Improve Your Bar Chart Visuals With These Five Tips
Hi, I’m da data guy!
I’ve been creating data visualizations for six years and over time, I’ve learned how to increase the effectiveness of my story telling by reducing my audience’s confusion. An example would be receiving questions like, “What’s this chart telling me?” or “How do I read this chart? Is it saying X is really that much larger than Y?”.
At first, I thought it was my audience who could not read a chart; I was wrong. The problem was me.
When building charts and visualization, I dove so deep, that I began taking shortcuts. I knew the answers to the questions that my audience would ask and by doing so, I started skewing the data. I wasn’t skewing it to portray my own bias’s or to have a feature look positive or negative. However, that is exactly what I was doing, unintentionally. Often times I thought I was doing my audience a favor by “helping” them out, but really I was making it worse.
Below are my top five tips to clearly communicate your data using a bar chart.
1.
When creating a bar chart, either vertical or horizontal, I would manually set the X or Y axis to a value that reduced the height of the bar chart. For example, if I’m creating a bar chart and my range is from 50–125, I would set the starting value of the axis to 40. Seems logical, why have it start at zero when the first value isn’t until 50? The problem is it’s now overemphasizing the distribution. If the dataset was larger, this becomes significant. The two images below highlight this.
I highly recommend starting the axis at 0 and only change the starting value of the axis when you or the audience wants to see the sensitivity of change in the data.
2.
Communicating where each column in the bar chart aligns with the X or Y axis points is critical. At first glance, it appears to be cluttered when adding the tick marks, however, it allows the audience an opportunity to trace the column they found interesting to the corresponding axis value. Omitting the tick marks does not give the audience the ability to accurately gauge the column’s value.
There is always an except to a guideline. If you have only a few columns, remove the tick marks and use data labels. This will present a clean chart.
3.
Formatting the column texts can be an issue when building bar charts. Some programs, such as Power BI, default to tilting the text 45 degrees when there is a certain number of columns. Others might change the text to be vertical, causing you to tilt your head to reach each column. I’ve published dashboards that have defaulted to displaying the columns formatted in this way and it’s caused issues for my audience.
A solution is turning the vertical bar chart to a horizontal bar chart and keeping the default length of the bars. Adjusting the length of the horizontal bars can lead you right back to the overemphasizing issue I talked about earlier. If your dashboard cannot fit the new chart, create a new workspace or ensure all charts have a beneficial purpose for your audience.
4.
Using a mixed bar chart and line graph is a visualization I have removed from my reputare. Often, a dual axis bar chart has a different starting position or increments the values on the second Y axis differently than the first Y axis. This increases the likely hood that your audience infers correlation or makes it harder to accurately read the visualization.
5.
Finally, be mindful when using a stacked bar chart. A stacked bar chart gives you details that sum to its total. When the audience tries to compare each category within the stacked bar chart, it quickly becomes difficult to accurately compare. If your goal is to compare the categories within each bar, I suggest using a small-multiples chart. A small-multiples chart gives you the ability to quickly compare categories using the same axis scale.
Conclusion
This wraps up my five tips on how to present your bar charts with less confusion to your audience. Remember, less is more; do not judge your effort on how simple your dashboard is.
Often, I find myself laughing at how I used to think, “If I can fit one more chart on here, I’ll be able to deliver more insights!”
Don’t be like my younger self.
If you’ve found this beneficial, please give me a follow or a few claps. I also highly suggest reading the Better Data Visualizations book as it’s helped me immensely.
As a final note, I’ve have also written a blog post that provides a step-by-step tutorial on how to create your own machine learning model to predict and create the probability of them leaving/exiting. Click the link above to learn more!
