Real-Time Feedback Framework for VR Training Modules
I led the feedback system for VR training. I defined what a callout is, when it can fire, and how loudly it can speak. Engineering and Art built against that model on headset, tablet, and web.
I replaced per-exercise debates with one system. Overlapping feedback went from eight to at most two. Reduced decision making time for both design and product.
Nobody owned feedback as a system
Real-time feedback is the signal a trainee gets in the seconds between an action and the system response. I joined a pile of one-off tags. Each new drill added another. Nobody owned the list.
PMs, designers, and engineers were making the same callout decisions per exercise. That was slow, and it did not compose. Live feedback and the after-action report sat on the same mess.
Eight things talking at once, none of them ranked
I audited a drill and found up to eight overlapping callouts. Three failure modes sat under the clutter: too much at once, the wrong thing at the wrong moment, and every message delivered the same way.
Live feedback and the after-action report looked like different products. Underneath, nothing decided what got shown, when, or how urgently.
I sorted the system before I designed any UI
I started with categories, not screens.
DECISION 1 · SORT BY FUNCTION
I sorted every existing tag into five functions: Instructional, Shot-Specific, System Event, Result, Mixed-Use. If you do not know what a piece of feedback is, you cannot decide when or how to show it.
DECISION 2 · PLACE IT ON THE DRILL TIMELINE
I placed each function on the drill timeline. Instructional belongs in preparation. Shot-specific and system-event belong while the dummy is active. Results belong after it disappears, when nothing else should compete.
DECISION 3 · RANK HOW LOUDLY IT SPEAKS
I found almost everything had been world-locked UI, so an instruction and a warning felt the same. I ranked five modalities, then paired them: voice-over and optional light for info and system events; HUD or attached UI for shot-specific, never both; standalone UI for results.
I named the tenets only after the pattern was real
I mapped every item against every modality it actually used. The same three questions kept showing up. Those became the tenets, not the other way around.
MINIMIZE COGNITIVE LOAD
At most two pieces of information at once. If two modalities fire on the same event, they say the same thing.
DESIGN FOR THE TRAINEE’S PERFORMANCE
If it helps in this drill, say it now. If it helps in the next drill, put it in the mini-AAR. If it helps in the next exercise, wait for the full AAR.
DRIVE CONSISTENT MENTAL MODELS
Instructional always gets the same pairing. Users should know where to look before the callout arrives.
