IPMVP option C: baselines that survive scrutiny

Option C is the whole facility approach in the International Performance Measurement and Verification Protocol. You take the utility meter, model what the building would have consumed without the project, and call the difference savings. It is the cheapest option to instrument and the easiest to argue about, because everything rests on a baseline that no longer exists.

That is the practical problem. Nobody disputes the meter reading. They dispute the counterfactual. This is how to build a baseline that holds up when a client, a lender or a program reviewer pushes on it.

When option C is the right choice

Option C suits projects where the savings are large enough to see through the noise of the whole building, and where isolating each measure would cost more than the certainty is worth. A deep retrofit touching several systems at once is the usual case.

It is the wrong choice when the expected savings are small relative to normal variation at the meter. The working rule most practitioners apply is that savings below roughly ten percent of total consumption are difficult to defend at the whole facility level, because ordinary year to year variation can be that large on its own. When savings are that small, or when only one system changed, isolation options A and B are more appropriate. Our overview of the IPMVP options covers where each one fits.

Building the baseline

A baseline is a model, not a number. It relates consumption to the things that drive it, over a period long enough to contain the full operating cycle of the building. Twelve months is the normal minimum, because anything shorter cannot contain both heating and cooling seasons.

1. Fix the measurement boundary. For option C that is the utility meter, and every fuel crossing it has to be in the model.
2. Assemble at least twelve months of consumption, verified against the bills rather than typed from them.
3. Assemble the independent variables for the same periods, aligned to the same read dates.
4. Fit the model and test it, then keep the residuals. You will need them later.
5. Write down the operating conditions of the baseline period, including anything unusual that you decided to keep or exclude, and why.

Weather is almost always the dominant variable, and degree days are the usual expression of it. The base temperature matters more than people expect, and choosing it by fitting rather than by convention is what separates a model that tracks from one that drifts. We go through that in heating and cooling degree days in energy analysis.

Routine adjustments

Routine adjustments handle the variables you expect to change every period: weather, occupancy, production volume, operating hours. They are built into the model, so the reporting period calculation applies them automatically. The protocol expects these to be defined in advance, in the M&V plan, rather than chosen after the results are known.

The discipline is to select variables that have a physical relationship to consumption and can be measured independently. Federal guidance for performance based contracts, published in the FEMP measurement and verification guidelines, is a useful reference for how far to take this in a contractual setting, and ASHRAE guidelines cover the statistical treatment in more depth.

Non-routine adjustments, with a worked example

Non-routine adjustments handle the things nobody expected: a floor goes dark, a tenant leaves, a server room is added, a chiller is replaced for reasons unrelated to the project. These are where option C disputes actually happen, because each one is a judgement call made after the fact.

Take an illustrative case. A baseline year at 4,000,000 kWh. The project is expected to save about 12 percent. Nine months into the reporting year, the owner adds a data closet drawing a steady 20 kW, running continuously. That addition consumes roughly 20 kW multiplied by 24 hours multiplied by 90 days, which is 43,200 kWh over the remaining three months of the reporting year.

Reported consumption for the year comes in at 3,560,000 kWh. Measured against the adjusted baseline, raw savings look like 440,000 kWh, or 11 percent, slightly under target. Add the 43,200 kWh of new load that has nothing to do with the project, and savings attributable to the project are 483,200 kWh, or about 12.1 percent. The project met its target. Without the adjustment it did not, and the argument would have been about credibility rather than arithmetic.

The data quality that decides the outcome

Option C is only as good as the meter history behind it, and most of the failures are data failures rather than modelling failures. Estimated reads that were never corrected distort a baseline month. Read dates that shift by several days make a monthly comparison compare different lengths of time. A meter multiplier change applied mid period silently rescales everything after it. A missing month filled in by interpolation becomes a fact nobody remembers inventing.

• Verify consumption against the source bill or the meter record, not a spreadsheet copy.
• Normalize read periods, or work in daily averages, so unequal periods do not distort the fit.
• Flag estimated reads and decide explicitly how to treat them.
• Keep the raw values alongside any corrections, with a note on what changed and why.
• Where interval data exists, use it to explain anomalies that monthly data cannot, as covered in interval data versus monthly bills.

Reporting the result

A defensible option C report states the boundary, the baseline period and its conditions, the model and its fit statistics, the independent variables and their sources, every adjustment with its basis, and the savings with an honest statement of uncertainty. It should be possible for someone else to reproduce the number from the report alone.

That is also the test for whether your data system is doing its job. If assembling the report means reopening old spreadsheets and re-deriving numbers by hand, the result is not reproducible, whatever the model says. For programs that fund retrofits on verified savings, the reporting requirements are usually stricter again, which we cover in M&V data for retrofit programs. Teams running verification across a portfolio of buildings rather than a single project usually want it sitting next to the energy data itself, which is what VE-MAP is built for.