Adoption is compoundingWeekly sessions jumped +62% after the Feb 19 training mandate and have held. Utilization is no longer the bottleneck — content depth is.
Content v1.3.0 lifted pass ratesThe Mar 21 content update (clearer extinguisher rack labeling) lifted certification pass rate by ~5 pts vs v1.2.x — measurable within two weeks via the SDK's version metadata.
Onboarding drop-off~8% of learners don't finish the tutorial walkthrough on the first try, concentrated at Grandview Hotel. Recommend a facilitator for first sessions there.
Weekly training sessions
Headset sessions per week · dashed lines mark the mandate email (Feb 19) and content v1.3.0 release (Mar 21)
Certification pass rate by week
Share of certification attempts scoring ≥ 80 · trend crosses the v1.3.0 release
Program funnel
Enrolled → tutorial → practice → certification attempt → certified
Pass rate by content version
Version-over-version comparison from super-metadata on every event
Every layer of the Insights SDK feeds this report
Data streams captured by FireSafe VR's deep integration
AssessmentsObjectivesInteractionsScenario levelsExperiencesCustom eventsCritical safety eventsSpatial dataTelemetryTimed eventsLogsIn-headset pollsCross-session storageDevice tagsPer-user identity
Remediation worksLearners who failed their first attempt improve by +19 points on average after targeted practice — the biggest gain of any cohort.
EMS trainees set the ceilingEMS / volunteer-firefighter trainees average ~94 with a 96% first-attempt pass rate — a useful benchmark for what "expert" looks like.
Hints predict struggleScores fall sharply up to ~3 hints per attempt, then plateau. Learners requesting 3+ hints are strong candidates for proactive coaching before their certification attempt.
Certification score distribution
Best score per learner · pass line at 80
Cohort outcomes
First-attempt pass vs eventual certification rate
Pass rate by attempt
Retakes convert — the program recovers most first-attempt failures
Hints used vs score
Each point is one certification attempt
Score gain, attempt 1 → 2
Average improvement among retakers, by cohort
Learner roster
Individual records via per-user authentication — search, sort, drill down (names are synthetic)
PASS technique is the weak skillOnly ~68% pass the Pull-Aim-Squeeze-Sweep objective (threshold 75). Sweep motion and step sequencing drive most failures — worth a dedicated micro-module.
"Know when to run" gap41% initially chose to fight the oversized warehouse fire and 33% engaged the ambiguous-smoke hallway instead of evacuating. Judgment scenarios need reinforcement, not just technique.
Grease fires confuse extinguisher choiceThe single most-missed interaction is selecting the wet-chemical extinguisher for the breakroom grease fire — 34% initially grab water or ABC.
Performance by certification objective
Five weighted objectives from the assessment tree · avg score and pass rate
Scenario difficulty
Avg level score and % making the correct fight-vs-evacuate call
Most-missed interactions
First-try failure rate across graded interactions
Median time to identify fire class
Timed-event data (time_to_fire_class_decision), by site
Violations are fallingSafety violations per 100 sessions dropped ~52% from the first month to the last — the clearest evidence the training changes behavior.
Server room drives wrong-agent errorsThe Class-C server room scenario accounts for nearly half of all wrong-agent violations (water/ABC on electrical). This is exactly the mistake that's catastrophic in real life — and it's being caught in VR.
Two site storiesAtlas Manufacturing cut wrong-agent violations ~60% after their safety lead reviewed weekly Insights reports with crews. Meridian Data Center spiked in week 6 when the new server-room scenario shipped, then recovered.
Safety violations per week, by site
EventCritical stream · stacked weekly counts
Violation types
90-day totals across the critical-event taxonomy
Violations by training scenario
Which situations trigger unsafe behavior
Violation rate per 100 sessions
Normalized for training volume — the leadership metric
Grandview is slower to decideMedian fire-class decision time at Grandview Hotel runs ~25% above other sites. High new-hire mix + no facilitator — recommend guided first sessions.
atlas-07 needs maintenanceOne Quest 3 at Atlas logs ~4× the frame-drop warnings of the fleet average. Auto-telemetry caught it before anyone filed a ticket.
Northstar runs a proctored program72% of Northstar Hospital certifications are proctored vs ~13% elsewhere — device tags + assessment metadata make compliance segmentation free.
Site comparison
Segmentation via ArborXR device tags — no app changes required
Sessions by headset model
Fleet mix from device tags
Device health — frame-drop warnings
Warn-level logs per 100 sessions · worst 10 devices
Device utilization
Avg sessions per headset over 90 days, by site
Payback in weeks, not yearsAt current volume the hardware investment pays back in ~5 weeks. Every learner after that is ~$194 cheaper than classroom delivery.
Time returned to operationsVR certification takes ~36 min vs a 2.5-hour classroom block — ~3,000 work hours returned per year at planned volume.
Risk reduction is the real prizeViolations caught and corrected in VR (wrong agent on electrical, failure to evacuate) are the incidents that cost six figures in real life. NFPA-average direct cost of a workplace fire incident exceeds $85k — model below assumes just one avoided every two years.
ROI model — adjust the assumptions live
Defaults reflect our program; edit any input and the model recalculates
Fully-loaded cost per certified learner
Delivery cost + wages for time away from work
Cumulative 12-month cost
VR line includes hardware up front — the crossover is the payback point
Learner confidence — before vs after
In-headset PollUser ratings (1–5), pre-training vs post-certification
"What was hardest for you?"
Post-session poll · confirms the PASS-technique skill gap independently
Learner sentiment
From in-headset exit polls
Would recommend (NPS)
—
from 1–10 recommendation poll
Rated training 👍
87%
thumbs poll, 70% response rate
"AI instructor helped"
78%
shown after ai_instructor_consulted
Avg post confidence
—
up from — pre-training