“If you torture the data long enough, it will tell you anything.” were the wise words of the late American mathematical statistician, John W. Tukey
It’s not too rare of an occurrence when we are asked to look at questionable data visualization presentations. Have you ever thought to yourself: “These charts just look too good to be true” or “Is the presenter trying too hard to push his motives, biases, and agenda?” or “Is this presentation purposely made to look so difficult to understand?”
Purposely or by accident, unfortunately, data visualizations these days are riddled with errors, mistakes, and bloopers. This was the topic of Yogi Schulz’s Analytics Unleashed 2021 presentation. Schulz is the Senior Consultant at Corvelle Consulting. He also came prepared with solutions to the most common of these bloopers.
“You can easily adopt these solutions to impress the audiences you present to” said Schulz as he began his presentation.
In essence, most of these errors are caused by one of these two scenarios. First, they are caused by the lack of thoughtfulness about how best to present data in the chart that will resonate with the audience. Sometimes, however, the author of the data visualization is using the chart to mislead the audience on purpose which is unfortunate and unethical.
The “Less is more.” quote by the late architect Ludwig Mies van der Roher was referenced by Schulz as he pointed out that “Overall, we should avoid cluttered charts and aim for a minimalist approach.”
Some of the most common bloopers Schulz talks about in his presentation are based on the mislabeling of legends and axis, chart annotations, and data clutters. He also shows strong examples of clear vs. unclear pie charts, and why bar charts or line charts are better in many cases.
Also, try to use a single Y-axis or at least a very few Y-axis instead of multiple Y-axis. Always ask yourself – is there really a relationship among the variables I want to show on one data visualization? Trying to save space or deliberately misleading your audience are not good reasons to use confusing and cluttered multiple Y-axis visualization.
However, one of Schulz’s strongest messages was to avoid the pitfalls of coincidental correlation vs. strong causation. “We need to be extremely skeptical of correlations because coincidental correlations will occur frequently during statistical analysis of every type of data,” warned Schulz.
Whenever possible you should be using motion in your data vs. statistical non-motion data visualization. This was the last point emphasized in the presentation.
If you missed Analytics Unleashed, don’t worry! It’s available for on-demand viewing.