How to present qualitative data
A friend of mine recently said that he thought one of the barriers to people using data visuals to shorten and summarize reports is that many are not just used to thinking about it; how could these facts and figures be presented visually? This goes particularly for cases where much of the data is qualitative in nature.
Because I think that this is probably very important, I am building this envisioning process into my data presentation course. There are many examples to work with, but I’ll give you a few here. I have provided several example in interactive D3.js forms (unfortunately this doesn't seem to be possible within LinkedIn post) and several others as static charts. Interactivity can be helpful, as we shall see.(1)
As with visual data analysis and presentation, we are taking advantage of the pictorial superiority effect and principles based on characteristics of visual perception and memory.
The “pictorial superiority effect” is just a fifty-dollar term for the fact that large parts of our brains and the activity therein are devoted to visual processing, and that what we see can override other senses.(2) The principles of visual data design are based on how this processing works, exploiting our tendencies to perceive certain forms immediately and to be limited to processing only so much at a time. There is lots of material on this in my course, past blogs, and the writings of key visual data proponents like Colin Ware, Edward Tufte and Stephen Few.
We should want to use this pictorial superiority to allow us to simplify and effectively present qualitative data when we can, just as we should do with quantitative data. I find that when you start to think about this, there isn’t as much purely qualitative data in most reports on policy, performance, evaluations, organizational planning and so on as you might have thought. Posing some questions of the data helps, and we’ll talk about that in a moment.
First of all, however, let’s recognize that a lot of qualitative information is associated with a typical quantitative chart. Incorporating that data and tying it in as closely as possible with the chart makes the best use of both kinds of data. This example connects the qualitative information directly to a specific part of the visual, using a tooltip ( in the interactive version). The information is focused and of necessity, brief.
But what about visualizing qualitative data itself?
Quantitative data comprises cardinal numbers, that denote quantity, and that can be compared as ratios in a given metric. Qualitative data does not have that characteristic. It can, and very often does, however, have ordinal or relational characteristics. So a list of findings, from the most important to the least important, displays an ordinal relationship. Textual description of an activity causing something else to occur displays a relational characteristic. If these characteristics are present, you can visualize the data.
I think the best way to determine whether or not this is possible is by asking questions about your data. For example, in determining if the data can be ordered, among the questions you might ask would be: “Are different states of this data preferred to others"? You probably already arrange this kind of data into tables with “positives” and “negatives”, or you could assign subtle colour coding to your bullet points, among other strategies.
A trickier case might be when you need to compare qualitative assessments of different programs or products across multiple dimensions. As long as you can reasonably assign each product’s characteristics to levels on a rough ordinal scale, you can present this information visually, avoiding a lot of text. The following example is a parallel coordinates chart.
This chart allows a quick and intuitive comparison of alternatives and their features. With a bit of annotation, either in strategically located text boxes in a static chart, or using tooltips, as in this D3.js chart, you can provide critical interpretive and contextual information (in the interactive version).
Another tricky case is where many things are related to some extent to each other. Asking yourself “Are these things inter-related, and can we rate the strength of their interrelation or assign an order like “this affects that”? Here’s an example of the latter, where the direction of causation goes from y-axis to x-axis (e.g. speed affects cost). Our rating of the strength of interrelation is reflected in the box shading. This rating could be as simple as “highly related”, “related” and “unrelated” and still yield an informative visual summary. In the interactive D3.js version, I have included a mouseover function to help ensure that the right axis labels are read for any given box. Annotations to the right of each row, or tooltips, could provide context on the relationships.
Other cases are more obvious.
When your qualitative data concerns things happening over time, the obvious solution is to use a timeline. This is much easier than paragraphs of text. Here is a static example:
What about plans or recommendations that are contingent on certain conditions, events or findings? A tree diagram, with branches from the points of contingency would help. Here’s another static example. Very complicated versions of this can be handled by making the chart interactive such that it collapses the branches that you don’t need to see at the moment.
How about progress toward a goal? Bullet charts will do nicely, even if you are just able to roughly order things in terms of completeness. And so on.
So, you can ask yourself questions, like:
"Can this be ordered"?
"Is there improvement or deterioration in the situation"?
"Are some of these things related to each other"?
"What happens if this happens"?
The list goes on.
You’ll find that very often the questions will lead you to see that you could summarize your information with only an assignment of order or relationship.
Finally, visual elements can also supply context to qualitative information - e.g. a map or diagram in an infographic page, or annotations built into a map. A few words, however, about “infographics” and “data decoration”.
Many examples of highly informative and engaging infographics are available. I like to promote examples, such as this one, from Alberto Cairo’s books.(3) This is a great approach to presenting a mix of quantitative and qualitative data, or giving context to either kind of data.
This excellent presentation of both quantitative and qualitative data, supporting each other and telling a clear story, is in direct contrast to much of what is unfortunately out there, where “data decoration” stands in for communicating with data. Effective data presentation, including presentation of qualitative data, does not include sticking bright numbers in front of text and stacking up pictures of little people, etc. Doing these things generally violates the key principles of visual data presentation, where the data/ink ratio is kept low and where shape, colour and contrast are used to work with our perceptions.
Few of us want to follow rules and guidelines for the sake of doing so. But if we have a few principles to work with, we can present our information succinctly and with impact.
So, there we are. Some thoughts on how to take some of those pages of text and condense them into effective and engaging visual presentations and summaries.
(1) D3.js is a javascript library that allows you to create just about any kind of data visualization. One of the great things about it is that D3.js versions of charts can be included in interactive format in any html page, like a website, without great expense or any need to post your work on a third party site.
(2) Surely we are all acquainted with “blind taste tests”.
(3) See Alberto Cairo’s books: “The Functional Art”, 2012 and “The Truthful Art”, 2016, New Riders.