Demand-Controlled Ventilation: The Occupancy and CO2 Data a Building Needs Before It Can Run It (EPBD 2026)

Walk a typical European office floor on a Friday afternoon and the air handling unit is still pushing design-rate fresh air into meeting rooms nobody has booked since lunch. The XY Sense index for April to September 2026 puts average office utilisation at 46 percent, with Tuesday at 55 percent and Friday at 35 percent, and 40 percent of meeting room sessions involving a single person. A ventilation system sized for full occupancy and run on a fixed schedule is therefore conditioning air for roughly twice the people who are actually breathing it, most of the week.

Demand controlled ventilation (DCV) is the obvious fix, and every supplier will tell you so. What the generic articles skip is the uncomfortable part: DCV is only as good as the signal that drives it. The LBNL review of 38 DCV studies found ventilation energy savings of up to 60 percent, but it also found 26 percent overconsumption in some cases. The difference was not the damper hardware. It was whether the building had trustworthy occupancy and CO2 data for each zone the system was trying to control.

What EPBD building automation actually requires from 29 May 2026

The recast EPBD, Directive (EU) 2024/1275, Article 13, is the deadline most facility managers are now working to. Its building automation and control systems (BACS) duty already applies to non-residential buildings with heating or air-conditioning systems over 290 kW since 31 December 2024, and extends to systems over 70 kW by 31 December 2029 where technically and economically feasible. From 29 May 2026, the BACS in covered buildings must be capable of monitoring indoor environmental quality. Not controlling it, not optimising it, but monitoring it: which means a measurement layer that records CO2 and related parameters per space, not a single sensor in the return duct.

If you are waiting for your national regulator to tell you exactly what counts, note that on 15 July 2026 the Commission sent letters of formal notice to all 27 member states for incomplete transposition. National detail is lagging. The directive text, however, is in force, and lenders and tenants in several markets are already writing EPBD Article 13 readiness into lease and refinancing conditions. Being able to show per-zone CO2 and occupancy records is the practical proof.

Which CO2 levels your office or shop must hold

The set-point you give the DCV logic depends on jurisdiction, and the spread is wider than most people assume:

• Denmark (BR18 for schools and daycare), Sweden (FoHMFS 2014:18) and Germany (ASR A3.6) cluster around 1,000 ppm, with Germany requiring further measures above 2,000 ppm.
• The UK HSE treats readings consistently above 1,500 ppm in an occupied room as a sign of poor ventilation.
• Spain's RITE works in increments above outdoor air: IDA 2 offices at 500 ppm above ambient, IDA 3 shops at 800 ppm above ambient, so your outdoor reference sensor matters as much as the indoor ones.
• Dubai Municipality guideline DM-HSD-GU119-IAQ sets 800 ppm CO2 as an 8-hour average for existing buildings, and Al Safat clause 502.02 explicitly requires demand-controlled ventilation with a CO2 set-point below 800 ppm and an alarm above 1,000 ppm in dense spaces.

A Gulf operator running a Dubai mall has, in effect, a stricter target than a Copenhagen office, and an 8-hour average means you cannot satisfy it with a system that only reacts once a room is already stuffy. Acceptable co2 levels office tenants will accept are also drifting downward on their own; the cognitive performance evidence in our 28 September post explains why some occupiers now write 800 ppm into their own workplace standards regardless of what the local code says.

Why CO2 alone makes DCV sluggish, and why occupancy alone makes it blind

Here is the practitioner observation that separates DCV projects that pay back from those that quietly get switched back to schedule. CO2 is a lagging indicator. In a 20-seat meeting room with a 60-minute booking, concentration can take 15 to 25 minutes to cross the set-point, and by the time the damper has opened and the air has turned over, the meeting is half done. The room was under-ventilated for the first third of the session and then, because CO2 decays slowly, over-ventilated for twenty minutes after everyone left. Multiply that across a floor and you get the 26 percent overconsumption LBNL warned about.

Occupancy sensors solve the lag. A people count at the door or a presence signal in the room tells the BMS that twelve people have just walked in, and the airflow can step up before CO2 has moved at all. But occupancy alone cannot verify the result. A count of twelve says nothing about whether the air handling unit actually delivered, whether a filter is loaded, or whether a tenant has blocked a diffuser with a storage cabinet. You need the CO2 reading to close the loop and to produce the compliance record the EPBD now asks for.

The working rule we apply: occupancy leads the control, CO2 validates it and logs it. Both feeds have to be zoned to the same spaces the dampers actually serve, which is where many retrofits fail. A single CO2 sensor at the floor level averaged across six meeting rooms and an open plan area will keep the open plan comfortable and let the small rooms climb past 1,500 ppm every afternoon.

Shopping centres: tenancy hours are the real schedule

HVAC is 40 to 60 percent of a mall's energy bill, and industry estimates put 30 to 40 percent of retail HVAC energy in unoccupied hours when central plant schedules do not follow actual tenancy hours. An anchor supermarket opening at 07:00, a food court busy until 23:00 and a fashion wing that is dead after 20:00 are all served by the same AHUs on the same timer. Footfall counting at mall entrances and zone level already exists in most centres for leasing purposes; feeding that same count into the BMS as a ventilation demand signal is usually a software integration rather than a new sensor rollout. The CO2 layer then handles the cinema, the food court and the prayer room, where density per square metre changes faster than door counts can show.

Building the data layer before the control layer

A sensible sequence for a building that has to meet Article 13 and wants DCV to pay:

• Map ventilation zones against physical spaces. Where one VAV box serves several rooms, decide whether to sub-zone or accept the compromise, and document it.
• Deploy occupancy sensors per zone. Ceiling-mounted people counters from partners such as Xovis give counts at entrances and in larger spaces; desk-level presence sensors (see our 11 October post) cover open plan areas. Counting accuracy is contractually a minimum of 96 percent and typically 98 to 99 percent when lighting, layout and visitor behaviour allow; for DCV that margin is more than sufficient.
• Deploy CO2 and temperature/humidity monitors per zone, with an outdoor reference where the code uses differential ppm. Milesight LoRaWAN units are a common choice where cabling to a BMS is impractical.
• Run three to four weeks of monitoring without control changes. This baseline is what proves savings later and is itself the EPBD monitoring evidence.
• Only then hand the signals to the BMS and commission the DCV sequences, zone by zone.

Vemco's indoor environmental quality monitoring and occupancy management solutions are built to deliver exactly these two feeds into one platform with BMS integration, so the same dataset serves compliance reporting, space planning and ventilation control.

Frequently asked questions

What is demand-controlled ventilation and how does it work? DCV adjusts outdoor air supply to a space based on measured demand instead of a fixed schedule. Occupancy counts or presence signals raise airflow when people arrive, and CO2 readings confirm the air quality target is being held and trim the rate when the room empties.

Does the EPBD require building automation in every building over 70 kW now? No. Under Directive (EU) 2024/1275 Article 13 the BACS duty has applied to non-residential buildings with heating or air-conditioning systems over 290 kW since 31 December 2024, and extends to systems over 70 kW by 31 December 2029 where feasible. The indoor environmental quality monitoring capability applies in covered buildings from 29 May 2026.

What CO2 level is acceptable in an office? Most European guidance clusters around 1,000 ppm, Spain's RITE IDA 2 uses 500 ppm above outdoor, and Dubai's DM-HSD-GU119-IAQ sets an 800 ppm 8-hour average. The UK HSE treats readings consistently above 1,500 ppm as a sign of poor ventilation.

How accurate are occupancy sensors for measuring room utilisation? Ceiling-mounted people counters deliver a contractual minimum of 96 percent accuracy, and typically 98 to 99 percent when lighting, layout and visitor behaviour allow. For ventilation control, that is far more precise than the fixed design occupancy the schedule assumed.

Can demand-controlled ventilation use more energy than a fixed schedule? Yes. The LBNL review of 38 studies found savings up to 60 percent but 26 percent overconsumption in some cases, usually caused by poorly zoned or drifting sensors and CO2-only control that reacts late and ventilates empty rooms. Reliable per-zone occupancy and CO2 data is the precondition for savings.

How do occupancy sensors protect employee privacy? People counters and presence sensors used for DCV output a number or a presence state, not images or identities. Processing happens on the device and only counts are sent to the building platform, which keeps the data outside personal-data categories in most jurisdictions.

If you need to show EPBD Article 13 monitoring evidence by May 2026 or commission demand controlled ventilation that actually cuts HVAC energy in empty rooms, talk to Vemco about a zoned occupancy and CO2 data layer for your building.