Are data dashboards vanity projects?
The complexities of data visualization literacy
Researched and written with the help of M Zide PhD and Sarah Zimmers
If you own a smartphone, you probably receive a notification about your phone’s weekly usage. If you open that notification, you will see a dashboard that shows usage graphs for your apps. At work, your organization may track sales, growth, and forecast in a colorful summary with a chart or two for you to report on your performance.
When monitoring your financial spending, your banking app most likely shows you a pie chart with your monthly charges segmented into ample categories. And your HR information system most likely shows a graph-based breakdown of your paycheck with all the payments and deductions. We are so used to data being everywhere that visualizations have become a norm in our interactions with technology.
At the same time, what is often overlooked is that these charts, graphs, and other visuals can be merely decorations if users are not data literate. Based on the research on data literacy, this article offers a critique of the overabundance of data visualizations and dashboards, talks about when to use them appropriately, and sketches some guidelines for identifying when an alternative is justified.
Definitions & Concepts
To better gauge people’s understanding of data visualizations, it’s essential to define the concepts of numeracy and data literacy.
Numeracy is understanding, calculating, manipulating, interpreting results, and communicating mathematical information. It involves reasoning, critical thinking, analyzing numerical data, and making informed decisions based on quantitative information. Such seemingly everyday concepts as date, chance, patterns, relationships, dimensions, shapes, numbers, and change form the core of numeracy comprehension. Importantly, numeracy lays the groundwork for data literacy, which builds to data visualization literacy.
Data literacy is often interpreted as a sub-type of numeracy and can be summarized as the ability to read, analyze, work, and argue with data. Visualizations present many difficulties for people working with data, and proficiency in such data manipulations has been coined as data visualization literacy.
Together, numeracy, data literacy, and data visualization literacy form a spectrum in which it may be difficult to separate where one type of literacy begins and another one ends. However, without basic numeracy literacy, the individual cannot be data-literate, and data visualizations are reduced to decorative images.
The Complexity of Data
With the definitions and concepts of data visualizations explained, let’s now explore the challenging aspects of working with data, which can be summarized as follows:
- Data is often messy and disorganized. Data comes from various sources and requires additional cleaning, verification, and organization steps. These steps require at least some critical thinking and an ability to sort, filter, and standardize the data.
- Data can be biased or incomplete. Humans are data’s primary creators and inspectors, so its results may be inaccurate or missing values.
- Data can be challenging to understand. Data visualizations, statistical measurements, tables, and spreadsheets all represent a variety of formats. Different formats influence the likelihood of arriving at a particular conclusion, the correct interpretation of which requires a high degree of critical thinking.
- Data can be challenging to communicate. More often than not, presenters rely on the visuals of data rather than describing its meaning and implications.
Reliance on data presentations, such as visualization, is usually based on the assumption that they are more intuitive. However, research shows that not all data visualizations are comprehensible. In one data visualization familiarity study, the type of visualization affected the comprehension of information: after much experimentation, it was concluded that both youth and adults are capable of interpreting only the most basic reference systems, such as charts and graphs (e.g., pie charts, bar graphs, and scatter plots). Maps and network layouts were demonstrated to be the least comprehensible. In another study, significantly more participants could correctly identify a value on a chart than make inferences between two points. An additional study found that even fewer people could locate the trends demonstrated on a line graph, suggesting such graphs may pose a more significant challenge depending on the audience.
Overall, different visualization types require different amounts of cognitive capacity. Bar graphs are associated with lower cognitive effort for comprehension, whereas network graphs, generally infrequent, can be too convoluted, requiring users to observe too many relations to understand. The table below provides examples of data visualization types and summarizes their use, ordered by ease of comprehension.
While data visualization may initially seem effective for conveying a point, the inherent complexity of data and the cognitive burden of data visualizations prevent users from complete comprehension. Beyond UI impact, we see low data literacy tendencies around the world, which on their own significantly impact businesses and economies.
Data Literacy Trends
Numeracy, fundamental to data literacy, shows many challenges in proficiency and acquisition, including for many people in the US. It is estimated that less than 10% of the general American population aged 16–65 are proficient in understanding numbers. On average, Americans’ numeracy proficiency is rated at a level 2 out of 5, which includes performing basic calculations and interpreting simple tables and graphs. The US ranks at the bottom of the list compared to other developed world countries. Because numeracy and data literacy are interconnected, researchers observe that low numeracy rates also spawn low data literacy rates influencing many professional domains.
Although data has become the new gold, and any white-collar job expects significant data literacy, few people feel comfortable working with data. A recent study found that only 24% of the US workforce feels confident in their ability to read, work with, analyze, and argue with data. This lack of data literacy costs businesses millions of dollars in lost productivity and efficiency, yet many businesses do not provide adequate training in data literacy. Employees may also lack the time or resources to learn about data literacy independently. Meanwhile, higher data visual literacy positively correlates with making correct inferences from unfamiliar data visualizations. In other words, in this gold rush of data, we send people mining without providing them with the proper tools. As a result, data literacy remains the main challenge in data management.
As data literacy thwarts the democratization of the data in the workspace, the real challenge remains to teach people the required data skills. The problems with literacy begin in kindergarten, where students are rarely taught how to read or create data visualizations. Yet, time and time again, we observe high reliance on data dashboards across B2B platforms, specifically visual data. However, such dashboards are often ineffective in conveying clear, actionable insight, nor are they easy to use. Most of the time, while they attempt to simplify some decision-making for the user, they put an additional burden on them by requiring them to interact with the data to know how to decipher it.
Outlook & Recommendations
Given the hurdles data literacy can present to their usefulness, dashboards tend to be vanity projects. Furthermore, we question the worth of most data visualization pages, as they are extremely costly to develop but usually earn low engagement from users (check out this helpful calculation of ROI per feature). One UX principle is that user engagement should not come at the expense of usability. Another pillar of product design is that user satisfaction derives from efficient task completion rather than engagement duration. Data visualizations and dashboards tend to contradict both of these principles by foregrounding elements that demand engagement but present poor usability hindering the platform’s efficiency, and generally aren’t usable to most of the population. To recover from this maldesign, we recommend taking a step back to assess the reasons for incorporating a costly dashboard.
Before investing in building data visualizations or dashboards solely because they seem necessary to remain competitive or because that’s what other B2B and B2C products are doing, it’s crucial to consider your users and their actual needs. While certain user profiles, such as data scientists, field scientists, researchers, and medical professionals, possess advanced data literacy and desire the freedom to explore various data visualizations, most B2B and B2C users lack such proficiency. For the general audience, a higher return on investment would be exploring automation and AI technologies to perhaps bridge the data literacy gap by providing actionable contextual insights throughout the platform. Such an approach capitalizes on the technology of your product and invests in better user engagement with potentially lesser cost and without relying on users’ literacy.
You might be wondering when visualizations are okay. Visualizations are great for modeling known relationships or hypothetical outcomes between multiple or mixed variables or large quantities because they help users summarize these relationships and identify trends. Data dashboards take these visualizations further by making them interactive, allowing users to swap out some variables for others, filter data points, and explore different relationships. When a dashboard is required, consult the plethora of research on better visual storytelling that exists to this day (for example, this post from IDEO U on how to tell a story based on your visual data). However, overall, there is little justification for having dashboards for people who may not be highly data literate.
It’s essential to prioritize solving users’ problems efficiently and providing them with actionable insights. Rather than assuming that dashboards are universally beneficial, take the time to understand your users and their objectives. Building dashboards should be driven by genuine user needs and literacy levels on the one hand and the actual data types on the other. After all, UI is simply a tool a user can use to accomplish a task, and since most do not know how to use data visualizations, a dashboard might be a tool best left in the shed.
Researched and written by Irina Wagner, PhD, M Zide PhD, and Sarah Zimmers.
