IPMVP Option C vs Option D: How to Prove Your AI-HVAC Savings Are Real

By Robin Chien | IPMVP-Certified | 15+ Years Managing $1.2B in Commercial Real Estate

Vendor dashboards are marketing tools, not evidence.

I say this as someone who has deployed 15+ AI-HVAC systems and reviewed dozens of vendor performance reports. The internal savings numbers that vendors display on their dashboards are useful for operational monitoring. They are not sufficient for any external stakeholder who needs to verify that your building is actually saving what you think it is saving.

🤖

Run this analysis on your building

Our AI agents use the same methodology. First query free — no credit card.    Try the Agent →

This is where Measurement and Verification (M&V) enters the picture --- and specifically, the International Performance Measurement and Verification Protocol (IPMVP), which provides the internationally recognized framework for proving energy savings are real.

Why M&V Matters

You need rigorous savings verification when the audience extends beyond your own engineering team:

Green bonds and PACE financing. Lenders require independently verified energy performance data before they will underwrite green building instruments. Vendor-reported savings do not satisfy underwriting requirements.

Regulatory compliance. LL97 annual reports, BEPS compliance filings, and emissions trading schemes demand auditable M&V documentation. Self-reported numbers from your technology vendor are not auditable by a third party.

Institutional investor ESG requirements. If your building ownership includes pension funds, sovereign wealth funds, or GRESB-reporting entities, they need verification that withstands independent audit. GRESB scoring specifically evaluates the rigor of your energy measurement methodology.

Portfolio-scale accuracy. At 10+ buildings, small measurement errors compound into large financial misrepresentations. A 3% overstatement of savings across 50 buildings can represent hundreds of thousands of dollars in phantom value.

Disputing vendor performance. If you suspect your vendor is overstating savings, an independent IPMVP analysis settles the question with data, not arguments.

Option C: Whole-Building Analysis

IPMVP Option C uses utility meter data to compare pre-installation and post-installation energy consumption. This is the most common M&V approach for AI-HVAC because the intervention affects the entire HVAC system.

How it works:

  1. Collect 12+ months of pre-installation utility data (your baseline period).
  2. Build a regression model relating energy consumption to independent variables --- primarily outdoor temperature (heating degree days and cooling degree days), but also occupancy if data is available.
  3. After the AI system is installed and operating, collect 12+ months of post-installation utility data (your reporting period).
  4. Use the regression model to predict what energy consumption would have been in the reporting period without the AI system, based on actual weather conditions.
  5. The difference between predicted and actual consumption is your verified savings.

Best for:

Statistical requirements: CV(RMSE) must be 25% or less for monthly models. Normalized Mean Bias Error (NMBE) must fall within plus or minus 5%. If your regression model does not meet these thresholds, your baseline is not defensible.

Cost: $5,000-$15,000 per building.

Timeline: 2-4 weeks of analysis (after data collection is complete).

Option D: Calibrated Simulation

IPMVP Option D builds a calibrated energy model of your building and simulates what energy consumption would have been without the AI system. It is more expensive and complex, but it solves problems that Option C cannot.

How it works:

  1. Build a detailed energy simulation model of the building using as-built drawings, equipment specifications, operational schedules, and weather data.
  2. Calibrate the model against actual pre-installation utility data until it reproduces measured consumption within acceptable tolerance.
  3. Run the calibrated model for the reporting period using actual weather data but without the AI optimization.
  4. The difference between the simulated baseline and actual measured consumption is your verified savings.

Best for:

Cost: $15,000-$25,000 per building.

Timeline: 4-8 weeks.

When Vendor M&V Is Sufficient

I want to be clear: third-party IPMVP verification is not always necessary.

Most AI-HVAC vendors provide internal M&V that uses pre/post comparison with weather normalization. Runwise, for example, uses this approach, and it aligns with industry practice. For internal decision-making --- deciding whether to continue the pilot, whether to scale to more buildings, whether the technology is working as expected --- vendor-provided M&V is often sufficient.

The gap opens when the audience changes. The entity reporting the savings should not be the same entity whose revenue depends on demonstrating savings. This is not an accusation of fraud --- it is a structural conflict that external stakeholders will flag.

Rule of thumb: If the person reviewing your savings report has never heard of your vendor and has no reason to trust them, you need third-party verification.

The M&V Plan: What to Include

Every IPMVP analysis starts with a written M&V plan. Document these elements before the pilot begins, not after:

  1. Baseline period: Minimum 12 months, covering both heating and cooling seasons.
  2. Reporting period: Minimum 12 months, seasonally matched to baseline.
  3. Energy metric: Total building energy (kBtu), HVAC-only (if submetered), fuel (therms), or electricity (kWh).
  4. Independent variables: HDD, CDD, occupancy, production volume.
  5. Regression model specification: Linear regression at minimum; changepoint models for buildings with both heating and cooling.
  6. Non-routine adjustments: Document any equipment changes, occupancy shifts, or envelope modifications and how they will be quantified.
  7. Uncertainty analysis: Fractional Savings Uncertainty below 50% at 90% confidence is the standard threshold.
  8. Reporting schedule: Monthly tracking during the reporting period, final annual verification report.

Budget It In From the Start

M&V is not an afterthought --- it is a pilot design requirement.

Budget $5,000-$25,000 per building depending on whether you use Option C or Option D. This is not trivial for a single-building pilot, which is one reason I recommend the 3-building pilot framework: you amortize the M&V methodology development across three buildings. The first building is the most expensive to verify. Buildings two and three are incremental.

If you are running a pilot across 3 buildings at $10,000 average M&V cost, that is $30,000 --- a meaningful number. But compare it to the cost of deploying AI-HVAC across 50 buildings based on unverified savings estimates. If the vendor overstated savings by even 5%, you have made a portfolio-scale decision on faulty data.

The Career Angle

IPMVP certification costs approximately $1,500 and takes about 3 months of study. For Facility Managers looking to differentiate their career trajectory, it is the highest-ROI professional investment I can recommend.

The building industry has a critical shortage of practitioners who can bridge operational technology and financial verification. Owners who deploy AI-HVAC need someone who can evaluate whether the savings are real. That person commands a premium --- and the certification pays for itself on the first project where you are the one signing the M&V plan.

Next Steps

Full verification framework: The AI-HVAC Pilot Guidebook, Chapter 5 covers the complete IPMVP protocol for AI-HVAC deployments, including M&V plan templates.

Career path: The Execution Playbook maps the transition from Facility Manager to Strategic Building Intelligence Officer, including the certifications and skills that compound over a career.

Robin Chien is an IPMVP-certified practitioner with 15+ years managing energy optimization across a $1.2B portfolio. He writes at ai-smart-buildings.com.