Skip to Main Content

Evaluating Online Information

Evaluating Data Visualizations

Trifecta Checkup

Use a set of criteria like the Trifecta Checkup to evaluate data visualizations. It is also just as important to evaluate the data source -- read more on that below.

1. What is the question?

Ask yourself: What is the purpose of the visualization? What is the claim?

2. What does the data say?

Ask yourself: Is the data relevant to the question? Can the question be answered with this data?

3. What does the visual say?

Ask yourself: Is the visualization accurate to the data?

 

The video below provides an overview of how data visualizations can be misleading. See also the Good vs. Bad Data Visualizations section below for examples.

Evaluating Data and Statistics

It's just as important to evaluate the underlying data and statistics being represented in the visualization. Statistics can be easily manipulated or shared out of context in all types of media. In general, when looking at a visualization, if the data source is not listed or seems suspect, don't trust what you see!

Use these questions based on the SMART Criteria (University of Washington) to evaluate a data source or visualization.

Source: Who or what is the source?

What are their credentials?

Motive: Why did they collect the data?

What is the potential for bias based on their motive?

Authority: Is the source authoritative or reliable?

Review: Review the collection methods and completeness.

Who or what is not represented?

Two-source Test: Double check all claims if possible.

Do other sources say the same thing?

Good vs. Bad Visualizations

The Y-Axis on a bar chart must always start at zero, because if it does not, other bars will double in size when the difference is not truly double. See the example below. In the first example, the value of 2.1 million appears double the value of 1.8 million, because the y-axis does not start at zero. In the second example, the y-axis starts at zero, so the differentiation between parks is visualized more accurately.

Bad Bar Chart Good Bar Chart
bar chart showing national parks visitation, but the 2.1 million bar appears more than twice the size of the 1.8 million bar, because the y-axis starts at 1.4 million rather than zero. national parks visitation, but now the y-axis starts at zero, which better represents the differentiation between parks.

Another thing to consider is when to use certain types of charts. For example, pie charts should only be used when visualizing parts that make up a whole -- they should be percentages that add up to 100%. Below, the first example adds up to 193% -- a better way to show the part-to-whole relationship is among each candidate, as shown in the bar charts on the right.

Bad Pie Chart Good Bar Chart
a pie chart that says 70% of respondents have a favorable view of Palin, 63% of Huckabee, and 60% of Romney. However, this does not add up to 100% so does not work as a pie chart. a bar chart showing % republicans by opinion of each candidate. Here, we have multiple bar charts (one for each candidate), that shows the percent favorable, unfavorable, or can't say. this shows better the opinions of each candidate.