Deciding Which Type of Data Visualization to Use

Apr 01, 2018

During this video, you will learn how to distinguish between the purpose of different types of data visualizations. Choosing a method of data visualization requires determining the most effective way to convey your data story.

There are as many ways to visually present data as there are stories it can tell. Choosing a method of data visualization requires determining the most effective way to convey your data story. This is not, however, to suggest that your visualization should in any way be used to distort or misrepresent the data itself. At base, your visualization must be informative, relevant, unambiguous, and true. That said, different types of visualization have different strengths.

Form, color, and position all have significant parts to play in determining the most effective way to tell your data story. In the words of statistician and data visualization expert, William Cleveland, some display methods lead to efficient, accurate decoding. And others lead to inefficient, inaccurate decoding. While many consider pie charts both basic to and indispensable for presenting data, they are best used sparingly, if at all. They can be used to display part-to-whole proportions and then only to compare a few limited categories of information such as line items in a project budget. Beyond four or five categories, pie charts tend to become confusing and difficult to understand.

Better and increasingly more commonly used are line and bar graphs, which depict data points connected by straight lines. As our eyes are well suited for comparing differences in two-dimensional location and differences in line length, these types of graphs are able to show distinctions and values with immediacy and clarity. Both are generally easy to read and comprehend at a glance. Line graphs are generally used to illustrate specific changes, such as sales figures over time. In the case of many variables, a stacked line graph can be used. The length, width, and color of the lines can all be adjusted to make differences more readily apparent and overall comprehension more immediate.

Note too, that bar and line charts can be oriented both horizontally or vertically, depending upon which best visually tells the data story. Similarly, bar graphs are useful to track changes over time for comparison analysis between limited numbers of variables. The length of each bar in the graph is sized in proportion to the section it represents and can be horizontal or vertical. It's recommended to display data in order using vertical bars, like a progression of years and horizontal bars to display data by category, such as bestselling car brands. For various categories of data, a grouped bar chart can be used when each categorical group requires two or more bars. These bars can then be color coded to represent a particular grouping.

For example, a business owner with multiple stores might make a grouped bar chart with different colored bars to represent each store. The horizontal axis might show the months of the year and the vertical axis would show the revenue for each store. Similarly, stacked bar charts are used to emphasize a comparison of groups of data relative to other groups. The focus of stacked bar charts is on the big picture, not the finer details. For instance, if a company wants to compare three of their best sellers over a period of four months, a stacked bar chart is ideal, providing an overall view of each item sales in comparison to the others.

But while line and bar charts are best able to tell data stories relating to categorical data, when that data corresponds to a few specific discrete categories, they're not suitable for every story. Consider scatter plots, which illustrate the relationship or dependency between two sets of data, and generally plotted on an X and Y axis. These are useful when there are many data points to work with, such as demonstrating the relationship between two variables, such as volume and cost per unit. If a scatter plot has more than two variables, it's referred to as a scatter plot matrix. The matrix is a group of scatter plots arranged into a grid. The grid is used to ascertain if there are linear correlations between numerous variables.

Other types of data visualizations include things like map charts, which are useful when data relates to geographical comparisons, and bubble charts, which are a type of scatter chart used when the data has three variables. The sizes of the bubbles are determined by the values of specific data points and they are then placed along the X and Y axis to represent the other two values.

There are still further, more specialized data visualizations. But the point is that, technically, any data can be displayed in any type of graph or chart. However, some are more suited to telling specific stories than others. Using a pie chart to show changing sales over time, for example, is ineffective. A simple bar graph is more effective. But no matter how simple or complex your data is, visualizations can make it more compelling and engaging, allowing your audience to get a clear picture of what the data is saying and how it can be used to inform business decisions.