October 2024 – December 2024
Overview
This was a Meta Quest 2 prototype for exploring CSV data in a 3D scatter plot. Our team of three built it in one semester.
My role
I owned the spatial interface and interaction design. I focused on the data controls and the point tooltips people used inside the scene.
Question
We wanted to explore whether moving around a dataset in VR could help people understand relationships between its values.
The challenge was to support that exploration without overwhelming controls or an obstructed view.
The constraint
With a Quest 2 and one semester, we focused on a small prototype that people could use with their own CSV data.
Framing the question
Original project documentation with the target users, how-might-we questions, and problem statement.
Scope
We prioritized importing a CSV, plotting it across three axes, and selecting points to inspect their values.
People had asked for custom chart styles, but we left them out so we could finish and test the core interaction within the semester.
Original documentation listing the needs the first version was designed around.
What made the first version
CSV import, three axes, and point inspection were built; custom chart styles were deferred.
Interaction decisions
I designed the controls people used inside the scene. Three decisions shaped the interface.
I grouped the scene controls into a few panels near the hands, leaving the plot in front of the user.
Controls within reach
This illustration places the control panel near the hands while keeping the scatter plot in view.
The panel brings CSV upload, rotation, and scale into one place. Sliders adjust the plot inside the scene.
In this illustration of the prototype, rotation and scale sliders change the plot while the controls remain in view.
Selecting a point reveals its three axis values in a separate tooltip, keeping the detail close to the data.
Iteration
Early layouts crowded the view with information close to the user’s face. I reduced the interface to a few panels: dataset controls with rotation and scale, and a tooltip for the selected point.
Early layout
Information placed close to the user’s face crowded the view of the plot.
Reduced layout
A few panels near the hands leave the scatter plot in view.
These views reconstruct the prototype’s layout changes; the early arrangement is schematic. Comfort and understanding were not measured.
Outcome
We completed a working Quest 2 prototype and gathered informal feedback from people familiar with Excel and Tableau. Feedback was positive about the panels and point tooltips. We did not measure analysis speed or accuracy.
I would ask participants to complete the same analysis tasks in the prototype and their usual desktop tool, then compare errors, completion time, and comfort.
Scope to what you can test.
A smaller first version gave us something concrete to discuss. The next study should use explicit analysis tasks to guide a decision about further development.
In VR, the view is part of the interface.
Designing for VR made panel position part of the layout problem. I would evaluate reach, view obstruction, and comfort together in the next iteration.
Product demo
The original walkthrough shows CSV data plotted in VR, scene controls, and point inspection.
Next project
Rebuilding VI’s Course Pipeline ↗