Field notes, engineering, privacy.
Long-form on what we're learning from pilots, what we're building, and the privacy thinking that drives the architecture.
Demand-controlled ventilation: CO₂, occupancy and the meeting-room spike
Demand-controlled ventilation explained: how CO2 monitoring and occupancy data drive ventilation to the people actually in the room, cutting HVAC energy while indoor air quality improves.
Emergency department overcrowding: measuring the waiting room without cameras on patients
Emergency department overcrowding is measured too late. How anonymous waiting-room occupancy analytics gives hospital patient flow teams live density, dwell and surge signals without a camera ever recording a patient.
The 6-star NABERS deadline: what the July 2026 government lease rules mean for building owners
From July 2026, new Commonwealth office leases require a 6.0-star NABERS Energy rating and all-electric operation. What the Net Zero in Government Operations rules mean for owners, and where occupancy data fits.
Computer Vision on the Edge, Explained
Computer vision on the edge, explained: how edge AI turns cameras into anonymous sensors for retail dwell, events occupancy monitoring and real-time alerts through an occupancy API.
Concourse Monitoring with Edge Computer Vision
Concourse monitoring and outdoor venue occupancy analytics with edge computer vision: how anonymous, on-device crowd sensing measures density, flow and dwell.
Human-Centred Design for Spaces That Work
Human-centred design shapes a building around how people use it. Here is how space user patterns turn workplace guesswork into evidence, without cameras.
Do occupancy sensors use cameras? Types, privacy and accuracy compared
Do occupancy sensors use cameras? Some do, most do not. Compare PIR, mmWave, thermal, ToF and optical sensors on privacy and accuracy, the Occivar™ way.
How to Measure Office Occupancy Without Cameras (or Badge Logs)
How to measure office occupancy without cameras or badge logs, using anonymous people counting on-device. No video stored, no identities, accurate data.
Occupancy rate vs utilisation rate: definitions, formulas and how to calculate each
Occupancy rate vs utilisation rate, with the formulas: how to calculate each, what a good office occupancy rate is, and how Occivar™ measures it anonymously.
Wi-Fi Occupancy Tracking Accuracy: Why MAC Counting Breaks
Wi-Fi occupancy tracking accuracy collapses under MAC randomisation, and the fixes are re-identification in disguise. Why device-free anonymous counting wins.
Victoria's WFH law starts 1 September 2026: what it means for your office floor
From 1 September 2026, every Victorian role that can reasonably be done from home will have the legal right to work from home two days per week. For property and workplace-experience teams, the question is no longer how many desks you have, it's which days they actually get used. Here is what the law changes, what stays the same, and what an honest occupancy answer looks like under the new regime.
We're now Occivar, same mission, same logo, new name
CrowdVision.ai is now Occivar. The mission, the architecture, the team, the logo and the legal entity are unchanged. Only the name on the door is new. Here is what changed, what did not, and what you need to do.
The future of smart buildings: how building optimisation evolves as the demand for space accelerates
By 2050, urban populations grow by two billion. By 2030, commercial buildings host 4.12 billion IoT devices. By 2050, 85% of the global building stock has to retrofit to zero-carbon-ready. The next decade of building optimisation isn't an upgrade cycle - it's a re-platforming.
Socceroos 2-0 Türkiye: what Australia's World Cup opener tells us about fan-zone crowd analytics
The Socceroos beat Türkiye 2-0 in Vancouver to open Group D of the 2026 FIFA World Cup. The result is on the front pages. The bigger crowd story - fan zones, live sites, transport hubs and stadium concourses - is the one venue operators across Australia should be watching closely as the campaign rolls on.
Office utilisation without a badge log, anonymous occupancy for hybrid work
The badge-swipe utilisation report has been wrong for five years. Hybrid work killed it. Here's how anonymous floor occupancy would give a corporate property team back the number they actually need, without auditing anyone's day.
Pacing the venue exit to the train platform, the transport-hub seam
The platform fits 200. The next train is four minutes. The venue is about to release 1,800 people. Privacy-first occupancy data makes the math possible. Here's how the seam between a venue and a transport hub would be designed.
Live occupancy analytics that survive a procurement review
The reason most occupancy analytics never reach a council, a university or a hospital floor is the procurement review. The fix is structural, the architecture has to be the privacy answer, not a policy clause appended to a surveillance camera.
Egress in under twelve minutes: the math behind safer crowd flow
Stadium egress is the most consequential ten minutes of any event. A privacy-first density layer can turn the post-match crowd into a routing problem, not a bottleneck. Here's the math and the architecture.
Threshold alerts that do not cry wolf
The first alerting system every ops team turns off is the one that pings every dwell spike. Useful alerts have hysteresis, calm-periods and a routing graph. Here is how the Occivar alerts layer is built, and where the defaults come from.
Anonymous occupancy monitoring for aged-care corridors
Falls happen in the corridor between the bedroom and the bathroom at 3am. Anonymous corridor occupancy can produce a useful safety signal without putting a camera in a resident's room. Here's the architecture.
Sustainability reports that walk into the NABERS submission pack
A sustainability report that an ESG lead trusts at 4pm on a Friday is one that already speaks NABERS, already shows the baseline, and already cites the counts. Here's how the Occivar reports layer is designed to be audit-grade out of the box.
BMS, Slack, Snowflake: connecting occupancy to the stack you already run
Occupancy data is only as useful as the system it informs. BMS for HVAC, Slack for ops, Snowflake for analytics, REST for everything else. Here is how the Occivar integrations layer is designed to fit a stack, not replace one.
Edge AI for buildings: latency, hardware, and the on-device guarantee
An on-device inference guarantee is not a marketing slogan. It is a hardware choice, a model choice and a network choice, and it is the part of the architecture that makes the privacy answer load-bearing.
What dwell time tells you that footfall never can
Footfall is a denominator. Dwell is the story. Here's the method a privacy- first dwell-time pilot would use to surface where the attention is, where the bottleneck is, and what neither a badge log nor a foot-counter ever reveals.
Plain-English queries against occupancy data: the assistant
An ops lead shouldn't need a SQL course to find out how the library was used last Friday. The Occivar assistant takes plain-English questions against anonymous occupancy data and answers in charts, schedules and one-line summaries, without any of the data ever ceasing to be anonymous.
Civic festival safety without facial recognition: the architecture
Density thresholds in civic squares don't need to know who anyone is to keep them safe. Here's how a privacy-first occupancy layer would be configured for a civic festival, and why councils can sign off on the architecture without spending the community-trust budget.
Right-sizing library hours against actual occupancy, the method
Every Australian university library opens at 9am by reflex. The case for shifting opening hours to match actual demand starts with measuring the demand. Here's the methodology a privacy-first occupancy pilot would use.
Why "privacy-first" needs to mean "architecture-first"
Every vendor calls themselves privacy-first. Then you read the data-retention policy. Here is how we built the answer into the silicon instead of the slide deck.
Right-sizing HVAC against real occupancy a worked example
A 28-day worked example from RMIT Building 80 9,420 kWh avoided, $2,830 AUD saved, 6.4 t CO2e avoided. One schedule change, zero comfort complaints, full methodology.
The BGIG-backed civic pilot we are scoping
How we plan to take a privacy-first occupancy architecture through council legal review, and why the architecture is the unlock, not the metrics.
A note on counting accuracy: ground truth and humility
97.8% counting accuracy means 22 missed counts in 1,000. Here is where the misses come from, and how we tell the customer up-front.
No biometrics. No demographics. No exceptions.
A clean technical breakdown of what we deliberately did not build, and why each absence is structural rather than optional.
Heatwaves & crowds: how AI detects "heat stress" before it's an emergency
When temperatures push past 40°C, crowds stop moving in predictable lanes and start migrating toward any available shade. Computer vision analytics detect the velocity drops and dwell spikes that precede medical emergencies.
Retail 2026: why "passive" security is the only way to boost margins
Loss prevention and store operations have been siloed for decades. In 2026, the same camera that measures conversion rates can detect operational hazards in real time - and the unified data is where the margin comes back from.
3 crowd trends shaping Australia this week
Record Australian Open attendance, the "high-spend, low-traffic" retail paradox and a national heatwave - three movement patterns from one week that show why static crowd planning is over.
The invisible risk: why "smart" venues need smarter security in 2026
Security cameras are network entry points. As venues integrate IoT sensors, crowd counters and facial authentication, they keep expanding their attack surface. In 2026, smarter security means less data, not more.
The 2026 AI arms race is here: why smart cities need "edge resilience"
Autonomous "agentic" malware can now scan networks and adapt tactics on the fly. Smart cities that centralise surveillance have built the perfect target. Resilience in 2026 means decentralising at the edge.
Rain, runs and risk: analysing crowd flow during the Hurricanes vs Stars knockout
Hobart Hurricanes beat Melbourne Stars by 3 runs (DLS) in a wet-weather knockout final. The drama on the field was real - so was the crush risk in the concourses every time the covers went on.
Record breakers and risk takers: what 103,000 fans at the AO teach us about crowd safety
The 2026 Australian Open shattered attendance records three days running, peaking at over 103,720 fans. Officials suspended ground-pass sales to keep density safe. The headline isn't on the court - it's at the gates.
The "ghost meeting" epidemic: why booking apps aren't enough
Every meeting room booked, half of them empty. The booking app says full, the floor says ghost town. Passive occupancy detection eliminates the reservation that nobody attends.
Stop paying for empty desks: the truth about "shadow vacancy"
Corporate real estate is the second-largest expense for most businesses, and badge swipes are still the most common utilisation signal. Shadow vacancy - the desks that "look used" but aren't - is where the lease money is hiding.
The lunch-rush data gap: optimising amenity spaces with vision AI
Retailers have run queue analytics for years. Corporate offices haven't. The cafeteria, the lobby and the breakout zones are where the hybrid-attendance anchor lives - or doesn't.