Siemens Says 20% Energy Reduction. Here's How to Verify That Claim (A Practitioner's IPMVP Guide)

This week, Siemens published a claim for its Digital Twin Composer platform: 20% energy reduction.

It's a confident number. It's also, without a methodology disclosure, unverifiable.

That's not unique to Siemens. "20% savings" has become the standard claim in smart building marketing — specific enough to sound credible, vague enough to be impossible to audit. Before you use it to justify a deployment budget, justify a sustainability claim to your board, or benchmark against your own portfolio, you need to know whether the number holds under professional scrutiny.

Here is how to verify it — or any similar claim — using the International Performance Measurement and Verification Protocol.

What IPMVP Is and Why It Matters

The International Performance Measurement and Verification Protocol (IPMVP) is the global standard for quantifying energy and water savings in buildings and industrial facilities. Developed by the Efficiency Valuation Organization and recognized by the U.S. Department of Energy, the European Commission, and major ESG reporting frameworks (GRESB, LEED, BREEAM), it defines four options for measuring savings — each with specific requirements for measurement scope, data collection, and uncertainty reporting.

IPMVP is not a best practice. It's the basis on which performance contracts (EPC), green bonds, and GRESB portfolio submissions are audited. A savings claim that doesn't specify an IPMVP option is like a financial statement that doesn't specify an accounting standard. It may be accurate. You have no way to know.

For a "20% energy reduction" claim from an AI building platform, the relevant question is: which IPMVP option was used, and does the methodology match the claim scope?

The Four IPMVP Options — Which One Should Apply

For a platform like Siemens Digital Twin Composer — a whole-building AI system touching multiple subsystems — the appropriate option is almost certainly Option C (whole-facility regression) or Option D (calibrated simulation). A claim based on Option A stipulation across a diverse portfolio would be methodologically inappropriate.

The Three-Step Audit for Any Vendor's Savings Claim

Step 1 — Establish What "20%" Is Being Measured Against

A percentage savings claim requires a baseline. The most common methodological weaknesses in AI building savings claims are baseline-related:

• Cherry-picked baseline period. Using the worst-performing 12 months as the baseline inflates apparent savings. A sound baseline covers a full calendar year (or multiple years) under typical operating conditions, not a period selected after the fact.
• Uncontrolled occupancy changes. If the building's occupancy dropped from 95% to 70% during the measurement period — as happened widely in 2021–2023 — energy use would decrease regardless of any AI intervention. The baseline model must account for occupancy independently.
• Boundary mismatch. A 20% reduction in HVAC energy is not a 20% reduction in total building energy. Lighting, plug loads, and process equipment may dominate consumption in some buildings. The claim boundary — HVAC only? whole facility? — must be explicit.

Ask the vendor: "What was the baseline period? What was the occupancy adjustment methodology? What energy systems are inside the measurement boundary?"

Step 2 — Assess the Regression Model Quality

For an Option C whole-facility claim, the regression model that normalizes for weather and occupancy must meet the IPMVP calibration standard. The key metric is CV(RMSE) — Coefficient of Variation of the Root Mean Square Error:

• Monthly data model: CV(RMSE) ≤ 20% AND R² ≥ 0.75
• Hourly data model: CV(RMSE) ≤ 30% AND R² ≥ 0.75
• ASHRAE Guideline 14 alignment: Model bias (MBE) ≤ ±5% monthly, ±10% hourly

A vendor that cannot provide these statistics on its regression model has not done Option C verification. It may have done a simple before-and-after comparison normalized for degree days — which looks like M&V but lacks the statistical defensibility required for audit.

Ask: "What is the CV(RMSE) and R² for your baseline regression? Can you share the model specification and calibration period?" If the answer is "that level of detail isn't something we typically share," the claim is marketing, not measurement.

Step 3 — Check for Independent Verification

Self-reported savings are not independently verified savings. The difference matters for ESG reporting (GRESB requires third-party verification for portfolio-level claims), for energy performance contracts (savings must be verified by a qualified independent party), and for any building where the claimed savings is being used to support a green bond or sustainability-linked financing.

The credibility hierarchy for a savings claim:

1. Independent M&V agent verified (highest). A qualified third party — typically a licensed professional engineer with CMVP or CPMP credential — has reviewed the methodology, audited the data, and signed off on the calculation.
2. Client self-reported with IPMVP methodology disclosed. The vendor has published the option, boundary, regression statistics, and baseline parameters. A competent buyer can audit the methodology even without independent verification.
3. Vendor-reported with methodology on request. Methodology exists and can be reviewed under NDA. Weaker than published, but verifiable.
4. Vendor-reported headline only (lowest). "20% energy reduction" with no methodology disclosure. Treat as unverified until documentation is provided.

The Siemens Digital Twin Composer claim, as published, falls in tier 4. That doesn't mean it's false. It means you cannot currently verify it without requesting the underlying methodology documentation.

What a Verified Claim Should Look Like

For reference, here is what an IPMVP-compliant savings claim statement looks like at minimum:

"Building X achieved a 17.4% (±3.2%) reduction in whole-facility electricity consumption compared to the 12-month baseline period (January–December 2023), normalized for weather (HDD/CDD) and occupancy using an IPMVP Option C linear regression model (CV(RMSE) = 14.3%, R² = 0.91, MBE = +1.8%). Measurement period: January–December 2024. Verification performed in accordance with IPMVP 2022 Edition, Chapter 4."

Note: specific percentage, uncertainty range, measurement boundary, baseline period, normalization variables, model statistics, measurement period, and applicable standard. All verifiable. All auditable.

Compare to: "20% energy reduction." One of these can appear in a GRESB submission. The other cannot.

Applying This to Your Own Portfolio

The Siemens claim is a useful test case, but the same framework applies to any AI building platform you're evaluating or any internal energy program you're reporting on.

Before your next ESG report, sustainability committee presentation, or vendor selection process, run your own savings claims through the three-step audit:

1. Is the baseline period, boundary, and occupancy adjustment methodology documented?
2. If whole-facility, is the regression model calibrated to IPMVP/ASHRAE Guideline 14 standards?
3. Has any part of the verification been independently reviewed?

If the answer to all three is no, you have a performance story, not a performance measurement. The difference surfaces in the audit room.

→ Ask the AISB CSIO agent to verify your building's energy savings claims — bring your vendor's claim documentation, your interval data, or your baseline parameters. The agent applies IPMVP Option B/C verification logic and tells you what holds and what doesn't. No fee, no form.

Related reading: The IPMVP Verification Framework: A Complete Guide to Options A, B, C, and D | The 176% Funding Surge in AI Buildings: What the Capital Signal Tells Us About Vendor Credibility