Episode 31: How to Decide If Your Visualization Should Be 3D - Featured Data Visualization by Ryan Baumann

Welcome to episode 31 of Data Viz Today. When is it beneficial to visualize your data in 3D? We know that people love the “cool” factor of 3D, but I think most people know to avoid it now. But in which cases could it be useful in understanding your data? In this episode, host Alli Torban explores the pros and cons of 3D viz. Featured data visualization project by Ryan Baumann shows a beautiful way to make custom 3D visualizations of spatial athletic data.

Listen on Apple Podcasts, Google Play, Google PodcastsStitcher, SoundCloud & Spotify.

  • Welcome! I'm Alli Torban.

  • 00:28 - Today’s episode is about how to decide if your visualization should be in 3D. When does 3D benefit your viewer?

  • 01:00 - In this episode, we’ll talk about the visualization that sparked my curiosity about the third dimension, how it was built, and what to consider when you’re tempted to make your dataset 3D.

  • 01:14 - Today’s featured data viz project is a book called “Athlete Data Viz” by Ryan Baumann. He’s a former pro-cyclist and an engineer at Mapbox where he collaborates with customers to develop new products. His side project is a data visualization tool for athletes called Athlete Data Viz, which is a website that Ryan created where anyone can visualize their athletics or training in a custom way on a map.

  • 03:03 - So how did Ryan get into this cool little side hustle? As an athlete, he’s always loved tracking and visualizing his performance on his bike rides. One day in 2014, a post on the Strava subreddit caught his eye… someone named Kevin was trying to build a heatmap of his Dad’s rides for Father’s Day. Ryan loved this idea and built out a proof of concept of how it could be done and reached out to Kevin to ask if he could help him.

  • 03:45 - How? He uses the Strava API along with Mapbox data and custom styles to gather the data for the app. A user logs into their Strava account and can pull their latest ride to visualize over different map styles. Listen for more of the technical aspects explained!

  • 05:30 - He really wanted to make something that felt more like art than business intelligence, and give people the joy of commemorating a great race or experience by creating something beautiful and unique.

  • 05:45 - Tools used: Ryan used Mapbox GL JS, javascript, Postgres + PostGIS extension, Python Flask for the backend API, and Heroku for the platform.

  • 06:08 - If you want to try this out and make a beautiful map of your athletic data, you can head over to athletedataviz.com.

  • 06:16 - I liked Ryan’s idea of making paths 3D especially when the height of the path is determined by elevation. That feels like a natural and helpful use of 3D to me.

  • 06:30 - But then I began to wonder what is it exactly that makes 3D visualizations a bad choice in most situations?

  • 06:50 - I’m currently reading Tamara Munzner’s Visualization Analysis and Design and she has a section is called “No Unjustified 3D”, which I found refreshing because it wasn’t just “Never 3D”, but a carefully laid out section about the pros and cons. I like the idea of keeping 3D in my toolbox as long as I also have the knowledge of its strengths and limitations.

  • 07:20 - My two big takeaways from Tamara’s analysis is that 3 dimensional visualizations

    • 1 - have the problem of occlusion and

    • 2 - have the problem of perspective distortion.

  • 07:30 - For #1, occlusion means to obstruct, so when you show data in 3D, some data might be obstructed by other data, which can be helped by making the visualization interactive, but this comes at the cost of the viewer’s time and cognitive load, since the viewer is having to construct a mental model by remembering past views, which is especially hard with abstract data.

  • 08:00 - And for #2, the perspective distortion problem, when you plot data in 3 dimensions, the objects that are further away appear smaller, which is a problem when you’re trying to compare the size or length of objects.

  • 08:15 - So you want to keep your data in 2D when relative position is important - when you need to precisely judge the distance or angle between objects.

  • 08:25 - A time when 3D is easier to justify is when the viewer’s task involves shape understanding of inherently 3D structures, which is usually spatial data.

  • 08:45 - This made me curious about what people are using 3D models for in the real world, so I asked my sisters, who are both engineers, whether they ever visualize 3 dimensional data, and one said she does because she runs simulations of air or water flow around an object so if she graphed it in 2D, she’d miss out on what’s going on on the other side, especially when the object isn’t symmetrical.

  • 09:15 - It was a similar case for my other sister who said she uses 3D models to see how atoms interact with magnetic fields and lasers, so seeing the full picture in all 3 dimensions is essential.

  • 09:27 - This is actually similar to the example that Tamara used in her book to show a well-justified use of 3D. She showed a visualization of streamlines of fluid flow through a volume. In these cases, you’re trying to understand the shape of data in a 3D space.

Example of 3D visualization of abstract data that works well. From Tamara Munzner’s book Visualization Analysis & Design.

  • 09:44 - When you try to apply 3D to abstract data though, it gets a lot trickier to justify.

  • 09:55 - Tamara had one great example in her book where she shows a visualization of voltage readings over time, but in the 2D view, all the readings are graphed over each other so it looks like all the readings are about the same, but this interactive viz allows you to kind of pull it out like a drawer and expand all the readings out like an accordian, revealing this 3D view that shows that half the readings actually look different than in the 2D view, but since so many readings were on top of each other, the fact that the shape of the readings was changing over time was obstructed.

  • 10:35 - The important benefit of 3D was the shape of the data as time progressed. You could see how the data changed in a way that wasn’t clear in the 2D view.

  • 11:02 - Cons of 3D:

      • 1 - When you extrude objects, you might block some data behind it - called occlusion.

      • 2- You introduce perspective distortion and it becomes difficult to compare lengths and angles accurately.

  • 11:17 - Pros of 3D:

      • 1 - You can quickly understand shape.

      • 2 - Helpful when the viewer needs to understand the 3D geometric structure of objects or a scene.

  • 12:00 - If you have a use case for extruding buildings or other shapes on a map, I actually just made a series of short video tutorials for Mapbox on how to do this.

  • 12:15 - My final takeaway is that I think Tamara Munzner summed it up perfectly in her book by naming the section “No unjustified 3D”. Not never 3D, but take note its limitations and keep it in your toolbox.

  • 12:35 - Finally, I asked Ryan what’s his advice to designers just starting out, and he said as a beginner using color in your visualizations, it’s too easy to tell a story you don’t intend or that even misrepresents your data. So keep your colors simple - start with well known designer palettes for data visualizations that you can find in tools like colorbrewer.

  • 13:00 - You can follow him on twitter @RyanBaumann and head over to athletedataviz.com to make your own visualization of your training data!

  • 13:12 - I’m finishing up another course, this one is on creating and editing visualizations in Adobe Illustrator, so if that’s something that you’re interested in learning, you can sign up for my newsletter to get notified of its launch and maybe a coupon code too… :)


Allison Torban3d, baumann