Virtual Occupancy
28% energy reduction across a UK building portfolio, with no new sensors installed.
28%
Energy Reduction
Portfolio energy consumption cut through predictive, demand-based control.
35%
HVAC Efficiency
Performance improved by predicting demand rather than reacting to it.
94%
Air Quality Consistency
CO₂ and indoor air quality held within target throughout operating hours.
Client
UK IoT platform provider
Market
Commercial real estate & the built environment
Scale
Hundreds of commercial buildings, schools and sports facilities
Duration
12+ months
Team
Cross-functional data & platform team
Engagement
Embedded Partnership
Delivery hub
Belgrade & London
Capability
Data Engineering
Stack
Python, AWS, event-driven architecture, RAG, MCP database integration
The challenge
Conflicting objectives
Balancing energy efficiency against indoor air quality.
Sensor gaps
Many buildings lacked comprehensive occupancy sensors, with no capex to add them.
Reactive control
Systems responded to current conditions rather than anticipating demand, wasting energy.
The approach
Virtual occupancy
Occupancy inferred from existing air quality, HVAC and usage signals, with no new hardware.
Predictive HVAC
Heating, cooling and ventilation forecast from weather and historical performance.
Joint optimisation
Energy and air quality optimised simultaneously rather than traded off blindly.
Conversational interface
Operators query building data in natural language.
What we would tell you before starting
Virtual sensing is an inference, not a measurement. It was accurate enough to drive control and cut energy, but not to serve as a metered compliance figure. Where the client needed a measured occupancy value for reporting, we said so, because that requires hardware.
Last reviewed