The Data Daily

5 Lessons on Data Visualization From a Data Physicalization Failure

Last updated: 01-29-2020

Read original article here

5 Lessons on Data Visualization From a Data Physicalization Failure

Just like an author must tailor the content and vocabulary of their work to match the group of people they want to reach, those who make data visualizations should do the same. When working on our physicalization, my group approached the data with other college students in mind — knowing that our classmates would be observing our final product. As a result, we wrote our placards at a collegiate reading level and expected our audience to be capable of noticing a high level of detail. However, we failed to take into account the confusion that resulted due to the novel form we created, and presented our information without a key to guide the user.

Think about what information the intended audience has already experienced: do they know what the labels on the axes mean? Would it be helpful to include definitions for words or symbols? What kinds of information do they interact with regularly and how is that presented? Putting new information in a recognizable form can put the audience at ease and allow them to interact with the data more deeply.

The physicalization I helped design was installed on the Data Science floor of the library at my college, which was accessible only to students affiliated with the Data Science Initiative. Because of these restrictions, only a limited number of people were capable of viewing our physicalization. This might seem like an extreme circumstance, but many factors can impact who can and cannot view data.

Has it been specifically shared with groups or organizations who hold a stake in the issue? For example, if the data visualization is about Virginia taxes, it may be wise to notify local news sources and legislators. It’s also important to think beyond discovery of the viz itself. Users with visual impairments like blindness, partial blindness, or color blindness may have trouble viewing some visualizations. Try testing visualizations for color blindness compatability using tools like Coblis or ColorBrewer and ensure that the accompanying alt-text for online visualizations is accurate and descriptive. (This could easily be a whole article unto itself, but for a good starting point you can refer tothis Storytelling with Data blogpostorthese guidelinesfrom theU.S. General Services Administration.)

As previously stated, my group’s project was put up in a space designated specifically for data science students, so we knew that anyone who approached our installation would be expecting some kind of data-based story and conclusion. Had the project been installed in some other space — like a public library or a children’s science museum — our viewers might approach the project with very different expectations.

Imagine the environment of the data visualization from the perspective of the viewer to help discern the tone that the visualization should use. If the visualization is accompanied by a very serious text, then it should use more serious language and refined imagery. On the other hand, if the visualization is placed in a lighthearted atmosphere, it might work well to play with color, fonts, and language.

It’s easy to imagine what kinds of materials would be available to create a physicalization. My project was built of wood, glue, motors, wires, motherboards, and a few other odds and ends. But to enhance our piece and make it more understandable to viewers, we probably should have used some paint for color coding and more labels so that people would better understand what they were looking at.

Materials in the physical world are intuitive, but it’s more difficult to think about the “materials” used to build visualizations in the digital or print realms. Instead of considering objects, try to think of materials as the tools employed: colors, fonts, icons, symbols, or even the layout of the visualization. If your intended audience usually associates the color green with environmental issues, then it probably wouldn’t be wise to make a visualization about technological advances with a primarily green color scheme. By the same token, consider familiar forms (like bar graphs, pie charts, and Venn diagrams) as platforms to easily communicate with your audience if they need information quickly.

Even if everything had worked mechanically with our physicalization, it still would have had some issues conveying the information successfully. We realized after building it that viewers needed to be able to see the piece both from the top and the sides to draw parallels between power production and consumption. Our physicalization sat on the floor and the tallest tower was only about two feet tall — far too low to expect someone to be able to observe it from the side.

Though visual obstruction of the data is a concern with data physicalization, data can be obscured in other ways in visualizations. Think about any data in the dataset that’s not included in the visualization: does it change the story being told? Similarly, be careful with transforming data, such as with a logarithmic scale. Some viewers may not understand it’s implications or even realize that the data has undergone a mathematical process.

Read the rest of this article here