Here are three common ways data visualizations can be misleading along with strategies for avoiding those pitfalls.
1. Shrinking Graphs
This doesn’t refer to a graph actually becoming smaller on the page. (Of course, displaying tiny graphs also makes it hard for your readers to distinguish details in your data. That’s a no-no as well.) Instead, a shrinking graph betrays its integrity in its baseline.
The baseline is the numeric starting point of the graph’s y-axis and is one of the most common ways data can be visualized in a misleading fashion. In fact, it’s extremely effective at making one group look better than another. In data visualization, these truncated graphs often make mountains out of molehills.
Let’s take a look at an example:
The chart below pits the product sales of Company A against those of Companies B and C. One glance and it’s clear that Company A has the top product. It’s doubling what Company B is doing in the same timeframe; 4.1 Million in sales is hard to beat, after all.
In fact, if you look at the values more closely, Company A is doing better than its competitors by a relatively slim margin. Laying out Company A’s 4.1 million next to Company B’s 3.8 million and Company C’s 4 million in this way is an obvious case of misleading data visualization.
If you take another view, before the graph underwent truncation, it’s clear as day that the data has been mishandled. And it’s an easy trick to spot, so changes like this won’t go unnoticed by outside viewers.
The best way to steer clear of these kinds of problems? Always start your graphs from point zero. Without truncating your charts, you won’t face the problems Company A will when its trick is (inevitably) sniffed out.
Of course, without truncation, charts can face the problem of reduced readability. Starting from point zero can make it hard to distinguish the actual differences in the data, defeating the purpose of data visualization.
This isn’t often a problem, especially when it comes to sales figures and similar data, where the actual number is the most important detail of the chart. However, there are exceptions to the rule. In most charts illustrating stock trading values, a miniscule fluctuation in a single number could wreak havoc in the industry. The differences matter, when it comes to certain areas of data visualization like this.
But as a general rule, avoid shrinking graphs.
2. Shifting Scales
Axis changing or manipulation is another common trick in the data visualization book. In a sense, it’s almost the opposite side of the coin of truncating data. It includes baselines and starts from zero, but it shifts and changes the scale of the chart so much that it doesn’t carry any more meaning.
Consider this chart on climate change that the National Review shared on Twitter:
Here, the scales have been shifted so much that the line is practically flat. The data seems to show that global temperatures aren’t rising, that climate change isn’t happening. All this has been accomplished by including values from -10 to 110 on the axis, completely diminishing any obvious changes in the data.
If you take a look at the next chart, made by more objective hands at Quartz, you’ll notice the data actually holds a lot of meaning in its small changes–changes that were eliminated in the previous chart.
This shows exactly how dangerous misleading data can be. Data manipulation can easily tilt the conversation one way or another.
In all honesty, this type of shifting scale is sometimes hard to pick up on, especially if the data isn’t scrutinized closely. However, as data becomes more easily accessible, it can be easily cross-referenced. With technology, readers could even make their own graphs—graphs that don’t possess the same problems in their axes or baselines.
3. Pick a Chart, Any Chart?
As data visualization becomes an increasingly important way of getting (and keeping) people’s attention, there’s greater variety in the types of charts and graphs being made. There are line graphs and bar charts, pie charts and spider charts. Today, charts are even available in 3D, which presents an altogether different set of pros and cons.
Exactly which type of graph should you use to visualize your data?
It can be hard to make that decision. But it’s an important one. The wrong choice can lead to very misleading or confusing data.
Here’s an example.
Look at this graph. Can you easily tell which candidate earned the popular vote? We can’t either.
In this form, it’s hard for the human eye to capture which slice of the chart is bigger than the other. And in this case, it’s vital to know exactly which is bigger.
To solve this problem, you could still choose this type of graph and just add a visible scale and lines to represent it. Including the percentage in numerical value also would make it much easier to distinguish Candidate A’s success from Candidate B’s failure.
However, it’s still not exactly the clearest picture.
Instead, if we take the data and change its form, we could present a much clearer, much less misleading set of data.
To avoid making these kinds of mistakes, you need to understand the graph’s purpose and usage conventions. In fact, the best way of making a graph clear, legible, and honest is to stick to what’s been done before.
However, to start you off, let’s focus on the purposes and conventions of the most commonly utilized charts out there.
The first being the simplest and most straightforward: the bar chart. These charts are universal. They allow for powerful distinctions to be made across categories or across discrete periods of time and hold immense power when it comes to comparing data in the marketing industry.
Another common type of data visualization is the line chart. It connects data from point to point and allows our eyes to track changes easily. Most importantly, it makes it easy to take note of trends in data or correlations between different sets of data.
The last of the big three is the pie chart. Within a set limit, usually a 100 percent point, the pie chart allows us to compare the parts of a whole. They can display how certain things have been allocated across a budget, for example.
This is only the tip of the iceberg when it comes to the art of picking a chart. While some of these chart profiles may be familiar to us, others might prove more challenging. In those cases, it’s best to rely on research and trial and error. Keep in mind that the choice isn’t just limited to the type of chart used. It’s also about the colors, axis, and many other factors.
For long-term success, it’s imperative to keep misleading data away from your visualizations. These days, the consequences of visualizing data senselessly are costlier than ever, whether it happens by accident or on purpose. So have faith in your chart, have faith in your data. And always aim for transparency and accuracy.
This post was written by Quincy Smith, part of the marketing team at Springboard, an online training company that provides mentor-led courses like their Data Science Bootcamp and Cybersecurity Career Track.