Mastering Data Storytelling with Visualizations
Machine Learning Reply GmbH
We fuel entrepreneurship by leveraging leading-edge technologies.
Author: Dilara Ademo?lu
Crafting Clear, Compelling, and Clutter-Free Insights??
At its core, data storytelling is about creating visualizations that convey a clear, impactful message. Storytelling is essential in business, with limited time and short attention spans. Most executives don’t have time to sift through raw data. Crafting a clear story with your data is the key to being listened to or influencing decision-making.??
This article will discuss important tips for improving your?data visualizations?and explain the concepts defining a data store. We will also discuss points to be mindful of while visualizing the data and what not to do to make your visualizations?more human-friendly.?
Data Storytelling: Setting the Scene for Your Audience?
An effective data story is understandable, quick, and impactful—a narrative that guides the audience directly to the insight. Businesses often create fixed reports or presentations to communicate these insights to customers or stakeholders, turning complex data into a clear and compelling message. In many ways, every presentation or report is an act of storytelling designed to help audiences make sense of complex information.??
Using visuals is one of the most effective ways of storytelling for human audiences, as interpreting visuals is an innate way of processing the world. Well-designed visuals allow us to grasp the facts behind the data quickly and intuitively. Let’s look into the purpose of visualizations first to design our visuals better.?
Understanding the Purpose of Visualizations?
The purpose of visualizations can be classified into two segments: form and function:?
?
Based on the above-explained segments, we can combine one form and one function to create different visualizations for different purposes.?
Static explanatory visualizations are often used to present straightforward insights that viewers can interpret at a glance—like a snapshot. For instance, simple line graphs or bar charts can clearly convey trends or comparisons without needing further interaction.?Static exploratory?visualizations let viewers interpret the data and find their own results in a static display of information. An example of this could be maps.?
The other end of the Form spectrum is the interactive visualizations. Interactive explanatory visualizations follow the lead of static explanatory format and present the insights at a glance. The difference between the two is that interactivity is usually provided through the hover function for presenting more details. An example is given below: an interactive line chart illustrating a particular KPI and allowing the user to observe the change details independently.?
?
However, not all data stories are meant to be fixed. The interactive exploratory visualizations encourage viewers to engage more deeply with the data. The basic example of these visualizations would be the classic dashboards, which allow the user to filter the data, change the type of visualization, and discover the insights they require by interacting with the data. This format is essential to data exploration, allowing users to investigate deeper and draw individual conclusions. One great example of this format is the Gapminder demographics visualizations by Hans Rosling. The exploration of country demographics becomes interactive by various inputs that the users can select and adds a touch of animation.?
?
Whether static or interactive, explanatory or exploratory, the goal of data storytelling is the same: to make data meaningful. When you present data with a clear narrative, your audience isn’t just looking at numbers or graphs—they’re seeing the story those numbers tell.?
Choosing the Right Chart for the Story?
Now that we’ve covered the purpose let’s get to a crucial question: what type of chart should we use to tell our story? The answer often depends on the use case and the audience.??
For example, horizontal bar charts work well for most data but may be difficult to read if you have long labels. In that case, a vertical bar chart might be better. If you’re visualizing percentages, a stacked bar chart or pie chart can be effective. Pie charts are especially helpful when a segment is significantly larger or smaller than others, while line charts excel at showing trends over time.?
To compare proportions, a pie chart might suit a simpler dataset, while a treemap is ideal when you want to maximize space. Tables can be powerful if you have only a few numbers to communicate, but they become hard to interpret if the audience must scan through extensive data. Selecting the right format comes down to both the message and the specific needs of your audience.?
?
Understanding Your Audience’s Perspective?
As a data storyteller, knowing your audience is just as important as knowing your use case. A well-crafted data story aligns not only with the data and visuals but also with the audience’s knowledge level, ensuring that your insights land effectively. Put yourself in their shoes and consider how they’ll interpret the visual. The way you visualize data would be different depending on your audience’s background and familiarity with the subject. For example, if your audience is business professionals you would want to display clear insights quickly, however, if the audience is business or data analysts then your visualization could be open to more exploration. Still, in any case, your data visualizations will be made for human decision-making rather than machine processing, no matter who your audience will be. This brings us to points you should consider while visualizing data. We can emphasize effective data storytelling (with data visualization) to three principles:?
?1. Make it as human-friendly as possible. Visuals should be intuitive and easy to interpret.?
2. Embrace minimalism. Minimize cognitive load by removing unnecessary elements like excessive white space, gridlines, or decorations.?
3. Highlight insights for clear takeaways. Focus on what matters most to avoid visual clutter.?
Use Human-Friendly Charts?
The human brain is remarkably adept at perceiving patterns, but not all visual encodings are equally effective. When designing charts, it’s crucial to play to these strengths, ensuring that your visuals are not just appealing but also easy to interpret.?
?For example, humans perceive numbers encoded by length or position far more accurately than those encoded by areas. This is why bar charts, which rely on length comparisons, are often more effective than visualizations like bubble charts or pie charts for detailed comparisons. That said, pie charts have their place—they’re excellent for emphasizing big or small proportions within a dataset. If your primary goal is to highlight that one segment dwarfs the others, a pie chart does the job well.?
领英推荐
?Let’s take a look into the following example, we asked the question “What AI tool or model have you primarily used for work or personal projects in 2024?” in a survey to our colleagues. Given 5 different options to answer, the distribution of answers turned out to be close to each other. If we tried to visualize this distribution, the pie chart looks nice, but we cannot interpret the differences between the answers clearly. If we convert the same data to a bar chart, then we can see the distribution of the answers much more clearly.?
On the other hand, when we visualize a yes-no question from our survey “Do you use GenAI tools for work or personal projects every day?” with a bar chart we can clearly see the difference between the answers, but if we choose to use a pie chart for this data, we would get an equally clear impression of the data, which can be easily highlighted.?
?
For datasets that demand more detail, treemaps can be a better choice, particularly when there’s too much whitespace on your canvas. Treemaps maximize space while maintaining clarity, making them ideal for hierarchical data or complex datasets.??
However, it’s important to avoid unnecessary complexity—this is where the temptation of 3D visuals often trips designers up. 3D charts can distort data and make it harder to interpret, especially if the third dimension doesn’t add meaningful information.??
In some cases even though there is information to visualize on the third dimension, it may still be better to avoid 3D, if the chart becomes difficult to read.?
?
?
?
When designing charts, encoding through position and color should be intentional and clear. Positioning should naturally guide the viewer’s eye to the key insights, while color should be used thoughtfully to enhance understanding, not distract. For example, in the below image, we see personality traits visualized as a bubble chart vs a dot plot. With the bubble chart, the sizes indicate the value and the colors indicate the traits, however, the position is used randomly. We can exclude this randomness of position by simply visualizing the same data as a dot plot to create a more understandable visualization.?
?
?
Finally, when using color scales, consider accessibility. Color blindness affects a significant portion of the population, and poorly chosen scales can exclude some viewers from fully understanding the data. Tools and guidelines for color-blind-friendly palettes ensure that your charts remain inclusive and effective.?
?
Embrace Minimalism: Getting rid of the Chartjunk?
Selecting the chart type correctly is not the only important point to make your visuals effective for data storytelling, getting rid of the unnecessary distractions, more commonly known as Chartjunk, is also important for your narrative. Uncluttered visuals help viewers focus on what matters, so use the chart elements carefully. Let’s go through some important points with examples again.?
?
The first tip to get rid of Chartjunk is to?avoid using unnecessary colors.?Unnecessary colors create a distraction for the audience and distract your story from the meaningful insights you wish to present.?
The same tip applies to using color scales as well. Color scales can be effective in giving certain insights when used correctly. There are many popular available color scales, such as the rainbow color scale. It catches the eye, looks pretty, and is given by default color scale on many visualization libraries, but it creates a great distraction. Please take a look at the bar charts below, which show US states' population growth. The color used in the first chart gives us no information, while the second chart uses color to visualize the geographical categorization. Even though the chart types are identical, it is clear to see that the second chart is more mindful of the use of color scales compared to the first one.?
?
?
Source: Claus O. Wilke, O’Reilly Media, 2019?
Another tip is to remove unnecessary grid lines?within your charts. A grid line can?become unnecessary if it overshadows your data lines. Using no grid lines could also hurt your story if your data values become difficult to read due to a?lack of grid lines. Use grid lines to support your visuals, not to overcrowd them.?
Use direct labeling whenever you can to make your story easier to follow. In the example below, we can observe the data lines and their labels easily on the second chart, whereas on the first chart, it requires glimpses back and forth.?
Clear Takeaways: Highlighting the Key Insights?
The final point for improving your visualizations is to highlight important takeaways if you can. The human mind can focus on a certain amount of information at once.?Presenting too much information in one chart could overwhelm your audience?and?lead to distractions.?
Remember our survey bar chart from above, let’s assume we want our audience to focus on one of the bars, as the highest percentage value. If we want to ensure our audience is paying attention to the point we wish to make, we can use color to highlight this information.??
We can also use the same tip for the population growth of the US states chart above. To present the highest and lowest population growth of southern states, Texas and Louisiana, in this case, we can apply lower opacity to the rest of the data points and highlight the two states to make sure our audience focuses on the insight we want to present to them.?
Conclusion: Crafting Visuals with Purpose?
The key takeaway? Just because you can visualize data in a certain way doesn’t mean you should. Every design choice, from the type of chart to the colors and dimensions, should enhance the clarity of your story and reduce the cognitive load for your audience.?
At Machine Learning Reply, we guide and support all our customers in creating better data stories for their use cases. Providing a wide portfolio to ensure our customers’ needs, whether in Business Intelligence reporting, dashboarding, or monitoring, is a priority.?
?