Designing a Practical and Compelling Data Visualization Designing a Practical and Compelling Data Visualization
Data visualization is the most powerful way to communicate data to both technical and non-technical audiences. It’s also the most “taken-for-granted” component of... Designing a Practical and Compelling Data Visualization

Data visualization is the most powerful way to communicate data to both technical and non-technical audiences. It’s also the most “taken-for-granted” component of analytics. Few analysts have formal training in data visualization best practices.

Exacerbating the problem is that the default options in applications like PowerPoint and Excel are poorly designed; they’re viewed as acceptable or preferred when they should be seen as a starting point.

The purpose of data visualization is to allow the viewer to quickly and easily pull out the most important information from the data. I’ve found it frustratingly difficult to find a non-textbook resource that focuses on practical, business-focused best practices for data visualization that draw on knowledge of psychology and human behavior. Most of the content is graphic-design focused or designing for Verbal Presentations.

Yet, this gap is where most presentations exist for analytics professionals today — you have content that needs to do more than “look good”, must be intuitive enough to exist on its own in a slide or dashboard, and should be clear enough to quickly pull out the right message.

I recently completed a course in my Business Analytics Masters program focused on Visual Influencing. Its intention is to bridge the gap described above. It’s about taking what we know on how to gain attention with design and implementing it into business presentations.

To illustrate the best practices, I’ll start with the default PowerPoint chart and table to answer a simple question: Which world region was happiest in 2019?

Remove Everything From the Default

Have you ever (or commonly) seen a slide that looks something like this?

Default PowerPoint chart and table

The chart and table contain all of the needed information to answer the question but nothing pops. There’s a bland bar chart, nondescript header, and text table. To answer which region is happiest, I need to visually compare Australia, North America, and Western Europe, then reference the table on the right to compare values.

Let’s consider a few other items too:

  • Why do we need a y-axis that’s labeled at every number? No number falls below 4 and we have the numeric values on the right.
  • Do we need the color legend that’s added by default since we only have one set of values?
  • What about the slanted region values? It’s a high cognitive load on someone viewing the graph for the first time.
  • Why are the values ordered alphabetically?
  • Are the background grid lines helpful or distracting?

A few simple clean up steps addresses the above and begins to improve the chart design.

The y-axis was removed and the values added to the top of each bar. By adding the labels to the top, we can remove the table on the right, and the increased space allows us to widen the graph and remove the slanting without reducing font size. The legend was removed too and chart title improved. In summary, everything was taken away except for the bars themselves and the country names.

It’s now easier — but not easiest — to answer our question.

Reducing the Cognitive Load

Aggregating the right data and choosing the right chart are critical steps to communicating the data, but we can do more. We’ll add pre-attentive attributes, which are subtle alterations that we can make to a visualization to have our most important information “pop”. We’ll also adjust the graph to have a more conventional ordering that mirrors how we think — that is, left-to-right.

Note the pre-attentive attributes used: Color / shading, boldness, text size

There are a couple important changes to the above graph:

  • The default color of blue can be challenging for people with color vision deficiency. Inclusive design means that we should build with the assumption that at least one viewer of our design may be affected by this. Err on the side of non-blue colors if your company’s brand guidelines allows it.
  • We’ve brought attention to the highest average score through the pre-attentive attribute of color. Personally, I feel a color pops when set against grayed out values. Note too that the data label colors have been grayed out other than the one for Australia.
  • Next, still on the labels, the size of Australia’s data label has increased to show hierarchy in addition to bolding to draw further attention to the value.

The chart-viewing audience knows exactly where to look even if they don’t yet know the question being answered.

Bring it to the Finish Line

The information pops and we’ve created a logical ordering. To finish, remember that we’re trying to inform someone about which region is happiest in 2019. This isn’t a grand unveiling of information, so why would we rely on them to answer the question when we can give them that information along with the data?

Instead of a descriptive chart title, we can give an informative headline. With the headline, it’s implied that the numeric value represents the average happiness score, so we can remove the chart title. I’ve also grayed out all of the regions other than Australia and New Zealand because we’re not interested in which region is second, third, and so on.

You may ask why to even include the rest of the data if we give the answer, which is fair. It serves as context for Australia’s value and oftentimes audiences will feel more comfortable with an answer when presented with the information themselves rather than taking for granted the answer you’ve given.

I added a source at the bottom because we don’t always know who will receive our data and what they’re trying to do with it; adding the source reduces an additional question that you may receive later. Even for internal presentations and data, sourcing where you gathered the information can help your audience (and also help you should you need to reference the data again at some point in the future).

Bringing it Together

Before and After

This example is equal parts dramatized and very real. Take a look at presentations you’ve seen recently. Does it look more like the top visual or the bottom? What are two improvements you would make to the most recent visual you’ve seen?

The most important advice I can give is to remove every additional piece of information and chart design that’s added to a visual by default and add back the minimal amount of information to convey your message. Then, use color, size, bolding, spacing, and other pre-attentive attributes to enhance your message.

It takes work to scrutinize every visual that you create, but it’s also immensely satisfying to see the improved discussions that it can generate and the improved outcomes of your work.

Originally posted here.

Jordan Bean

Jordan Bean

Jordan is an analytics professional with an interest in data storytelling, visualization, and simplifying complex topics into interpretable messages. He's currently pursuing a Masters degree in Business Analytics at Wake Forest University while working in analytics for Liberty Mutual Insurance. Prior to that, he worked in consulting for Private Equity firms on strategy and buyouts. Feel free to connect at https://www.linkedin.com/in/jordanbean/ to talk through any thoughts on articles or breaking into the data field.