Design practices that see how their buildings actually perform.
Architects, MEP engineers and sustainability consultants specify a measurement layer into the project, then compare the model with measured use system by system. Post occupancy evaluation runs on data, and the practice builds a performance record it can show.
What design and architecture teams are up against.
Water, energy, air and operations carry equal weight. Each one runs the same loop: monitor, measure, identify, act, verify.
Water: Water design checked against use
Fixture and demand assumptions compared with metered mains and sub-meters, the area where design feedback is usually weakest.
Energy: The model against the meter
Modeled loads per system compared with weather-normalized measured use, so a gap is located to a system rather than reported for the whole building.
Air: Comfort read alongside consumption
Temperature, humidity and CO2 per zone measured next to the energy that delivered them, so indoor quality and load are judged together.
Operations: Post occupancy evaluation with numbers
A measured record across the first seasons in use, ready for the aftercare and seasonal commissioning window while it is still open.
Problem, solution, impact.
How Aqueduct resolves each problem for design and architecture, and what changes when it does.
Fixture and demand assumptions compared with metered mains and sub-meters, the area where design feedback is usually weakest.
A live water balance across meters and sub-meters localizes the loss, routes the fix and confirms it closed at the meter.
Loss found from the balance, not the bill, and every fix verified.
Modeled loads per system compared with weather-normalized measured use, so a gap is located to a system rather than reported for the whole building.
Weather-normalized baselines per building and asset show the waste in money; schedules and setpoints adjust where control is enabled.
Savings stated in money and confirmed at the meter, not estimated.
Temperature, humidity and CO2 per zone measured next to the energy that delivered them, so indoor quality and load are judged together.
Air quality per zone, overlaid with occupancy, with ventilation raised when it matters and verified back in range.
Targets held during occupied hours, with the energy cost counted.
A measured record across the first seasons in use, ready for the aftercare and seasonal commissioning window while it is still open.
Faults diagnosed in plain language and routed into your CMMS with the fix; early signs of failure schedule maintenance first.
Fewer reactive call-outs and closures confirmed by the equipment.
One incident, from first signal to verified fix.
An illustrative example of the flywheel on a typical site.
Meter and sensor schedule specified at design stage, mapped to the zones in the energy model.
Commissioned with the BMS and tagged to site, floor and zone. A weather-normalized baseline built in the first season.
Divergence: measured heating in one wing running well above the model, traced to schedule and setpoint.
Finding shared with the owner and the FM team, with the tuning change recommended.
Measured use moves toward the model. The corrected assumption goes into the practice's next design.
Four connected products on one data layer.
Replace what should be replaced. Connect the rest.
- Occupant surveys as the only POE
- Design-stage rating used as the only evidence
- Ad hoc data requests to owners
- BMS through BACnet and Modbus
- Energy model outputs for comparison
- Owner and FM reporting
Three equal ways in: keep your BMS, replace it, or build Aqueduct in at design stage.
The questions each seat asks first.
“Can we prove our buildings perform, not just certify?”
“Which of my modeling assumptions were wrong, and by how much?”
“Can post occupancy evaluation carry measured data instead of a survey?”
Standards and regulations we supply evidence for.
We supply the measured inputs. You run the program and file the report.
Six assumptions every model makes, and how each one gets tested
An energy or water model is a stack of assumptions. In most projects none of them is ever checked against what the building did. Measurement turns each one into a question with an answer.
How many people, and when
Models assume an occupancy profile per zone. Real buildings fill late, empty early and run half-full on Fridays. The difference shows up in ventilation, lighting and plug loads.
- Occupancy or CO2 per zone, compared with the modeled profile
- Weekday and weekend patterns read separately
- Zones that never reach design occupancy identified
- Measured
- occupancy counts or CO2 per zone
- Model
- occupancy schedule per zone type
- Aqueduct
- plots the two side by side for each zone
When plant actually runs
Plant schedules in the model rarely survive commissioning. Air handlers start early, run late and run through holidays because nobody set the calendar.
- Run hours per AHU, chiller and boiler
- Out-of-hours operation flagged against the design schedule
- Holiday and weekend running made visible
- Measured
- run state and energy per plant item
- Model
- operating hours per system
- Aqueduct
- shows run hours against the schedule, week by week
What the zones are held at
A degree or two of setpoint drift, or heating and cooling deadbands that overlap, can move a building far from its model without anyone noticing.
- Zone temperature and humidity against design setpoints
- Simultaneous heating and cooling found by zone
- Setpoint changes logged so their effect is measurable
- Measured
- zone conditions from sensors and BMS points
- Model
- design setpoints and deadbands
- Aqueduct
- finds zones held outside the design band
How well the plant converts energy
Models use rated efficiencies. Installed plant runs at part load, with fouled coils and staging that was never optimized, so delivered efficiency can be well below the rating.
- Chiller and boiler energy against load served
- Staging and part load behavior visible
- Drift over seasons tracked against the first baseline
- Measured
- plant energy with flow and temperatures where metered
- Model
- rated efficiency curves
- Aqueduct
- compares delivered with rated performance per plant item
What tenants plug in
Small power is often the least certain input in a model, and in office and lab buildings it can rival HVAC. Separating it at the board shows what the assumption should have been.
- Small power metered per floor or tenancy
- Base load at night compared with daytime use
- Feeds a corrected plug load density for the next project
- Measured
- small power per board
- Model
- assumed W per square meter
- Aqueduct
- turns the measured load into a density you can reuse
How much water the building really uses
Water design usually stops at fixture counts and a per person demand. Metered mains and sub-meters show the real split between domestic, cooling, irrigation and leaks.
- Mains reconciled against sub-meters, so leaks are not counted as demand
- Cooling tower make-up separated where present
- Irrigation measured against weather
- Measured
- mains and sub-meter flow
- Model
- fixture counts and demand per person
- Aqueduct
- keeps a live water balance for the building
A measurement specification that survives procurement
These are the lines that make the comparison possible later. Each is cheap at design stage and costly as a retrofit.
| Element | What to specify | Why it matters later |
|---|---|---|
| Main incomers | Revenue grade meters on every incomer, readable over Modbus or BACnet | Whole building use for benchmarking and disclosure in every market |
| HVAC plant | Separate metering for chillers, boilers or heat pumps, pumps and major AHUs | Locates a performance gap to the plant rather than to the whole building |
| Lighting and small power | Separate boards or circuits per floor or tenancy | Tests the plug load and lighting assumptions, and supports tenant billing |
| Water | Mains plus sub-meters on cooling make-up, irrigation, kitchens and major risers | Closes a water balance, which is where design feedback is weakest |
| Zone air | Temperature, humidity and CO2 sensors per zone type | Reads comfort and ventilation next to the energy that delivered them |
| BMS integration | Open protocols, accessible points and an agreed naming convention | Lets the measurement layer read the BMS without a custom project |
| Data rights | Owner holds the data, with agreed access for the design team | Lets the practice keep the comparison and publish what the owner allows |
How the model meets the meter
Map model zones to meters
Each zone and system in the model is matched to the meters and sensors that will measure it. Where the two do not line up, either the schedule or the comparison is adjusted before anything is installed.
Commission and tag
Devices are installed with aQ Atria and tagged in aQ Core to the same site, floor and zone structure the model uses, so the comparison does not need manual mapping later.
Build the measured baseline
Aqueduct builds a weather-normalized baseline from the first months of operation, using regression on use against cooling and heating degree days. Weather stops being an excuse for the gap.
Compare system by system
Measured use per system is set against the model's prediction for the same weather. Divergences are ranked by size and traced to occupancy, schedule, setpoint or plant behavior.
Feed it back
Findings go to the owner and FM team for tuning during aftercare, and the corrected assumptions go into the practice's next model.
The systems where modeled and measured usually part ways
Not every system diverges equally. These are the ones worth metering separately, and what the comparison typically turns up.
Central cooling
Part load operation and staging that the model assumed would be optimal.
Chiller and pump energy, with flow and temperatures where available.
Compares delivered efficiency with the design curve across the season.
Heating plant
Boilers or heat pumps running outside the heating season or against cooling.
Plant energy and run state.
Flags heating that overlaps with cooling, zone by zone.
Ventilation
Fresh air delivered at design rate to zones that are half empty.
AHU energy and run hours, zone CO2.
Shows where ventilation could follow occupancy, and the energy it costs.
Lighting
Controls overridden, daylight sensors disabled, lights on overnight.
Lighting circuits per floor.
Finds after-hours lighting and compares it with the design schedule.
Small power
Night base load and plug density higher than assumed.
Small power boards per floor or tenancy.
Turns measured load into a density the next model can use.
Domestic water
Leaks and running fixtures counted as demand.
Mains and riser sub-meters.
Keeps a live balance so a leak shows as a residual, not as demand.
Cooling towers
Make-up and blowdown well beyond what the cycles of concentration should need.
Make-up and blowdown flow.
Reconciles each tower and flags drift in cycles.
Irrigation
Fixed schedules watering through rain.
Irrigation sub-meter with local weather.
Compares irrigation with rainfall and flags waste.
Closing the loop on one project
Start with a single scheme reaching design stage, or one completed building whose model you still hold.
Agree the question
Decide what the practice wants to learn from the building, and which rating the owner will face. That sets the meter schedule.
Write it into the specification
Meter and sensor schedule, open protocol requirements and data rights go into the MEP specification and the BMS package.
Commission with the BMS
Devices installed and tagged with Atria, checked at commissioning, and the first baseline started before occupation.
Measured POE
Season by season comparison with the model, tuning findings to the FM team, and a post occupancy report with numbers.
A performance record
The comparison repeats across completed projects, giving the practice a record that compares buildings with each other and with their models.
From a portfolio of intentions to a record of outcomes
Validated assumptions
Occupancy, schedules, plug loads and plant efficiency corrected from evidence instead of carried forward by convention.
Measured POE
A post occupancy evaluation built on energy, water and indoor air data across the first seasons, not only an occupant survey.
A defensible bid claim
Measured results across your own completed projects, which a competitor cannot match without the same data.
Aftercare that tunes
Soft landings with evidence, while the cheapest window to fix a performance problem is still open.
Research material
Comparable measured data for papers, teaching and conference work, used as the owner allows.
No delivery risk
Milvian AI and its partners install and support the layer. The practice specifies it and reads the results.
Learn moreSources9 references cited on this page
- NABERS, National Australian Built Environment Rating System
- Commercial Building Disclosure Program, Australian Government
- MEES tightened to EPC B for large commercial buildings (GOV.UK minimum energy efficiency standard guidance)
- Energy Savings Opportunity Scheme (ESOS), GOV.UK
- National Energy Code of Canada for Buildings 2020, National Research Council Canada
- Energy Conservation Building Code, Bureau of Energy Efficiency, India
- Al Sa'fat Dubai Green Building System, Dubai Municipality
- ANSI/ASHRAE/IES Standard 90.1, ASHRAE
- ENERGY STAR Portfolio Manager, US EPA
Proven in live sites across sectors.
Anonymized results, verified at the meter.
sites in 21 countries run on Aqueduct.
and 9 clinics in one of India’s top hospital chains.
US airport running water intelligence on Aqueduct.
of commercial real estate under Aqueduct.
Features that matter here
Common questions
Usually the owner or developer, because they operate the building and carry its rating. The practice's role is to decide whether measurement is in the specification at all, which is why referral is the usual starting point.
No. It is vendor agnostic and reads BACnet, Modbus and vendor APIs from whichever BMS the project buys. It sits above the BMS rather than replacing it.
Yes. An existing building can be metered and baselined from its own history or uploaded utility bills, then compared with the original design intent where the model is still available.
With the owner's agreement, the measured comparison against your model, the post occupancy record and the corrected assumptions. Over several projects that becomes a performance record the practice can cite.
Yes. Water is usually designed from fixture counts and a demand assumption with no post completion check, so it is often where design and use diverge most. In water stressed markets it is also the utility under the most pressure.
Through the same platform the owner uses, with access the owner grants. Comparisons can be exported for reports, teaching and publication.
No. A design-stage model is enough to start a comparison. Where the model is calibrated later, the measured data is what it is calibrated against.
Aqueduct regresses measured use against cooling and heating degree days to build a weather-normalized baseline. The model's prediction and the measured result can then be compared for the same weather.
Yes. Metered mains and sub-meters form a live water balance, which is compared with the design demand. Leaks show up as a residual rather than inflating the demand figure.
No. Installation and support sit with Milvian AI and its partners. The practice specifies the layer and reads the data the owner shares.
With the owner's agreement, yes. Many owners are comfortable with anonymized results, and the practice can cite them in bids and research.
See Aqueduct on your own facility.
Bring a site, a utility bill or a question. We will show you where the loss is.