article Choosing the Right Visualization
Choosing the right visualization for your data can be a daunting task. You may put in countless hours gathering requirements, identifying the data needed to solve your business question, and triple-checking your calculations for accuracy. As you near the finish line, you select the first visualization that seems to suit your data. You don’t spend too much time on this step because you believe the data speaks for itself.
While that may be true to an extent, choosing a visualization can have a large impact on how well your data is received. Selecting a poor visualization can quickly undo your hard work and cause unnecessary confusion for your audience. Picking a good visualization can help paint a more polished, easy-to-understand story and create a lasting impression on your audience.
This doesn’t have to be an unsettling task if you have a clear idea of the end goal in mind. Let’s walk through a few common visualizations for expressing your data story.
TO SHOW HOW VALUES COMPARE TO EACH OTHER:
This common visualization is effective at comparing categories or trends over time. It tracks which groups are most common and how certain groups compare against others. Typically, column charts are used with ordinal variables that follow a natural progression.
Horizontal Bar Chart
Similar to column charts, bar charts can be utilized to address long category labels that may overlap on a vertical chart. They are also better for visualizing nominal variables and look best when sorted from highest to lowest value to distinguish the difference between values.
These graphs are best for depicting trends or progress over time. More specifically, they are good for showing relationships using periodical data. Adding points to the lines can also assist the viewer to follow the pattern.
Tables can be utilized when precise values are required but are not recommended for showing change over time. This can also aid in showing data where the scale of certain figures is too large to accurately portray the difference between the figures on a chart.
TO SHOW HOW VALUES RELATE TO EACH OTHER:
This visualization shows the proportions of an occurrence using a color legend. The variation in color should be added using any hue or tint that easily provides visual cues to the viewer about how the event is clustered over an area.
This graph shows pairs of numerical data, with one variable on each axis, to find a relationship between them. If the variables are correlated, you may see patterns within the data or clusters of data points.
TO SHOW THE COMPOSITION OF DATA:
These are similar to line charts, but area charts utilize colored regions to show how the parts of a whole change over time. It’s best to use a 100% stacked area chart when possible to show composition, instead of using a regular area chart which can be misleading.
Pie and Doughnut Charts
Pie and donut charts can be used to visualize a part-to-whole relationship; so the total sum of all segments must equal 100 percent. In general, pie charts should only be used to display 5 or fewer categories to avoid confusion. However, pie and donut charts should be avoided whenever possible, as differences in areas and angles can cause confusion.
Stacked Column or Stacked Bar Charts
These charts illustrate the composition of data over time. A drawback to using this visualization is difficulty comparing all but the first series. Adding value labels to each category helps distinguish the difference. Limiting the number of categories also helps this visual from becoming convoluted.
TO SHOW HOW THE DATA IS DISTRIBUTED:
These look similar to bar charts, but measure frequency, rather than trends. The x-axis is comprised of bins, which are groups of values. The y-axis represents the frequency count.
This visualization displays the distribution of data based on the full range of variation. It includes the maximum and minimum values, median and range between the first and third quartiles. Box plots are especially useful for showing outliers.
This method is similar to visualizing a box plot, except that it plots the numeric distribution of data and its probability density. While they are less popular, they are more informative than box plots.