A UX Researcher's Workflow for Cross-Session XR Analysis
This page is written for the UX researcher or analyst who owns the weekly XR readout. Not the SDK, not the AI internals — the workflow. If your job is to turn a week of headset sessions into a defensible recommendation for the next sprint, this is what the tool looks like from your seat.
Segmentation you can actually defend
Cohorts in an XR product are messier than in web analytics: same user, different headset, different room, different play posture. The analysis view lets you segment by build version, headset model, controller vs hand-tracking, standing vs seated, first session vs returning, and by custom tags your product team ships in the SDK config. Every chart carries its cohort definition inline so a stakeholder reading a screenshot can see exactly what was and wasn't included.
Before/after, not just after
Most XR analytics tools show you one release. The comparison view lets you pin a baseline build and diff every downstream release against it — completion, hesitation, comfort markers and spatial hotspots — with the delta shown in the same units. When you argue that the v2.3 tutorial rewrite reduced hesitation at the airlock, the number is already in the format your PM needs.
Reports that survive being emailed
A shareable report in this workspace is a versioned artifact — locked cohort definition, locked date range, exported to PDF for leadership or to CSV for the data team, with the underlying query definition attached so a colleague can rerun it three months from now without guessing which filter you had selected. This is the part of 'analyze XR behavior' that turns a screenshot culture into an institutional-memory culture.
When to hand the question back to the AI
Not every question is worth a manual analysis. If you find yourself running the same cohort comparison every Monday, promote it to a saved view; if the answer is 'something changed but I don't know what', hand it to the AI decision layer at /ai-xr-analytics. This page is about the workflow you drive; the AI page is about the workflow that drives itself.
What you get
- Cohorts by headset, controller mode, posture, tag, build
- Baseline-vs-release diffs in the units your PM already uses
- Cohort and filter definitions attached to every export
- PDF for leadership, CSV for the data team, saved views for repeats
- Locked, reproducible reports — screenshots are not a system
This page is the researcher's workflow. /track-xr-user-behavior is the capture layer that fills the warehouse; /ai-xr-analytics is the automated interpretation layer that surfaces what to look at first. Analyze XR Behavior is what you do in the middle.