Researched by BEAST Library Curator | Verified by Harper | Quality: 8.5/10
BLUF: Occupancy analytics has quietly crossed from "nice dashboard" to hard M&V-grade input — CBRE's 2026 study of 303 million sq. ft. shows utilization up to 53% and desk-sharing as the new norm. But most facility teams still point that data at only one job: desk math. The faster, larger payback in 2026 is to feed the same occupancy stream into your HVAC control loop, where occupant-count demand-controlled ventilation (ODCV) is cutting HVAC energy up to 40% and peak demand charges 10–25%. Here is how I'd wire it in 90 days.
The number that should change your budget conversation
For four years, "we don't know how full the building actually is" was a legitimate excuse. In 2026 it isn't. CBRE's 2026 Global Workplace & Occupancy Insights — drawn from client portfolios totalling 303 million sq. ft., average portfolio 5 million sq. ft. — puts the numbers on the table:
| Metric | 2023 | 2024 | 2025 | What it means for you |
|---|---|---|---|---|
| Average office utilization | 35% | 38% | 53% | You are conditioning space for 100%, using half |
| Peak utilization | — | — | 80% | First time at target since early 2020 — peaks are real again |
| Orgs setting explicit utilization targets | — | — | 87% | Nearly half target 76–85% |
| Clients where >40% of staff have no assigned desk | — | — | 69% | Sharing ratios of 1.01–1.49 (a third push past 1.5) |
The gap between a 53% average and an 80% peak is the whole story. It says two things at once: you have far too much conditioned space most of the time, and you have genuine crush moments you cannot ignore. A static HVAC schedule can serve neither — it over-conditions the empty Monday and under-serves the packed Wednesday. That gap is exactly what an occupancy-driven control loop is built to close.
Why your occupancy data keeps under-delivering
Walk into most buildings that bought sensors in the last two years and you'll find the data doing one job: feeding a space-planning deck once a quarter. That's real value — sharing ratios are now informed by job function (83%), utilization data (78%), and supply/demand data (68%) per CBRE — but it's a lease-cycle payback measured in years, and it needs a real-estate decision-maker to act.
The HVAC payback needs no lease event and no boardroom. It is an operational change your controls contractor can make this quarter. Yet the occupancy stream and the BMS almost never talk. That disconnect is the single largest piece of stranded value in a 2026 smart building.
The reframe: feed HVAC first
Occupant-count demand-controlled ventilation modulates outside-air and conditioning against how many people are actually in a zone, not a nameplate design occupancy. The published savings are not marginal:
| Lever | Documented saving | Source basis |
|---|---|---|
| ODCV (occupant-count DCV) vs. fixed ventilation | up to 40% HVAC energy | Occupancy-based ventilation studies (MDPI Sustainability 17/9/4140) |
| CO₂-modulating control vs. simple ON/OFF | 60% lower consumption, IAQ maintained | Demand-controlled ventilation field assessment |
| Peak demand-charge reduction via DCV | 10–25% of peak kW charges | Commercial DCV peak-shaving analyses |
| Whole-system HVAC analytics (schedule + FDD + Cx + predictive) | 18–35% (avg 22%) | Aggregated commercial case studies |
Schedule optimization alone — simply not conditioning genuinely empty hours — is 5–12% of that, and it is the lowest-risk starting point. You are not touching comfort setpoints on day one; you are removing conditioning from space the sensors prove is empty.
The 90-day play
- Weeks 1–3 — Instrument one floor, not the portfolio. Pick your worst average-vs-peak floor. API-first occupancy platforms now deploy in weeks, not months, feeding data straight into the BMS and BI tools.
- Weeks 4–8 — Baseline, then schedule-optimize. Establish an IPMVP Option C (whole-facility) or Option B (retrofit-isolation) baseline before you touch a sequence — otherwise you cannot defend the savings. Then implement occupancy-gated scheduling: the safe 5–12%.
- Weeks 9–13 — Layer in ODCV on the AHU serving that floor. Modulate outside-air to live count, hold your CO₂ ceiling, and log the demand-charge delta across at least one peak billing period.
Every step is reversible and every step is metered. That is the difference between a control-sequence change and a science experiment. See our Library for the companion M&V-standards briefs that pin the baseline discipline.
Choosing the sensor: privacy-first is now table stakes
If your occupancy data is going to drive real-time building control, it will sit under legal and works-council scrutiny — especially in APAC. The 2026 market has converged on privacy-by-design, but the modalities differ, and the modality is the compliance story:
| Platform | Sensing modality | Privacy posture | Best fit |
|---|---|---|---|
| Butlr | Thermal (heat signature) | No PII, enforced at the hardware layer — no cameras, no device tracking | Wall-to-wall coverage, fast deployment, deep stack integration |
| XY Sense | On-device spatial (X/Y coordinates only) | 100% anonymous coordinates, no images ever, GDPR-compliant by design | Teams wanting spatial detail beyond headcount, air-quality on the roadmap |
| VergeSense | Camera + on-device inference | Imagery processed on-device; only anonymized occupancy leaves the sensor | CRE teams wanting rich planning analytics |
| Basking | Wi-Fi + access-control fusion | Infrastructure-based, no new hardware | Portfolio-wide real-time view using existing network data |
For a Taiwan or wider APAC deployment I would default to a modality where the privacy claim is hardware-enforced (thermal or coordinate-only), not a software toggle. Under Taiwan's PDPA and the tightening APAC consent regimes, "the camera never stores an image because we configured it that way" is a weaker legal position than "the sensor physically cannot capture one." That distinction is worth more in a works-council meeting than any dashboard feature.
What I'd do if this were my building
My portfolio is APAC, my grid is Taipower, and my tenants skew toward the office-advocate end — people come in, and the peaks are real. So I would not chase the desk-reduction headline first; I'd chase the kilowatt. Concretely:
- Aim the first pilot at demand charges, not rent. In a Taipower commercial tariff, the peak-kW component is where ODCV pays back inside a single billing cycle. A 10–25% peak reduction on a TSMC-adjacent office or a mixed-use tower is a line item finance will notice this quarter.
- Keep the space-planning payoff as the second act. Once the sensors are in and trusted for control, the sharing-ratio conversation (1.01–1.49 seats) rides on the same infrastructure for free — a rare case where the operational win and the real-estate win share a capital line.
- Refuse any sensor that can't survive a privacy audit. If it can capture an image, it will eventually be asked to prove it didn't. Hardware-enforced anonymity removes that entire class of risk.
The one-line version for your next capital review: we already own the occupancy signal that can cut our HVAC bill up to 40% and shave peak charges 10–25% — it's just plugged into the wrong system. Fixing that is a 90-day controls project, not a lease negotiation.
Have a question about this topic? Ask our CRE AI Agent →