#30: Panel bars
When we use stacked columns (and bars), we force the audience to do two things simultaneously — filter out the category they want to focus on and analyze data. Once they find something interesting, they have to refocus to read the label and then… repeat the whole process. This task is difficult because we overload our cognitive load by doing two tasks simultaneously. Additionally, we risk creating an optical illusion if we present numerous categories — like all countries in the EU. Having multiple bars makes assigning them to the country headache-inducing.
This is the case when you try to analyze the relative population change among different types of regions in European countries. Eurostat squeezed nearly 60 columns into one chart to present how the population of each of the three types of region will change during the upcoming 30 years. Theoretically, this gives diverse comparing options — we can analyze the change in one country, compare it with the average, or focus on the specific region type shift. But take a moment and try to do this to see how insightful that chart is.
Elements that work in this chart
✔ Using standardized units
Converting numerical values into percentages allows for comparing trends between categories that differ by order of magnitude. This conversion will enable us to easily compare data for populous Germany and sparse Croatia. Additionally, using the absolute values would require providing the base value to determine the accurate scale.
✔ Providing aggregated value for context
The integration of the average value for the European Union in the graph gives another point of reference. We can compare changes within one country (when analyzing one country level), across the countries (examine the same category across two countries), and at the Union level (when comparing the single country’s value with the average).
Elements that don’t work in this chart
✗ Showing everything at once
The main disadvantage of grouped bar charts is that they squeeze all data into a limited area. By mixing different information, there is no indication of what the viewer should pay attention to. In addition, if we show many categories — in our case, 32! — all the bars are compressed, and it’s hard to tell where the data for each country begins and ends. Despite using a variety of colors, it’s hard to focus on analyzing any value other than the one by which the countries are sorted (predominantly urban regions).
✗ Text orientation
Using a column chart is good if we have short categories’ names. In cases like ours, it forces us to rotate the label, which is literally painful to read.
Step-by-step improvements
✎ Clean up the chart
The first step is to remove unnecessary elements like the illustration and colored background. While cleaning the chart, we should also make the labels more readable. Turning the chart 90 degrees and switching from a column chart to a bar chart is the simplest solution.
✎ Split the grouped chart
The next step will be to separate the chart into three separate graphs. Each will plot the population change within one type of region and allow easier comparison within regions. The separation allows us to increase each bar’s visibility by tripling its width. It also makes the sorting logic visible.
✎ Allow easier comparison
Instead of showing the EU average as separated bars on top, we can represent it as a reference line. That way, we can easily compare the length of each bar with the average. It’s more precise than trying to compare the lengths of descent bars. Another improvement is graying out the negative part of each scale. Because we are not using the same range, the coloring works as a separation of the upward and downward trends and as a reference point across the regions. Lastly, we can simplify the footnote and incorporate the annotations into the chart.
✎ Work on formatting
Since the regions are already encoded by position (a separate graph represents each region), we can remove the color coding and the legend. Instead, we can use color to show countries below the European average. Choosing similar colors — darker orange for the reference line and lighter for the bars below the average — is intuitive while still well contrasted. Lastly, we can format text elements to create a visual hierarchy by de-emphasizing subtitle, axes values, and countries’ names.
Read the full article here
