IPMVP Option C baseline adjustments: routine, non-routine, and the utility data behind them

Option C is the M&V option that lives on utility bills. Instead of metering a chiller or a lighting circuit, you take the whole-facility meter, model how the building used energy before the project, and compare that model's prediction with what the meters show afterward. It is the natural choice when several measures interact and when the question the owner asks is "did the bill go down."

The catch is that a whole-facility meter records everything: the retrofit, the weather, the new tenant on the third floor, the server room that arrived in month eight. Savings only mean something once those effects have been adjusted out, and every adjustment is built from data that has to be complete, continuous and aligned with the bill periods. For the four options side by side, see measurement and verification with IPMVP.

The equation everything else serves

The FEMP M&V Guidelines write the whole method in one line: Savings = (Baseline Energy minus Post-Installation Energy) plus or minus Routine Adjustments plus or minus Non-Routine Adjustments. The purpose of the adjustments, in the same document's words, is to express both baseline and post-installation energy under the same set of conditions. The 2024 draft of Version 5.0 keeps the same equation and notes that adjustments can be made to either the baseline or the post-installation period. What makes Option C hard is that the two energy terms come from utility meters you do not control, and the adjustment terms come from data the M&V plan has to specify in advance.

What Option C is and when it fits

EVO's IPMVP Core Concepts 2022 defines it directly: the meters measuring the supply of energy to the whole facility can be used to assess performance and savings, and the measurement boundary encompasses the whole facility. Options A and B isolate a retrofit; Option D uses a calibrated simulation; Option C is the only one that reads the same meter the invoice does.

FEMP is specific about when it fits. Version 3.0 says it suits complex equipment replacement and controls projects where predicted savings are relatively large, greater than about 10 to 20 percent of the site's energy use on a monthly basis, and the 2024 Version 5.0 draft tightens that to greater than about 10 to 15 percent of the consumption measured by the utility or submeter, warning that smaller savings risk being lost in the noise of monthly data unless the model is very highly predictive. Both editions ask for at least 12, and preferably 24 or more, months of pre-installation data for the baseline model and at least 9, preferably 12, months of performance data. The 5.0 draft adds that because buildings change, Option C is often best used for 2 to 3 years before switching to a retrofit-isolation approach. The fit, then, is several interacting measures, savings large enough to see through monthly bill variability, and a facility that is not about to change use.

Routine adjustments: building the baseline regression

Routine adjustments handle the variables everyone expects to move. FEMP describes them as accounting for expected variations in independent variables such as temperature, humidity, meals served or production, usually with regression, with the method fixed in the M&V plan before award. In practice the baseline model is a regression of monthly (or daily, where interval data exists) consumption against one or more drivers over 12 to 24 months of bills. Weather enters as heating and cooling degree days computed for each bill period; base temperatures and variable choice are covered in degree days in energy analysis and weather normalization. Occupancy, hours and production enter where they vary month to month and can be measured.

FEMP adds three rules that decide whether the model is usable: every independent variable that affects consumption must be specified, whether or not it is in the model; driver data must correspond to the billing meter reading dates; and models are best built on whole-year sets, 12, 24, 36 or 48 months, so the seasons balance. Once fitted, the model is driven with the performance period's weather and other drivers to produce adjusted baseline energy, and the difference from metered energy is the avoided energy. FEMP is blunt that bill-to-bill comparison without regression is unreliable and not recommended on federal performance contracts.

Non-routine adjustments and static factors

Static factors are the things the model assumes stay put. FEMP 5.0 defines them as factors within the defined measurement boundary that are not expected to change during the performance period, such as occupancy and operating hours, and lists the usual suspects: square footage, occupancy, operating hours, equipment loads, configuration or function. When one changes, a non-routine event has occurred. EVO's practical exercise describes non-routine adjustments as covering changes in static factors which could not reasonably be foreseen and considered when the M&V plan is written.

That exercise shows the failure mode. A facility with a baseline of 200 MWh a year installs four variable frequency drives expected to save 18 MWh a year, 9 percent of baseline, verified under Option C. A significant new seasonal space-heating electricity load is added after the baseline period, and the whole-facility meter shows negative savings. The lesson is not the arithmetic: the M&V plan lacked the baseline data and equipment schedules needed to size the adjustment, and by the time the gap appeared, that data could no longer be gathered.

A worked example: a CHP plant replaced by heat pumps

Consider a campus that verified an envelope and controls retrofit under Option C, with electricity and gas as separate whole-facility meters. Two years into the reporting period, the owner retires the combined heat and power plant and installs electric heat pumps. This is a change in a static factor, not a routine driver: the gas meter falls sharply, the electricity meter rises, and neither movement has anything to do with the original measures. The method is the same as for any non-routine adjustment. Document the event with dates, the equipment removed and added, and its rated capacity. Quantify the effect on each meter separately, by engineering calculation or by sub-metering the new plant, aligned to the bill periods it affected. Adjust the baseline on each meter so the comparison is like for like again, recording the calculation and its uncertainty, and have both parties sign it off, because a fuel switch of this size can easily exceed the savings being verified.

The utility data requirements underneath

FEMP names the three data types Option C runs on: utility billing or other metered data, independent variables such as cooling degree days, and information on unrelated changes at the site. The first has four properties a regression cannot do without.

• Complete: every meter and commodity inside the boundary, for every period, including tenant and sub-meters that feed the total
• Continuous: bill periods that abut with no gaps and no overlaps, with estimated reads replaced by actuals when the true-up arrives
• Unit-normalized: gas in one energy unit across the series, electricity energy separated from demand, time-of-use blocks kept where the tariff has them (FEMP prefers time-of-use data because it shows more about consumption patterns)
• Weather-aligned: degree days computed for the exact bill period, not the calendar month

The 5.0 draft adds a pre-check: review the previous 12 to 36 months of bills for anomalies and seasonal variation before the baseline is fixed. A rebill, a meter swap, a rate-class change or a run of estimated reads inside the baseline window shows up as model error and gets mistaken for savings, or for their absence. Keeping the data honest across a multi-year reporting period is covered in M&V data for retrofit programs.

What the M&V plan has to say

FEMP 5.0 says the plan should list all the possible static factors that might affect savings, the thresholds that trigger a non-routine adjustment, and the method for making one. A defensible Option C plan documents at least the following.

1. The measurement boundary and every meter inside it, with account and meter identifiers
2. The baseline period (whole years) and the source and version of every bill used
3. The independent variables, their sources, their frequency and how they map to bill periods, plus the regression form and its validation targets
4. The static factors, their baseline values, monitoring frequency and the thresholds that trigger a non-routine adjustment
5. The method for quantifying a non-routine adjustment, who approves it, and the reporting period and format

How RETScreen and other tools consume this

Most teams do not fit the regression by hand. RETScreen's performance-analysis workflow, energy management information systems and M&V spreadsheets all take the same inputs: a continuous consumption series per meter, a weather series aligned to it, and a log of events. What they cannot do is repair the series: a gap becomes a missing point, an overlap a doubled month, a unit change a step that looks like savings. Preparing the import is covered in using utility data in RETScreen. This is the part MartinAI is built for: it reads every field on every bill and meter file across electricity, gas, water and steam, checks each period for continuity, estimation and unit consistency, aligns the series to the bill dates, and keeps the source document linked to every figure.

Option C is honest about what it measures: the whole bill, adjusted. The adjustments are only as credible as the utility data behind them, collected continuously, normalized once and documented from day one of the baseline.