching tools to make your graphics look professional.
Want more on this topic?
- Making Better Sense of Data With Good VisualizationsData visualization experts Alberto Cairo and Kaiser Fung give tips for using graphs that do your data justice.
- Kaiser Fung: Discover What's Hiding In Your Data With Exploratory Data AnalysisData visualization expert Kaiser Fung presents a series of cross-industry case studies that will showcase the ways in which exploratory data analysis enables you to get the most out of your data.