Analytics for VR Interactions
There are a handful of standard analytics tools to help creators improve their VR applications, but as the experiences become more interactive, the tools will become more fragmented and content-specific.
Right now they fall into two categories: traditional web analytics like Google Analytics and Mixpanel, and tools made specifically for VR applications. While web analytics can be useful in VR applications, we’ll focus our discussion on tools designed specifically to improve VR experiences. And we’ll see how choosing the right way to analyze a VR app becomes more difficult as the apps get more interactive.
Let’s start by considering what makes analytics hard. Analytics seeks to quantify and summarize aspects of a population (the users) by aggregating business metrics, often about how products are being used. While statistics tells you the right way to interpret aggregations of these metrics, analytics is also concerned with choosing the metrics you should look at in the first place. In VR there is a wealth of sensor data available for analysis, but it’s not always clear which data will help you understand how users are engaging with an application. This is the challenge of VR analytics.
The choice of the most appropriate analytics tool often depends on how interactive your VR experience is. When we say “interactive,” we mean that it involves “the actions or input of a user” (following this definition). In the context of VR, we’ll interpret an interaction to mean that the user is altering something in the virtual world, not simply changing their observation of it. So a user moving her head to explore a 360 photo is not interacting with the experience, even though the display updates with each movement. If she switched the photo, that would be an interaction.
360 Video
So by our definition, 360 video isn’t interactive since the user doesn’t modify the virtual world other than to press the start and stop buttons. This lack of interaction is one reason some people don’t consider 360 video a true VR experience. But these constraints are an advantage for analytics, because fewer degrees of freedom make the analysis much simpler—both in terms of implementation and interpretation.
Since users can look around, but can’t move or touch anything, we know that they each saw the same spherical image at each point in time. The only variable is where they looked on that image. Aggregating where all users are looking at a given point in time is a perfect application for heat maps. For example, in Vrtigo’s 360 video tool, a sequence of heat maps at each point in time is played back over the original video, allowing content creators to see what areas were viewed the most.
So it’s possible to compare the experience of each user to the population since they all see the same spherical image at each point in time. This type of population-level analysis, tied closely to the original asset (the video in this case), is hard to do with other forms of VR since individual users rarely take the same path through the experience.
It’s worth pointing out that while heat maps are one of the most common and easily understood tools for analyzing activity on a two-dimensional surface, they’re not the only option. As we’ve discussed in previous posts, we’ve created a number of tools to quantify how users are engaging with 360 videos, including measures of the audience’s focus and quantified hot spots.
Fully Interactive VR
At the opposite end of the spectrum from 360 videos lie fully interactive VR experiences. These use state-of-the-art technologies to bring the user as close as possible to full immersion in the virtual world. They are rendered in 3D, make use of hand controllers, and track movement at room scale. While the experiences may be superior to passive videos, one downside is that these new degrees of freedom complicate the job of the analytics tools. It can be difficult to answer simple questions like, does engagement with a tourism app look the same as engagement with a virtual roller coaster?
And even if you could agree on a set of engagement metrics for a highly interactive app, the large number of possible interactions means that no two paths through the experience are likely to be the same. Unlike 360 video where everyone shares the same timeline, here it’s possible that every user will take a different path through the virtual world. So effective analytics tools need to measure engagement across different types of experiences, relying on fewer commonalities between each user’s journey.
It’s therefore unlikely that a single, one-size-fits-all tool will solve all the analytics problems content creators have:
Rather, there will be many tools, each tailored to a specific type of experience. Most likely these tools will make use of machine learning to classify types of behavior, thus reducing the large volume of VR sensor data to categories that are easier to interpret.
Partial Interactivity
So if 360 video can be analyzed using familiar tools, and fully interactive VR requires specific solutions for different experience types, what can we say about VR experiences that are neither fully passive, nor fully interactive?
One such type is what we’re calling “rendered interactions” in the above chart. Unlike 360 video, these experiences are rendered in 3D and often allow the user to move around the space. But their interactivity rarely extends beyond navigation and pushing buttons. This type of app is common in WebVR, which is still being developed and doesn’t support full interactivity at the level of native VR apps yet.
Another type of experience that’s neither pure video nor fully interactive is what we‘re calling “branching 360 video.” These are choose-your-own-adventure style 360 videos where viewers encounter branch points between video segments, allowing them to decide what direction to take the narrative. This lends itself well to existing analytics approaches since it uses 360 video, which benefits from heat map analysis, and it produces a graph of user interactions, which benefits from path analysis. We’ll talk more about branching 360 video in the coming weeks when we release a tool for analyzing those experiences at Vrtigo.
There’s no single analytics tool appropriate for all VR applications, but there are specific types of experiences that benefit from existing approaches. Having fewer interactions makes the analytics job easier, but applying analytics to fully interactive experiences is also possible, provided the tools are specific to the type of experience. Most of these tools have yet to be invented, so the next few years will be an exciting time for analytics in VR.
Vrtigo is the next generation VR analytics platform for content creators, editors, producers, and marketers.
