Intelligence Hub

Deep insights into AI, smart buildings, and commercial real estate technology

Technical Architecture

4-Layer Sensor Fusion Model

True occupancy intelligence requires fusing multiple sensor modalities. Click each layer to explore how data flows from raw signals to actionable building intelligence.

Layer 1 — HVAC Telemetry

Supply/return temps, valve positions, airflow rates, VFD speeds. The baseline mechanical signal.

What it measures: Zone temperatures, supply air temperature, return air temperature, chilled water valve position, fan VFD speed, damper positions, and thermal load calculations.

Accuracy alone: 40–55% for occupancy detection. Thermal lag creates 15–30 min blind spots. Cannot distinguish between solar load and human presence.

Key limitation: Reactive, not predictive. Tells you what happened, not what's happening now.

BACnetModbus5-min intervals

Layer 2 — CO2 / IAQ Sensors

CO2 concentration as a proxy for occupancy density. Adds the "breath signature" of a zone.

What it measures: CO2 parts per million (ppm), typically 400 ppm (outdoor baseline) to 1200+ ppm (high density). Also VOCs, PM2.5, humidity in advanced IAQ sensors.

Accuracy with Layer 1: 60–70%. CO2 buildup correlates to person-count but has 8–15 min latency due to air mixing. Windows and ventilation rates introduce noise.

Key insight: CO2 delta (rate of change) is more valuable than absolute level. Fast rise = recent arrival. Slow decay = zone clearing.

NDIR sensors1-min polling400–5000 ppm

Layer 3 — Wi-Fi / BLE Positioning

Device association counts from APs. Uniquely identified device count = person count proxy.

What it measures: Connected and probing devices per access point. RSSI triangulation for zone-level positioning. Dwell time per zone. MAC randomization handled via association-based counting.

Accuracy with Layers 1+2: 80–88%. Real-time (sub-minute), unique device count, zone granularity. But: visitors without Wi-Fi, multi-device users, and guest vs. employee create noise.

Key advantage: Only modality that provides spatial flow data — movement patterns, zone transitions, peak corridors.

802.11axBLE 5.010-sec refresh

Layer 4 — Motion / PIR

Passive infrared and active motion detection. The ground-truth binary: someone is here, or they aren't.

What it measures: Binary presence/absence (PIR), motion direction (dual-element PIR), people counting (ToF/LIDAR sensors at doorways). Sub-second response time.

Fused accuracy (all 4 layers): 92–97%. PIR eliminates the "phantom occupancy" false positives that CO2 and Wi-Fi create. Doorway counters provide ground-truth calibration for all other layers.

The fusion payoff: HVAC alone = 45%. Add all 4 layers = 95%+. That 50-point accuracy jump is the difference between 15% and 28% energy savings.

PIRToF/LIDAR<1 secGround Truth

Fusion Accuracy Model

Layers CombinedAccuracy
HVAC Only~45%
+ CO2~65%
+ Wi-Fi~85%
+ Motion (Full Fusion)~95%
Each additional sensor modality addresses the blind spots of the previous layer. The marginal accuracy gain diminishes, but the marginal energy savings compounds.
Market Intelligence

PropTech Market Pulse

Real-time tracking of the trends shaping AI-powered building intelligence. Data refreshed weekly from industry sources.

4-Layer Sensor Fusion Architecture

Click each layer to explore how fusion eliminates false positives

1

HVAC Sensors

Temperature · Humidity · Airflow · Valve Position

45%

The foundational layer — HVAC sensors provide direct mechanical system data including supply/return air temperatures, relative humidity, volumetric airflow rates, and actuator positions. As a standalone signal, they achieve ~45% occupancy accuracy due to thermal lag and zone bleed-through artifacts.

2

CO₂ Monitoring

Concentration · Ventilation Rate · Air Quality Index

+20% → 65%

CO₂ concentration serves as a direct proxy for human metabolic activity. When fused with HVAC data, it eliminates false occupancy readings from solar heat gain and equipment load. Ventilation rate correlation adds +20% accuracy, bringing the cumulative to 65%.

3

Wi-Fi Analytics

Device Count · Dwell Time · Zone Density

+20% → 85%

Wi-Fi probe requests and association data provide real-time device counts per zone. Dwell time analysis distinguishes between transient foot traffic and sustained occupancy. Zone density heatmaps enable predictive pre-conditioning. Adds +20% cumulative accuracy to 85%.

4

Motion / PIR

Presence Detection · Movement Patterns · Zone Coverage

+10% → 95%

Passive infrared (PIR) and ultrasonic motion sensors provide the final verification layer. Movement pattern analysis distinguishes between occupied and recently-vacated spaces. Combined with all three previous layers, quad-fusion achieves 95% occupancy accuracy — the threshold for reliable demand-controlled ventilation.

Cumulative Accuracy

100% 75% 50% 25% 0%
45%

Accuracy by Fusion Depth

Single Layer

45%

Dual Fusion

65%

Triple Fusion

85%

Quad Fusion

95%

CRE & PropTech Market Signals

Key indicators driving smart building investment decisions

Regional Investment Distribution

North America
45%
Europe
30%
APAC
25%

Vendor Market Share

JCI
22%
Siemens
19%
Honeywell
17%
BrainBox AI
12%
Others
30%

Practitioner Tools & Resources

Decision-grade frameworks for CRE operations leaders. Free access — email-gated for serious practitioners.

📊
AI-HVAC TCO Benchmark Report
Total cost of ownership analysis across 7 leading AI-HVAC platforms
⚖️
Energy Compliance Risk Matrix
Regulatory risk scoring across APAC, EU, and North American jurisdictions
🗺️
Decarbonization Roadmap Template
12-month implementation framework for net-zero building operations
🔧
FDD/Predictive Maintenance Case Pack
Real-world fault detection data with ROI proof points from 4 deployments
Vendor Evaluation Matrix
Weighted scoring framework for AI-HVAC and smart building platforms
🧮
AI-HVAC ROI Calculator
Input your portfolio metrics, get IPMVP-grade savings projections