How to Create Compelling Visualization?

Gerren Keith-Davis 28 Feb, 2023 • 7 min read

Introduction

Visualizing data is both an art form and a science. Some books provide their best case on creating a compelling narrative for what makes visualization appealing. Still, these texts may fall short since oftentimes; the research is based on survey data (which does not always reflect truth). The science behind most of these texts is qualitative in nature, and that form of analysis may not be applicable across a spectrum. In addition, you may also find that you’re being asked to deliver for a project with vague direction.

With express candor, I can assure you that there are no universal rules to developing a compelling visualization, and you will often be thrust into projects with little direction. If there are no universal rules, then how does a visualization become compelling?

“Compelling” is a story that is told by the use case for visualization. Though there are no universal rules to a compelling visualization, there are ways to ensure your development is useful.

Compelling visualizations typically include the following:

  1. Intuitive and simple
  2. Firm science and math
  3. Stakeholder approval

This article was published as a part of the Data Science Blogathon.

Table of Contents

Intuitive and Simple

First and foremost, the purpose of any visualization is to tell a story (and visualization happens in a number of ways). So with that in mind, do not limit yourself to strictly bar graphs and line graphs. There are times when simple text will convincingly tell the story. More specifically, your intuitive visualization should be able to convey a message at a glance. Meaning that, without words and within 10 seconds, someone should be able to understand the meaning of your visualization. Implicitly, your visualization must be simple enough to understand.

For instance, suppose you are trying to visualize your project’s Annual Revenue for 2010–2022. It would be very easy to deliver a spreadsheet with two columns for a year and annual revenue) and that would [technically] tell the story of the data. However, from tabular data, it is unclear if there is any trend or if there was any year that may be abnormal. It would be better to deliver a bar graph or scatter plot with an explicit trend line while ensuring that your numeric values are represented in your project’s currency. In this case, the visualization will not only display the annual revenue for your project, but it is also simple enough that anyone would be able to understand the meaning.

Do this

"Visualization

Instead of this

"Visualization

Though the above example is trivial, you can imagine which one would be more appealing to tell the story of the data. Also, note it may be easy to notice a steady increase in annual revenue from the table shown, but in your actual projects, datasets may not be this small. For example, you may find yourself working on visualizing daily sales of your favorite snack food for the previous three years.

Lastly, do your best to avoid using overly technical terms in any visual, trust me. Your job is never to reveal your data’s r-squared value but rather to explain what it means “in English” or your given language. If you want to explain your data’s high variance, a numeric value is not helpful; instead, you can show graphically that it is hard to find any trends in the data because the data points are too scattered or dispersed.

Here is an added benefit of intuitive and simple visualizations, it makes people want to look more. When your work is easy to understand, your audience will not feel overwhelmed or confused. As a visualization expert, you should be able to explain your work without using too many academic terms. In the United States, there is a saying when someone is speaking in terms, only experts would know, “Tell me in English.” Essentially, this means the expert, for all their hard work to gain knowledge, has no ability to communicate with the people they are trying to help. Keep your visualizations intuitive and simple so as not to overwhelm your audience, and when they want to understand more, be ready to speak simply about the findings and insights you have worked hard to generate.

Firm Science and Math

Data visualization is important, but one of the unsung heroes (the component that may be overlooked) is actually the analysis that allows for visualizations. It is similar to going to a concert to see your favorite musician without considering all the people it took to build the stage the musician stands on. Without the proper science or math (depending on your project), your visualization may compel people to ignore your work. The role of the science and analysis may not be yours, but as someone responsible for creating the face of your data, you have to have an eye for when data quality may be lacking. As such, you must review the data you are actually trying to visualize! This is not quite Exploratory Data Analysis, but it is more of a peer review. Remember, you’re working on a team to deliver a product, and if you can catch a mistake early, you save time and energy for many involved.

It is not enough to view the data, though; the analysis is also imperative so that you can properly build your visualizations. In the example above, where you were asked to visualize annual revenue, how do you know what type of trend line to use? There are various types of trend lines [including but not limited to], linear, polynomial, and exponential. If you select a polynomial trend line, how do you know to what degree your polynomial equation should stop? You may have also heard this described as a degree of freedom.

In the above example, we used a moving average with a trailing period of two, but there could have been any number of trend lines attached; however, what makes sense for your project is solely up to you… at least during the development process (keep in mind that you are the technical expert, so these types of decisions typically rest with you however, always leave room for input from peers). Below are various types of trend lines that could be applied to the same graph.

Simple Linear Trend Line

"Visualization

Polynomial Degree 10 Trend Line

annual revenue

Exponential Trend Line

annual revenue

This example illustrates the importance of the science and math behind visualizations and is not as simple as dragging and dropping elements. Given the example, if you were to submit a graph with an exponential trend line, you would be telling an incomplete story. The trend line does not capture all the data elements.

Along with the science and math behind the visualizations, sometimes, you have to include data in separate files that helps to explain why the data is the way it is. For example, in the given an example, we see that for the year 2022, the business did not do very well. What factors contributed to this? Was there anything outside of normal business operations that could have created such a downturn? Were there any insurmountable external factors? For this reason, it is always helpful to include a breakdown of your data, and this can be accomplished in any way, but it is imperative that there be some rationale behind the data. Essentially, you want to allow the data breakdown to answer questions for you to prevent a revolving door of Q&A between you as the developer and any member of leadership (if someone has to continue to come to you for questions on your visualizations, you are not sufficiently visualizing data).

Stakeholder Approval

Ultimately, to make your visualization compelling, you have to get the approval of the project lead. I mentioned earlier that technical decisions are yours… during development. The truth is, ultimately, you have to earn the approval of your project lead. The project lead typically reports updates and results (or you may get to if you are not shy of presentations) to their lead; that process can repeat itself to the head of your organization or your investor.

You should aim to create visualizations that are beyond your wildest dreams the first time, though in reality; there is a lengthy process of development and review before acceptance, so be prepared to continue developing your visualizations.

Advantages of these tips:

These tips I have listed may seem basic. However, when these simple tips are followed, it can lead to opportunities you never thought possible. In some cases, if your work is great, you can present to many different teams in your organization, from managers to presidents; the possibilities are numerous. Presenting to leadership may seem trivial, but do not underestimate the power of getting in front of your leadership, even if it is only once. In addition to presentation opportunities, as your work improves, you will become a recognized expert (meaning you are proving your worth to your organization), which helps build a positive reputation for you and builds trust in your peers.

Conclusion

Most likely, your data will continue to evolve anyway, and as that happens, you may need to make adjustments. You may also find yourself recreating visualizations because you’ve migrated to another software as different tools increase in popularity. Basically, the development process for creating compelling visualizations is a continuous cycle.

As you continue to go through the development process, remember that your visualizations should be:

  • Intuitive and Simple: With a quick glance (and with no words), your visualization should tell a clear story, and then you should be able to explain your insights clearly (and simply).
  • Firm on Science and Math: Though you are creating data visualizations, your work is no less important than a data scientist, and your science and math need to be just as good.
  • Stakeholder Approved: You have to seek the approval of your project leaders in order to consider your work great truly; without their support, you may not succeed very much.

Though there are no universal rules, I’ve listed here a few checks you can implement to create compelling visualizations. These tips are simple, but if you’re familiar with Python, execute the command import this and see that the third line reads “Simple is better than complex” (among the many other applicable lines).

The media shown in this article is not owned by Analytics Vidhya and is used at the Author’s discretion.

Frequently Asked Questions

Lorem ipsum dolor sit amet, consectetur adipiscing elit,

Responses From Readers