Ventilation timing
the short version
From isolated reports to a learning loop
Problem
Incidents end and the lessons evaporate within hours. After-action reviews, when they happen at all, get filed and forgotten.
Approach
Five connected surfaces spanning hours to years: on-scene hot wash, formal review, searchable library, trend analytics, training loop.
System
A calm, light-mode counterpart to the live incident dashboard — built for reflection and pattern recognition, not split-second decisions.
Outcome
Every keystroke, voice note, and timeline event becomes a tagged data point that compounds into institutional memory.
The knowledge that leaves with the rig
When a fire department returns to quarters, the most valuable thing it carries back is rarely recorded. Memory of an incident degrades sharply within the first 24 hours, especially under stress. The decisions an Incident Commander made, the moments that nearly went wrong, the tactic that worked better than expected — most of it dissolves before anyone writes it down.
Departments tend to do the immediate debrief reasonably well, the formal after-action review inconsistently, and long-term trend analysis almost never. The result is that the same mistakes recur across crews and years, and good practice spreads by word of mouth rather than by design. The opportunity was not another reporting form — it was a system that treats every response as a data point in a continuous learning process.
Why the lessons disappear
Problem
Reviews are treated as paperwork rather than learning. Findings sit in filing cabinets, detached from the timeline data that explains them, untagged and unsearchable. Nobody can tell whether a new training rollout actually changed anything six months later.
Brief
Build the reflective companion to the live command dashboard. Capture the debrief while memory is fresh, structure the formal review around timeline replay, and route every finding into a library that powers trend analysis and closes the loop back into training.
Reframe
Stop designing a report. Design a memory the whole department can search — one that proves whether its own lessons are working.
Five surfaces, five time horizons
- I
Hot wash
On scene, within the hour. A blameless, voice-first capture of what we expected, what happened, and what needs work — before crews disperse and memory fades.
- II
AAR workspace
Within two weeks. A collaborative review built around timeline replay, where evidence is linked to the exact moment and findings are categorized as Sustain, Improve, or Investigate.
- III
Library
Permanent. Every finding becomes a tagged, searchable lesson — by incident type, building construction, tactical phase, root cause, and outcome.
- IV
Trends
Monthly and quarterly. Aggregate patterns surface across many incidents, and an intervention-effectiveness view tests whether changes actually moved the numbers.
- V
Training loop
Continuous. Findings route into SOP revisions and drills, tracked back to the incidents that prompted them, completing the cycle.
Calm, not urgent
Where the live dashboard is dark and built for two-second glances, this is light and built for an hour of focused reading. If it feels urgent, it is wrong.
Tagged for the future
Every insight carries metadata from the moment it is captured, so patterns can be found across years of incidents that no single person would ever connect.
what shaped the design
Learning from those who run toward it
- Ride-alongs
- IC interviews
- NIOSH LODD reports
- NFPA 1561 review
- Hot wash observation
- Near-miss pilots
The most dangerous moment is when the commander stops getting new information — silence usually means something is being missed.
How an incident becomes institutional memory
The review, built on replay
The commander scrubs through the incident replay; the building diagram updates to show crew positions at that exact moment, and the event list filters to what happened around that timestamp.
Findings sort into Sustain, Improve, and Investigate — the categories that decide whether a lesson becomes a drill, an SOP change, or an open question.
Flag any moment as a decision point and it appears in the findings panel, already linked to its evidence — timeline, radio audio, and hot-wash insight together.
A memory you can search
Filters on the left help find a specific lesson; the patterns sidebar on the right turns the archive into an early-warning system that flags when a root cause is trending the wrong way.
The colored edge on each card does the categorization work at a glance — green to sustain, amber to improve, red to investigate.
Every lesson sits alongside the SOPs, training modules, and external research (NIOSH, UL FSRI) that already address the same problem.
Proving the lessons work
Aggregated over many incidents, the data answers questions no single review can: are response times improving, which benchmarks consistently slip, where do near misses cluster.
A vent-coordination drill that rolled out in July and made the numbers worse triggers a very different conversation than ‘we delivered the training, we are done.’
The intervention-effectiveness view compares finding rates before and after each change — the honest accountability most departments never build.
Two halves of one product family
Live ops dashboard
- Dark mode, built for glare and night
- Urgent — pulses, countdowns, Mayday red
- Glanceable in two seconds under stress
- Decision speed is the entire job
Learning system
- Light mode, read at a desk in daylight
- Calm — no alarms, no urgency
- Built for an hour of focused reflection
- Depth and pattern recognition are the job
The tagging taxonomy
Incident typeResidential, commercial, industrial, wildland interface, hazmat, or vehicle — the broadest filter for finding comparable events.
Building constructionType I through Type V, from fire-resistive to lightweight wood-frame, which shapes nearly every tactical decision.
Tactical phaseSize-up, attack, search, ventilation, overhaul, or RIT activation — locating exactly where in the operation a finding belongs.
Root causeIndividual skill, crew procedure, doctrine, equipment, communication, or systemic — the bucket that routes the intervention.
Outcome severityNo issue, near miss, injury, or line-of-duty death — weighting how urgently a pattern demands attention.
InterventionThe SOP revision, drill, or doctrine change a finding produced — tracked back to the incidents that prompted it.
sustain · improve · investigate
The finding matrix
Sustain
What the department got right and must keep doing — the practices worth protecting as crews and leadership turn over.
Improve
Investigate
Air-consumption spike
Improve
PAR cadence
Investigate
Smoke change at 22:14
Improve
Radio discipline
Investigate
Floor-1 visibility loss
Improve
Hose stretch timing
Investigate
Water supply delay
anonymous near-miss reporting
The load-bearing wall: psychological safety
Most learning systems fail not on features but on culture. Firefighters will not report things that could embarrass them or their crew, and near misses — the precursors to real injuries — are exactly the data that goes unreported. The whole system rests on whether people trust it enough to be honest.
So the near-miss flow was designed to be genuinely anonymous, voice-first, and finishable in about sixty seconds. It is the most visually prominent action in the member view, because making it easy and visible is what drives the volume of reports that makes everything downstream possible.
None of the analytics matter without it. A blameless review process, leadership that models admitting its own mistakes, and visible closing of the loop — when someone raises an issue and later sees it addressed — are the conditions under which the technical system actually gets used. Design and culture had to be built together.
What carried the design
01
Calm
Inverting every signal from the live dashboard — light not dark, slow not urgent — so the surface itself invites reflection.
02
Tagged
Capturing structured metadata at the moment of insight, so patterns emerge across years no individual could connect.
03
Anonymous
Protecting the reporter so the precursor data — near misses — actually surfaces instead of staying hidden.
What this design did not solve
Gap
Adoption is a cultural problem a UI cannot fix alone. Without leadership protecting time for reviews and modeling blameless behavior, the best workspace in the world stays empty.
Open question
Integration debt is real. CAD, SCBA telemetry, and AVL each remove manual entry but add complexity, and small volunteer departments may never have those feeds at all.
The honest measure is not whether lessons are captured, but whether the next incident goes differently.
What the work taught me
Safety-critical design is mostly about restraint.
The temptation in a serious domain is to make everything feel serious. The opposite was true: the gravity of firefighter safety is honored by calm and clarity, not by alarm.
The interface is the easy half.
A learning system is a cultural artifact wearing a software costume. The screens were the tractable part; designing for the trust that makes them usable was the real work.
How it came together
Phase
Discovery
Ride-alongs, IC interviews, and review of NIOSH and NFPA material to map the real after-action process.
Phase
Define
Framing the five surfaces and the metadata taxonomy that ties them together.
Phase
Design
Wireframes and flows for all seven core screens, desktop and mobile.
Phase
Pilot
A controlled rollout with one department, one shift, heavy observation.
Phase
Iterate
Tuning the hot-wash flow and near-miss reporting based on what crews actually used.