4 tips for designing data visualizations people want to see
Data visualization techniques have rapidly evolved over the last 10 years thanks to powerful evangelists such as Stephen Few, Edward Tufte, and David McCandless. With their guidance (and that of many others), analysts have learned to leverage color, fonts, titles and various charts to better tell a story with our data. As our visualizations have become more refined, we have learned they can evoke a powerful emotional response from the audience—but is it the right one?
Data visualizations and feelings
An example of how data visualizations can evoke an emotional response is illustrated by an analysis of the world’s deadliest animals. Both of these images use the same data set, but the approach in design elicits a different response from the audience.
The data visualization below delivers a more powerful message emphasizing humans and mosquitoes as the most dangerous animals on earth.
- • Images – The smiling snakes, alligators and humans do not align with the dangerous message in the analysis.
- • Color – By representing the mosquito with red, the chart on the right conveys a sense of danger.
- • Charts – The horizontal bar chart on the right shows the scale of impact better than individual data points on the left.
With three simple choices, the designer was able to influence how the audience perceives the data. As you build your own visualizations, you need to ask yourself what type of emotional response you are trying to elicit from your audience.
The challenges of ‘personal’ data
Analysts working with Human Resource or Performance data must be extra careful in how they approach data visualization. Instead of the data set representing a large population (like in Data Science), the data may track the performance of an individual. Because the subject matter is personal, the same visualization may evoke feelings of joy or pride in one person, but lead to powerful feelings of resentment, anger, or shame in another.
“Shame is an existential feeling of unworthiness,” Dr. Bryn Jessup says. “When people feel shame, they believe that they are ultimately an inadequate person or an unworthy person.” If your data visualization has the potential to single out the performance of an individual, you must be extremely aware of the potential impact it may have. Will the individual feel shame and immediately dismiss the analysis, or will they view the data pragmatically?
In a Seattle Tableau User Group meeting in early 2020, Karl Moser, a Master Change Facilitator at Providence Health, talked about how his healthcare dashboards are perceived by clinicians. He emphasized that shame is often the most powerful form of resistance when introducing new dashboards and highlighted multiple ways to reduce it in his audience.
1. Highlight outliers, not poor performance
When your data represents the performance KPI’s of an individual, be extremely conscious of how you describe the variance in the results. Instead of describing a performance KPI as either “positive” or “negative”, think about highlighting the individuals as “outliers”, and provide the user the ability to explore the data to better understand the result. In the chart on the right, I’ve identified Cindy Stewart as an outlier and provided additional detail into why this may have occurred. This is a less confrontational approach than the chart on the left which simply identifies Cindy as a low profit customer.
Potential Conflict: Cindy is colored red and is associated with the ‘Low Profit’ group. Both design choices could promote shame.
Highlights Outliers: Cindy is listed as an ‘outlier’ and an additional chart provides more detail. We now know that a 70% discount was applied to one of her orders which contributed to the low profit.
2. Show progress, not failure
Instead of using fixed reference points which may indicate failure, use KPI’s that also show what is required to achieve success. Conversations with an individual can be more productive when the discussion focuses on the success KPI. Take the following two statements:
1. 'Draymond, you are 52% behind target and are failing to meet your goal.’
2. ‘Draymond, you are behind target, but only need $5,148 to reach your goal.’
The second statement is much more supportive and will lead to a more productive conversation than the first. Try to frame your dashboards so they support these objective conversations.
Potential Conflict: Draymond is highlighted in bright red and focuses on being 52% behind goal.
Shows Progress: The chart above uses less confrontational colors and focuses on the $5,148 Draymond needs to achieve his goals, not being 52% behind.
3. Use neutral colors
The use of color in your data visualization can have a meaningful impact on your audience. Instead of using traditional stoplight color schemes to highlight performance, use a neutral color schemes which are less likely to evoke feelings of shame. Blue, grey, white, and even green have a much smaller chance of creating a negative response in your audience. Try to avoid using negative colors (such as red) when the data represents an individual’s performance because it may lead to feeling of anger or shame.
It should also be noted that traditional stoplight palettes can create difficulties for individuals with color processing differences.
4. Include trends to start conversations
Instead of using a fixed KPI, include historical trends to provide a more comprehensive picture of the individual’s performance. Visualizing performance trends often leads to better conversations. In the example below, the first chart uses spark lines to show the trend in sales performance. The end-user can now see that Russel’s sales may be low, but he has shown a steady level of improvement over the last 4 weeks. This is a much less confrontational approach than the conclusions presented in the second chart.
Potential Conflict: Russel is listed as a poor performer in negative, red colors.
Using Trends: Russel is still ranked at the bottom of sales performance, but the last few weeks are trending up.
In conclusion, you should stay aware of how words, colors, formats, and context presented can have an emotional impact on your viewer—especially when end user data is on display. The guidelines above should improve the acceptance of your visualization, which in turn should lead to better conversations and productivity.
Special thanks to Simon Scarr, Karl Moser, Seattle Tableau User Group (SEATUG), Andy Cotgreave, Lauren Carrane, and Anil Celik.
Need help with data visualization?
See how Logic20/20 can help.