3 Principles to Keep Data Visualization Simple for Users
07/27/2016 by Jaime D’Agord Data Communication
Data visualization allows the mind to interpret data to gain insights. To be the most efficient in bringing out takeaways, time is wasted if end users have to figure out what they are looking for or cannot tell what they are seeing. Use the following principals when designing your reports and dashboards.
Before looking into ways to improve reports, it would be good to first look into what shouldn’t be done to graphs and charts that could hinder visualization. In Edward Tufte’s book, The Visual Display of Quantitative Information, he references what he calls chart junk on multiple occasions. What he means by chart junk is anything added to the space of the graph that adds nothing to the data. When looking at a graph we want to see the data and nothing else. This includes grid lines, borders, axis marks, legends and labels that can already be understood in the graph, and any other items that do not add any context to the data being seen. These features are only going to slow down the first steps of the perceptual process.
In the above figure, you can see a side-by-side example of chart junk being used (left) and taken away (right). Notice how the data labels and gridlines make you focus more on details that aren’t relevant to what the graph is trying to show. The exact numbers each month don’t mean as much as noticing the changes for each category every month. Also, the right chart has removed axis lines and labels which allows more space for the actual chart itself. The labels for each axis do not need to be on the report if they can be easily understood from the title.
In addition to chart junk, Tufte also talks about variance over detail. With this point, he means that the details around the data are much less important compared to being able to recognize the differences in the data. Essentially, this comes down to knowing what is wanted out of a certain graph. If the focus is only a select few values of a certain element, then only show those ones instead of everything and build the chart around being able to display an outlier if it comes along. After all that is the end goal, to show the major differences in the important data to then be able to do further investigation.
Once all of the unnecessary features are removed from the charts and only what is needed is being shown, then changes can be made to the data elements to help speed up the perceptual process. One visual aspect that is important surfaces, but the surface is a 3D component that we see in our real-world environment. When dealing with a 2D-space, the only way to create a surface on objects is to change the shading and texture of the objects. When applying these qualities, the objects on the screen get a popped outlook that makes them more visually appealing to the eye.
In the above figure, a style was added to the right bar graph that added some texture to the bars which in turn makes it easier to view. This may not seem like that big of a difference now, but when viewing a large report with multiple dashboards and graphs the change can make it easier on the eye in noticing positioning, shape, and layout. Thinking back to the second step of the perceptual process, these changes will make the data stand out even more to notice the discrepancies. Also, when it comes to shading and lighting of objects, the lighter the object is compared to its surroundings is always going to make it easier to view. This makes it better to use dark screens for your reports with lighter objects.
Another effective way of being able to visualize data from Tufte is the use of multi-functioning elements. When using pie charts and bar graphs, only one measure is being interpreted at a time. There’s nothing wrong with that, but you can save space by using charts that use data points to handle multiple measures. Then you are clearing space for another dashboard or graph and packing more data into a smaller area. Bubble plots and heat maps are examples of objects that can handle additional data fields by using the size (bubble plot) or color density (heat map) of the categories that they are representing.
In the figure to the left, you can see that the size of the bubble gives us an additional measure of the number of visits, but we can also use the X and Y axis for a total of three measures all fit into one chart. Depending on the data and how it is related, using these types of graphs can be very effective in getting a lot of quality data on the screen to be used for analysis. Then when it comes to the third step in the perceptual process, even more, data is being consumed at once to come to a conclusion.
Additional points to consider building out more effective charts:
Developers can start creating meaningful charts by knowing how to keep chart junk out, using the more eye-catching styles, and packing as much relevant data into the reports as possible. SAS Visual Analytics is a great tool for this type of reporting since it gives the developer all the creative options needed to produce effective reports.