Bar charts are one of the most common and effective ways to visualize data, especially when you want to compare categories, frequencies, or proportions. However, if you are not careful, you can also distort the data and mislead your audience with your bar charts. In this article, you will learn how to avoid some of the most common mistakes and pitfalls that can ruin your data visualization.
One of the most important decisions you have to make when creating a bar chart is how to scale the vertical axis. The scale should reflect the range and distribution of the data, and avoid exaggerating or hiding differences. A common mistake is to use a truncated scale, which cuts off the lower part of the axis and makes the bars appear larger or more dramatic than they are. This can create a false impression of the magnitude or significance of the data. To avoid this, you should always include zero as the baseline of your bar chart, unless there is a valid reason not to. You should also avoid using a logarithmic scale, which compresses the higher values and expands the lower ones, unless you are dealing with exponential data or you want to highlight the relative changes.
I believe, starting a bar chart from zero is essential to maintain accuracy and fairness in data representation. It prevents the exaggeration of differences between values, ensuring that the chart provides a truthful and unbiased view of the data. This consistency in scale is a fundamental principle for creating reliable bar charts
What about embedding different scales or scale levels in one chart, sort of categories for some sort of comparison, for instance, percentage score by gender and then by age group in a section of the bar chart and another respectively...I see this a common practice at times and I don't know what to think of it beyond the fact that it's misleading, to say the least.
Never truncate the axis of a bar chart. I can not emphasize this enough. It would most likely create a misleading chart. You may do it for a line chart but never for a bar chart because it tricks our brain into thinking that the variations are bigger than what they actually are. Bar charts can be horizontal or vertical. The scale will be mostly determined by the length of the bars. If they are similar, then you can choose to label them directly instead of using an axis so people can understand the differences more clearly.
Generally I would concur, zero is the safe bet. ?? However, there are times (especially using percentages) that setting a baseline of 50% might help highlight the trends you want to illustrate. The key is you be consistent in the other charts surrounding that analysis. Don’t confuse people with different baselines on each chart.
o avoid distorting data in a bar chart, you should follow best practices in data visualization and ensure that your bar chart accurately represents the data without misrepresenting the information. Here are some key guidelines to prevent data distortion in a bar chart. Use the Right Chart Type Start Bars at the Same Baseline Use Appropriate Colors and Patterns Provide Context Use Charting Software Avoid Gaps Between Bars Keep Bars Thin and Uniform Label Your Axes Clearly Maintain a Consistent Scale
Another factor that can affect how your bar chart is perceived is the width and gap of the bars. The width of the bars should be proportional to the data they represent, and the gap between them should be consistent and clear. A common mistake is to use variable widths or gaps, which can create confusion or bias. For example, if you use wider bars for some categories than others, you might imply that they are more important or more frequent. If you use different gaps between bars, you might suggest that there is a relationship or a hierarchy between them. To avoid this, you should use equal widths and gaps for all bars, unless you are using a stacked bar chart or a histogram, where the widths and gaps have a specific meaning.
The bars should be equidistant from each other unless we are visualizing another variable with that distance. For instance, if we are showing the amount of sales by year, and some years have zero, then a bar will be missing that year, but it is for a purpose. Ideally, bars should be wider than the space between them.
To avoid data distortion in a bar chart, ensure proportional scaling, appropriate bar width, and starting the y-axis at zero. Use clear labeling, avoid 3D effects, and minimize gaps between bars. Maintain a consistent bar order, thoughtfully use colors, and provide annotations for clarity. Testing with diverse audiences helps confirm understanding. These practices guarantee that your bar chart accurately represents data without misperception, aiding your audience in interpreting the information correctly.
Using the right bar width and spacing is crucial for a bar chart. It ensures that the chart is clear and easy to read, allowing people to compare data without confusion. If you get the width and spacing wrong, it can make the chart hard to understand. So, make sure to use the appropriate measurements for the best results.
Consistency in bar width and spacing is essential for an unbiased bar chart: Match bar widths to the data they represent. Maintain uniform gaps to prevent misleading implications. Only vary widths or gaps for stacked bar charts or histograms where they convey specific information.
A third aspect that can distort your bar chart is the use of 3D effects or other embellishments, such as shadows, gradients, or patterns. These can make your bar chart look more attractive or fancy, but they can also interfere with the accuracy and clarity of the data. A common mistake is to use 3D bars, which can create optical illusions or distortions, such as making some bars appear longer or shorter than they are, or hiding some bars behind others. This can affect how your audience compares and interprets the data. To avoid this, you should use 2D bars, which are simpler and more honest. You should also avoid using unnecessary or distracting embellishments, such as colors, shapes, or labels that do not add value or meaning to the data.
Have you ever heard of the data-to-ink ratio? This concept was introduced by Edward Tufte, and it simply refers to how much ink we use to represent the actual data. In this context, 'actual data' refers to the most crucial data elements necessary to convey the data story. While 3D embellishments may enhance the aesthetics of your charts, in my experience, they tend to introduce clutter into visualizations without significantly contributing to the data-story. The fewer 3D elements you use, the better the focus on the actual insights.
I said this before, but I will say it again - Do Not Use 3D Bar Charts. Representing a 3D object in a 2D space doesn't work. Keep everything 2D in a 3D space (screen or print output.) Once you start introducing 3D to your visualizations, you dramatically increase the time it takes to interpret what the chart is saying. Additionally, you run an extremely high risk of your users misinterpreting your 3D bar chart. This is just an unnecessary capability that I highly recommend staying away from - it causes more problems than it solves.
3D look might seems cool, but it can actually mess up how you see the data. Stick to a regular, flat design for your bars so that everyone can easily understand and compare the information in your chart. It’s all about keeping things straightforward and clear.
You can do the test yourself, prepare a graph, and then make a 3D version of it. Are you able to identify the relative differences between the bars in both charts equally? Colors, shapes, and labels are extremely useful when used purposely. If they don't add value they are adding noise and should be eliminated. But a well-used color can highlight the main insight, and there are a lot of elements that work as well (look for gestalt principles).
??!!! K.I.S.S. (keep it simple) It requires little effort to confuse people. The sign of a true professional is how simply they can illustrate their point.
A final aspect that can improve or ruin your bar chart is the way you label and title it. The labels and title should provide enough information and context for your audience to understand the data and the message you want to convey. A common mistake is to use vague or misleading labels or titles, which can create confusion or misinterpretation. For example, if you use abbreviations or acronyms that are not familiar or explained, you might lose your audience's attention or trust. If you use a title that does not match the data or the purpose of the bar chart, you might create a false expectation or a wrong impression. To avoid this, you should use clear and concise labels and titles, which describe the data and the main point of the bar chart.
Don't miss out on the opportunity to have a title that highlights the insights. A descriptive title like "Count of sales by the state in 2023" is less efficient than "Texas has doubled the sales of California for the third year in a row". You can add a subtitle explaining that the information showcased corresponds to 2023. The labels should add value, remember that if it doesn't add value then it's noise and should be eliminated. You probably don't need to label every bar, but the smallest, biggest, and maybe another that has an interesting story to tell. In that way, you are focussing the attention on those bars and the audience can have a clear idea quicker.
Providing context through titles, axis labels, legends, and reference lines are essential. These elements collectively contribute to a more effective and comprehensible bar chart, ensuring that viewers can easily understand the data being presented and how to interpret it accurately.
While labels and titles may convey your data, I believe that citing your sources is more crucial for demonstrating the trustworthiness of the information you present. Clearly listing where your findings are derived from allow stakeholders to dig deeper into the data on their own and cross reference your findings to prove to themselves how trustworthy your insights are.
Avoid adding unnecessary decorations that might distract from the data. Keep it clean and simple so that the important information shines through clearly, making it easier for users to grasp the message you’re conveying. Also, be mindful with color selection to make your chart inclusive and user-friendly.
Two simple mind hacks for internal audiences. 1 - Use your business’s brand colors on charts and graphs ?? ?? ?? it will evoke emotion because it’s familiar 2 - As much as you might hate it, most executives still live in Excel and PowerPoint. As pretty as your dashboards are, sometimes giving your stakeholders the data in excel or PPT can increase their trust in it (even though nothing else changed)
Avoid using borders, gridlines, colors, fonts, axis titles, and any other element that adds noise instead of information to the story.
Consider the orientation of your chart. In some cases you may be able to make better use of the space using horizontal bars, and fit labels in an orientation that readers may find easier to read and understand.
Working for major firms like KPMG and having had to provide data visualization tools as part of IT solutions the most underpinning principle to keep in mind is that your are communicating an idea. And with all business transactions what your audience needs to know is that you are marketing a pitch to them which you belive is a good deal. They need to know that due diligence is part of their responsibility and that you have assured them that you are clear about your interests and come in good faith. Data visualizations adds to the charisma of a presentation, however the pitch itself should be substantial and maintain integrity.