Occupancy analytics: the complete guide for portfolio organisations

Occupancy analytics is the practice of continuously measuring how many people use a physical space, where, and for how long, and turning that measurement into operational, financial and compliance decisions. It has moved in a decade from a niche workplace tool to core infrastructure for anyone running a large property portfolio, and in the last two years something new has happened: regulators arrived. Climate disclosure regimes now demand evidence about how buildings perform, government tenants set energy ratings a building cannot reach while conditioning empty floors, and every one of those obligations lands hardest on the organisations with the most floors.

This guide is the full picture, written for the people who own that problem: heads of property, workplace strategy, sustainability and operations leads in organisations with big footprints, universities, councils, health networks, retailers, governments and corporates. It covers what occupancy analytics actually is, how the technology works, why portfolio scale changes the economics, the mandate landscape from the UN level down to Australian law, and how to implement it without stepping on the privacy rake. Where we have opinions, we declare them; we build one of these platforms, and this is the guide we wish every buyer had read before talking to any vendor, including us.

01 · WHAT IS OCCUPANCY ANALYTICS?What is occupancy analytics?

Occupancy analytics is the measurement and analysis of how people occupy physical space over time. A sensing layer counts people, a data layer turns counts into metrics, and a decision layer turns metrics into actions: resize the lease, retime the cleaning, right-size the air conditioning, open the second entrance, report the intensity figure.

The discipline stands on a small set of metrics, and confusing them is the most common analytical error in the field:

• Occupancy is how many people are in a space at a moment, against its capacity. A 400-desk floor with 180 people is at 45 per cent occupancy.
• Utilisation is how much of the time a space is used at all. A meeting room booked solid but empty half the time has 100 per cent bookings and 50 per cent utilisation. The difference between these two numbers is where most portfolio savings hide, and we unpack the formulas in occupancy rate versus utilisation rate.
• Peak versus average determines what you actually provision for. Buildings are sized, cleaned, conditioned and staffed for peaks; portfolios that only know their averages systematically over-provision.
• Dwell time is how long people stay, which converts a count into behaviour: a lobby with high traffic and low dwell is a corridor; the same lobby with rising dwell is a queue.
• Density is people per unit of area, the safety and comfort metric, and the one crowd-safety thresholds are written against.

What separates occupancy analytics from a headcount on a clipboard is continuity and granularity: per zone, per hour, every day, across every site, comparable on one screen. One floor measured for a month is an anecdote. A portfolio measured continuously is an asset register for space itself.

A worked example, because the metrics only bite with numbers

Take a 4,000 square metre floor with 400 desks, leased at $650 per square metre per year: a $2.6 million annual commitment. Badge data says 310 people came in on Tuesday, so the floor "needs" its 400 desks. Zone-level occupancy tells a different story: the peak concurrent count on that same Tuesday was 228, held for under an hour; the all-day average was 154; two of the six zones never exceeded 40 per cent of their local capacity all week; and Friday's peak was 61 people, on a floor conditioned, lit and cleaned for 400.

Run the same arithmetic the portfolio way. If that pattern holds across twenty similar floors, the estate is carrying roughly 170 desks per floor of peak headroom that is never used, call it 1,700 square metres per floor of provisioned-but-unoccupied space, or $1.1 million per floor per year in rent alone before energy, cleaning and fit-out amortisation. Multiply by twenty and the number stops being a facilities observation and becomes a board paper. None of this arithmetic is available from badge entries, because an entry is not presence, and it is certainly not zone-level presence over time. That distinction, entries versus occupancy versus utilisation, is the entire analytical foundation of the field, and it is why the sensing layer has to measure zones directly rather than infer them from a door.

The same worked-example discipline applies in the other direction, which buyers forget: occupancy data also tells you where you are under-provisioned. The two zones pinned at 95 per cent density every mid-morning, the meeting-room bank with a queue outside it, the library level that hits its safe-occupancy threshold every exam period: these show up in the same dataset, and fixing them is usually cheaper than the complaints they generate.

02 · HOW THE TECHNOLOGY WORKSHow the technology works

Every occupancy analytics system is a pipeline with four stages, and the choices at each stage determine accuracy, cost and privacy posture.

Sensing. Someone has to count. The main modalities are passive infrared (detects motion, cannot count), doorway beam and thermal counters (count crossings at thresholds), network-based counting from Wi-Fi or Bluetooth signals (counts devices, not people), and camera-based computer vision (counts people directly, with zone-level position). Each has a legitimate niche; we compared them honestly in do occupancy sensors use cameras? and, for the network-based option, why MAC-address counting breaks. The short version: doorway counters drift in multi-entry spaces, device counting swings with phone policy, and vision is the only modality that measures zones directly, which is why the decisive question in this category is not capability but architecture.

Inference. Where the counting happens matters as much as how. Edge AI runs the computer vision on the sensor itself: frames are processed on the device and discarded, and only the resulting numbers ever leave. The alternative, streaming video to servers for analysis, creates a surveillance system with analytics attached, and everything from your legal review to your network design inherits that fact. We wrote the deep technical version in computer vision on the edge, explained.

Aggregation. Counts become time series; time series become baselines; baselines make anomalies visible. This is where a platform earns its licence fee: zone-level analytics, heatmaps and trend baselines that a property team can read without an analyst translating.

Action. Data that never leaves a dashboard is a hobby. The counts need to reach the systems that act on them: the building management system for occupancy-driven HVAC, messaging and paging for threshold alerts, the warehouse for analysis, the reporting layer for the compliance pack. Integration depth, BMS, Slack, Teams, Snowflake, webhooks, REST, is covered in connecting occupancy to the stack you already run.

03 · WHY PORTFOLIO ORGANISATIONS NEED IT MOSTWhy portfolio organisations need it most

A single site can run on intuition; a manager who walks the floor knows roughly how it breathes. A portfolio cannot. Fifty buildings, hundreds of floors and thousands of zones produce more operational variation than any team can observe directly, and three structural forces make measurement non-optional at scale.

The arithmetic of small waste. At portfolio scale, minor inefficiencies compound into major line items. One floor conditioned for 200 people while hosting 80 wastes a modest sum; the same pattern across 300 floors is millions of dollars and thousands of tonnes of CO₂ a year. Hybrid work made this the default condition of corporate real estate: attendance concentrated into anchor days, leaving mid-week peaks that size the building and Friday troughs that expose it. The gap between what organisations lease and what their people occupy has a name, shadow vacancy, and it is invisible precisely because every individual instance of it is small.

Decisions with long tails. Leases, fit-outs and consolidations are decade-scale commitments made on evidence about how space is used. Badge logs and booking systems, the traditional evidence, measure entries and intentions, not occupancy; we covered why in office utilisation without a badge log and the ghost meeting problem. A consolidation decision made on badge data inherits its errors for the length of the lease.

Comparability. The portfolio question is never "how busy is this building" but "which of these forty buildings earns its keep". That requires the same metric, measured the same way, everywhere, which is an argument for treating occupancy as infrastructure with one methodology, rather than letting each site buy its own answer.

04 · THE MANDATE LANDSCAPE: FROM THE UN TO YOUR LEASEThe mandate landscape: from the UN to your lease

For most of its history, occupancy analytics was sold on savings. What changed is that the world's climate frameworks reached the building sector, and then reached into it, and the resulting obligations run in an unbroken chain from international agreements to the clause in your next government lease.

The international layer

The Paris Agreement commits its signatories to emissions trajectories that are impossible without decarbonising buildings, and the UN's Sustainable Development Goal 11 makes sustainable cities and communities an explicit target of the 2030 agenda. These frameworks do not name occupancy sensors, obviously. What they do is generate national law, and the pattern of that law is consistent: measure building performance, disclose it, and improve it. The European Union's Energy Performance of Buildings Directive is the template, noting that buildings account for roughly 40 per cent of EU energy consumption; its recast requires member states to drive stock toward zero-emission buildings with measured, disclosed performance. The International Sustainability Standards Board's climate standard has become the global baseline for corporate climate disclosure, and jurisdictions including Australia have legislated local versions of it.

Why this matters to a property team: performance-based regulation runs on operational data. A design certificate says what a building could do; the regimes now arriving ask what it did, this year, with evidence. Demand, when the building was actually occupied, and by how many people, is half of that evidence.

The Australian layer

Australia has moved from encouragement to obligation on three fronts, and portfolio organisations sit in scope of all three.

Mandatory climate reporting. Under the Australian Sustainability Reporting Standards, AASB S2 makes climate disclosure part of financial reporting. Group 1 entities (two of: $500 million revenue, $1 billion assets, 500 employees) began reporting from January 2025; Group 2 (two of: $200 million revenue, $500 million assets, 250 employees) entered the regime from 1 July 2026, this month, at the time of writing. Scope 1 and 2 emissions are disclosed under assurance, and for office-heavy organisations, scope 2 is substantially a story about building energy, which makes occupancy the denominator behind the number: energy per occupied hour is a defensible intensity metric; energy per square metre of mostly-empty floor is a confession.

Government leasing standards. The Net Zero in Government Operations strategy sets the bar for leasing to the Commonwealth: from July 2025, new leases over 1,000 square metres require a maintained 5.5-star NABERS Energy rating, and from July 2026 the requirement is 6.0 stars and all-electric operation, per the Department of Climate Change, Energy, the Environment and Water. We analysed what that deadline means for owners in the 6-star NABERS deadline; the one-line version is that 6.0 stars is not reachable by scheduling tweaks, it demands plant that runs against measured demand.

Performance disclosure. The Commercial Building Disclosure program requires energy ratings to be disclosed when sizeable office space is sold or leased, and NABERS ratings themselves are built on twelve months of measured operational data, re-earned annually. A rating regime based on operation, not design, is a regime in which how the building was actually used is evidence, every year, forever.

What occupancy data actually contributes to compliance

Precision matters here, because vendors in this market routinely overclaim. Occupancy data does not calculate your emissions; metering does. What occupancy contributes is the demand side of every energy story a portfolio has to tell:

1. Intensity metrics that survive scrutiny. Energy per person-hour of actual use is the honest efficiency measure, and it requires knowing occupancy. Floor-area intensity flatters empty buildings.
2. The evidence behind the reduction claim. "We cut HVAC energy 14 per cent by aligning plant to measured occupancy" is an auditable sentence with a data trail. The occupancy-driven ventilation mechanics are the how.
3. The sequencing logic for capital. Electrification and plant replacement should be sized to measured demand, not nameplate assumptions; measurement first means buying smaller plant once, a point we argue in the NABERS deadline analysis.
4. Utilisation honesty in the annual pack. Climate-conscious investors and boards increasingly ask the awkward question behind scope 2: why are we powering this much space at all? Occupancy data is the only credible answer, in either direction; audit-grade occupancy reporting is what that looks like in practice.

The investor layer: pressure without legislation

Alongside the statutory regimes sits a quieter force: capital. Real estate investors increasingly benchmark portfolios on sustainability performance through frameworks aligned with the Task Force on Climate-related Financial Disclosures and its ISSB successor standards, and fund-level benchmarking of building portfolios is now routine in institutional real estate. None of this is law, and none of it names sensors. But an owner reporting into these frameworks answers the same questions the statutes ask, how much energy, serving how much actual use, trending which way, and answers them annually to the people who price their capital. Portfolio organisations feel this earlier than legislation: the mandate arrives in the fund questionnaire before it arrives in the gazette.

05 · OCCUPANCY ANALYTICS AND THE ENERGY STACKOccupancy analytics and the energy stack

Because the mandate chain runs through energy, it is worth being concrete about the physical mechanisms that connect a people-count to a kilowatt-hour. There are four, in descending order of typical value:

1. Conditioning schedules versus actual presence. Most commercial HVAC runs to a timetable commissioned years ago for a fully occupied building. Aligning run-hours and setpoints to measured occupancy, including demand-controlled ventilation driven by CO₂ and counts together, attacks the roughly 40 per cent of building energy that HVAC represents, at exactly the hours when the building is emptiest and the waste is purest.
2. After-hours baseload. The building that "closes" at 6 pm and still draws half its daytime load at 9 pm is a portfolio cliché. An occupancy feed makes the after-hours question answerable zone by zone: who is actually here, and what is running for them?
3. Plant sizing at replacement. Electrification programmes replace gas plant with electric plant, and the cheapest kilowatt is the one you do not buy capacity for. Sizing new plant against measured demand rather than nameplate assumptions is the single largest capital saving occupancy data enables, and it only works if the measurement precedes the engineering.
4. The floor-consolidation dividend. The largest energy saving in the portfolio playbook is not running a floor better; it is not running it at all. Consolidating Friday operations onto half the floors, or exiting a building the data shows the organisation never needed, dwarfs every optimisation above it, and it is a decision nobody makes without defensible measurement.

The honest caveat, which we repeat everywhere including our accuracy work: the energy savings depend on the building's controls actually being able to respond. A portfolio with 1990s pneumatic controls will harvest less of mechanism one than a building with a modern BMS, which is why the pilot phase should include the controls audit, not just the counting.

06 · WHAT IT LOOKS LIKE ACROSS A PORTFOLIOWhat it looks like across a portfolio

The same measurement layer answers different questions in different buildings, which is the point: one methodology, many verticals.

• Corporate offices use it for right-sizing, anchor-day management and the consolidation case: which floors earn their keep, which leases end.
• Universities apply it to libraries and study spaces, timetabled space against actual attendance, and campus planning.
• Councils and cities measure civic buildings and public spaces to defend service hours and justify investment, the use case behind our City of Melbourne backed pilot work.
• Retail portfolios trade in dwell and conversion across store networks.
• Health and aged care run it for safety and flow in privacy-critical environments, where camera-based recording is a non-starter.
• Venues, stadiums and transport use live density for crowd safety and egress, where the count is a safety instrument, not an efficiency one.

07 · THE PRIVACY QUESTION, ANSWERED STRUCTURALLYThe privacy question, answered structurally

Nothing kills an occupancy programme faster than a workforce or a council discovering it is being watched. This is not a communications problem; it is an architecture problem, and it has an architectural answer.

Counting people does not require identifying anyone. A system that runs inference on the device, stores no video, transmits no video and recognises no faces produces data with no individual in it, which changes every downstream conversation: the Privacy Act analysis shortens because anonymous counts are not personal information, the workplace-surveillance consultation becomes honest and easy, and Australia's new statutory privacy tort has nothing to attach to because no footage of anyone exists. The full argument, no biometrics, no demographics, no exceptions, is the position we build to, and privacy by design documents it at the architecture level for legal review.

For a portfolio organisation the strategic point is bigger than compliance: the privacy architecture is what makes one methodology deployable everywhere, the office, the library, the hospital corridor, the civic square, instead of stalling wherever sensitivity is high. The buildings where measurement is hardest to get approved are usually the ones that need it most.

08 · IMPLEMENTING IT: FROM ONE FLOOR TO THE WHOLE ESTATEImplementing it: from one floor to the whole estate

Having watched procurement from both sides of the table, our advice is boring and firm: sequence it.

Weeks 1 to 6: one site, instrumented properly. Pick a floor or building with a suspected gap between provision and use. Sensor every zone, not just the entrance, and run against your existing data sources, badge logs, bookings, meters, without changing anything. The divergence between what you believed and what you measure is the business case, and it should include a ground-truth accuracy audit: a count nobody has verified on your site is a rumour with a dashboard.

Months 2 to 4: act on one building, publicly. Retime the HVAC against measured demand, consolidate the Friday floors, fix the worst meeting-room bank. Publish the before-and-after internally. The point is to prove the loop from data to action to saved dollars, because that loop, not the dashboard, is the product.

Months 4 onward: roll out with the boring disciplines. One metric methodology portfolio-wide, per-zone coverage as the default, integration into the BMS and reporting stack from day one, and the full cost structure priced over five years before signing anything: licence triggers, recalibration, re-zoning and exit terms decide the real bill, not the sensor quote.

09 · HOW TO EVALUATE A PLATFORM: THE PORTFOLIO BUYER'S FRAMEWORKHow to evaluate a platform: the portfolio buyer's framework

Shortlists in this category compare poorly because vendors describe themselves in incompatible units: one quotes per sensor, one per square metre, one bundles hardware into a service fee. The framework that cuts through is five questions, weighted for portfolio use, and it works on any vendor in the category, us included.

1. Zone truth. Does the system measure each zone directly, or infer it from doorways and devices? Direct zone measurement is what makes heatmaps, density safety and per-zone HVAC control real rather than modelled. Inference-based systems are cheaper and honest vendors will say where inference is good enough; the failure is not inference, it is inference sold as measurement.

2. Architecture of the sensitive step. Where does the counting happen, and what does the device emit? On-device inference emitting counts only is the answer that scales across a portfolio's most sensitive sites. Anything that records or streams footage carries its governance program with it into every new building, and the buildings where it stalls are predictable.

3. Portfolio-grade data model. One zone taxonomy, one metric dictionary, cross-site comparability, API access to raw counts, and export without ransom. Ask to see the same report for two different sites; if the vendor cannot, the portfolio view does not exist yet.

4. Total cost shape, not total cost number. The five-year per-zone-per-year figure, and which lines recur: licence trigger, recalibration, re-zoning, integration support. The arithmetic and the trap-list are in our cost guide; the short version is that recurring lines compound at portfolio scale and one-off lines amortise.

5. Evidence culture. Does the vendor volunteer accuracy audits, publish their misses, and document what leaves the device in verifiable terms? In a field this easy to overclaim, the willingness to be checked is itself the signal. It is the reason we publish our accuracy notes and the reason this guide names failure modes that include our own category's habits.

10 · THE SEVEN FAILURE MODES OF PORTFOLIO OCCUPANCY PROGRAMMESThe seven failure modes of portfolio occupancy programmes

We have watched enough of these programmes, ours and the other kind, to catalogue how they fail. Every one of these is avoidable, and every one is common.

1. The permanent pilot. One floor gets instrumented, produces an interesting report, and the programme never scales because nobody designed the methodology to be portfolio-ready: different zones defined differently per site, no comparable baseline, no owner. Scale the methodology from day one even when the hardware covers one floor.
2. The dashboard with no verbs. If the data does not reach the BMS, the alerting stack and the reporting pack, it decorates a screen until the licence lapses. Decide which three decisions the data will drive before the first sensor is mounted, then wire the integrations that serve those decisions.
3. Badge-data comfort. The organisation already "has occupancy data" from access control, so the programme dies in business-case review, and the estate keeps making decade-length decisions on entry counts that overstate presence and say nothing about zones. The rebuttal is a four-week parallel measurement; the divergence funds the programme.
4. The per-device pricing trap. A licence model that charges per sensor punishes exactly the zone-level coverage that makes the data useful, so coverage gets thinned to save licence cost and the analytics quietly stop being trustworthy. Interrogate the licence trigger before signing.
5. Privacy blowback. A camera-shaped device appears on a ceiling with no notice, a staff representative asks what it records, and the answer is complicated. The programme is then paused, sometimes permanently, regardless of the architecture's actual merits. Anonymous-by-architecture plus proactive notice makes this failure structurally unavailable; anything else makes it a matter of time.
6. Unverified accuracy. The counts are taken on faith, someone eventually spot-checks a room, finds the number wrong, and confidence in the entire dataset collapses portfolio-wide. One bad audit unwinds a year of adoption. Ground-truth validation at deployment, per site type, is cheap insurance against the most expensive failure on this list.
7. Methodology drift. Site A counts the lobby, site B does not; site C changed its zones in March; the January-versus-June comparison is quietly meaningless. Portfolio analytics is a data-governance exercise wearing a hardware costume: one zone taxonomy, one metric dictionary, versioned, owned.

11 · FREQUENTLY ASKED QUESTIONSFrequently asked questions

What is occupancy analytics in one sentence? Continuous measurement of how many people use each zone of a building and for how long, turned into metrics that drive space, energy, safety and compliance decisions.

How is occupancy analytics different from a people counter? A people counter is a device; occupancy analytics is the system around it: zone-level measurement, baselines, comparability across sites, and integration into the systems that act on the data. The counter answers "how many came in". The analytics answer "which spaces earn their keep".

Do occupancy sensors record video of staff or visitors? They should not, and the good ones structurally cannot. On-device architectures process frames and discard them, emitting only counts; our comparison of sensor types covers which modalities record and which cannot.

Is occupancy analytics legal in Australia? Counting is legal everywhere in Australia; the legal questions attach to recording. Anonymous counts are not personal information under the Privacy Act, while stored footage is, and carries workplace-notice and privacy-tort exposure. The full analysis: is people counting legal in Australia?

What does an occupancy analytics system cost? The sensor is the smallest line. Licence model, installation, calibration, governance and integration decide the real bill, which is why we price the question over five years per zone in the occupancy sensor cost guide.

How accurate are occupancy analytics systems? Good camera-based edge systems reach high-nineties percentage accuracy in realistic conditions, but the honest answer is always site-specific: density, mounting height, occlusion and lighting all move the number. Any accuracy claim should come with a ground-truth audit on your site; here is how we handle ours, including what the misses look like.

Does occupancy analytics require new cabling or network changes? Typically one PoE run per sensor and a segregated VLAN. The integration effort concentrates on the BMS side, where point-mapping in older buildings deserves days in the plan, not hours. The full checklist lives in the cost guide.

How long before the data pays for itself? The measurement typically surfaces its first defensible saving within one quarter: a conditioning schedule that assumed presence the data disproves, or a consolidation the numbers finally make arguable. The larger paybacks, lease decisions and plant sizing, land at the portfolio's own decision cadence; the data's job is to be already there, trusted and comparable, when those decisions open.

Where should a portfolio organisation start? One floor, fully zoned, measured against existing beliefs for a month, with a ground-truth audit. Then act on one building and publish the result internally. Scale the methodology, not the pilot.

12 · THE BOTTOM LINEThe bottom line

Occupancy analytics used to be a savings play, and it still is: the gap between provisioned and occupied space remains the largest unmanaged cost in most portfolios. What has changed is the second driver: a mandate chain running from the Paris Agreement through European directives and ISSB-aligned standards into Australian law, government leases and annual ratings, all of it converging on the same requirement, evidence of how buildings actually perform. Organisations with large portfolios will produce that evidence either way; the only question is whether they will also own the instrument that produces it, and use it to decide instead of merely disclose.

We build that instrument, anonymously, by architecture, and we are happy to be one of the vendors you put the hard questions to. Start with a walkthrough of your portfolio's worst floor; it is usually the most profitable conversation in the building.