Measurement and Verification with IPMVP: Did the Retrofit Actually Save Energy?

You replace the chillers, retrofit the lighting, and tune the controls. Next year's bill is lower. Did the project save energy, or was it a mild winter, a half-empty floor, and a lucky rate change? On the invoice alone you cannot tell, and that uncertainty is exactly what kills confidence in energy projects and performance contracts.

The International Performance Measurement and Verification Protocol (IPMVP), maintained by the Efficiency Valuation Organization, is the internationally accepted framework for answering the question properly. It does not tell you which meter to buy. It tells you how to structure a savings claim so that it holds up: what the baseline is, what you adjust for, and how you report the result.

The core idea is that you never actually measure savings. Savings are the absence of energy use, so they cannot be metered directly. You estimate them by comparing what the building did use against what it would have used without the project, adjusted to common conditions. This piece covers that logic, the four IPMVP options, and the adjustments that make or break a claim.

The equation everything rests on

Every IPMVP savings calculation reduces to one relationship, per the EVO principles:

Savings = (Baseline Period Energy - Reporting Period Energy) +/- AdjustmentsIPMVP core concept

The baseline period is measured before the retrofit, and it captures not just energy but the conditions that drive it: weather, occupancy, operating hours, production. The reporting period is measured after. The adjustments bring both periods to a common set of conditions so the comparison is fair. Skip the adjustments and you are just comparing two different years, which is the mistake IPMVP exists to prevent.

The four IPMVP options

IPMVP offers four options for determining savings. The choice is a trade-off between cost, accuracy, and how much of the building you need to look at. The U.S. Department of Energy's FEMP guidance and EnergyCAP's summary both lay these out clearly.

Options A and B: retrofit isolation

When you only care about the performance of one measure, you isolate it. Option A meters the parameter that matters most and estimates the rest. In a lighting retrofit you might measure the connected wattage precisely and estimate operating hours from a schedule. Option B is the same idea with everything measured, no estimates, which costs more but removes a source of dispute. Both suit projects where the affected system can be metered cleanly and the rest of the building is irrelevant to the claim.

Option C: whole facility

When several systems change at once, or the interactions between them matter, you measure the whole building at the utility meter. Option C is well suited to deep retrofits and to programs that pay on realized whole-building savings, such as Ontario's Save on Energy performance programs. Its strength is that it captures everything, including interactive effects. Its weakness is sensitivity: because savings are a fraction of total use, you need enough baseline history and solid adjustments, or the savings signal drowns in normal variation.

Option D: calibrated simulation

Sometimes there is no usable baseline. A new building, a gut renovation, or a project where the old meter data is unreliable. Option D builds an energy model, calibrates it to whatever metered data exists, and uses the model to represent the counterfactual. It is the most technical option and demands modelling skill, but it is often the only route when the baseline itself is missing.

Routine and non-routine adjustments

Adjustments are where M&V earns its credibility, and they come in two kinds.

• Routine adjustments account for factors you expect to vary and can model, above all weather. Correcting the baseline to the reporting period's degree days is the classic routine adjustment. Production volume in an industrial setting is another.
• Non-routine adjustments account for things you did not expect and did not build into the model: a new tenant, a change of use, added server load, a wing taken offline. These are handled case by case, and failing to account for them is the most common way a savings claim quietly becomes wrong.

Choosing an option in practice

1. Ask whether you care about one measure or the whole building. One measure points to A or B; the whole building points to C.
2. Check your baseline. Twelve or more months of clean whole-building data makes Option C viable; no usable baseline pushes you toward Option D.
3. Weigh the savings against total use. If the project saves a small fraction of a large bill, whole-facility noise may swamp it, favouring isolation.
4. Match the rigour to the stakes. A performance contract with payments tied to savings justifies more measurement than an internal check.

How MartinAI helps

Options B, C, and D all depend on clean, continuous whole-building energy history, and that is usually the hard part. MartinAI reads utility bills and interval data into gap-checked records per building and per meter, which is exactly the baseline and reporting-period data an IPMVP Option C analysis needs. Aligning consumption to billing periods, flagging estimated reads, and reconciling units all happen before the numbers reach the savings calculation.

Because the same platform aligns each building to local degree days, the routine weather adjustment at the heart of most M&V is supported directly rather than assembled by hand each time. And because occupancy and cost intensities sit next to consumption, the non-routine changes that would otherwise corrupt a claim, a new tenant or a change of use, are visible instead of hidden. Clean inputs do not write the M&V plan for you, but they remove the data arguments that stall most savings disputes.

Conclusion

M&V is not paperwork. It is the difference between a project you can defend and a lower bill you have to take on faith. IPMVP gives you a shared language: a baseline, a reporting period, routine and non-routine adjustments, and four options that scale from a single lighting circuit to a whole-building model. Pick the option that matches what you are trying to prove and the data you actually have, decide your independent variables before the work starts, and report savings as avoided energy. Do that, and 'did the retrofit save energy' stops being an opinion.