A measurement and verification plan defines before implementation how savings will be separated from changes in weather, production, occupancy or operation. It is particularly important for non-standardised actions and projects whose result depends on measured data. There is no generic plan for every case: boundaries, baseline, periods, instruments, adjustments and uncertainty must respond to the applicable methodology and current procedure.
Set boundaries and baseline before implementation
Define the measurement boundary before selecting meters. Draw the energy flows entering and leaving the affected system, identify interactive effects, and state whether savings will be calculated for an isolated asset, a subsystem or a whole facility. Record what remains outside the boundary and why. The applicable non-standardised methodology or other governing route determines the required approach; a familiar M&V convention cannot override it. Secure technical review of the boundary before implementation changes the system being characterised.
Build the baseline from a period that represents normal pre-action operation. Inventory interval consumption, production, weather, occupancy, schedules, setpoints, shutdowns and maintenance events, then assess completeness and time alignment. State inclusion and exclusion rules before viewing the result. A long period is not automatically representative, and a short stable period may still miss seasonality. Document anomalies rather than deleting them silently, and preserve raw extracts, query parameters, timezone and meter multipliers used to create the analysis dataset.
Design measurements and data governance
Specify the baseline equation and its rationale in reproducible form. Name the dependent energy variable, independent drivers, functional relationship, coefficients, units and valid operating range. Report diagnostics relevant to the selected methodology and inspect residuals for structure, not merely a headline fit statistic. Separate model development from final application where possible. If engineering calculations supplement regression, expose every conversion and assumption. Lock the approved equation before post-action outcomes are used, unless a controlled correction addresses a demonstrable error.
Create a meter schedule for every required point. Record identifier, location, measured quantity, range, accuracy class, pulse or communications factor, sampling interval, clock source, calibration status, installation date, owner and data destination. Confirm current-transformer ratios and sign conventions on site. Plan commissioning tests that compare expected and observed readings under known conditions. A dashboard is not the source of truth unless its transformations are documented; retain native exports and configuration evidence needed to reconstruct displayed values.
Define adjustments, quality and uncertainty
Govern data acquisition as an operational process. Assign responsibility for daily or weekly completeness checks, drift alarms, clock corrections and backups. Define how gaps are flagged, who may estimate them and which method is permitted for each gap length or operating state. Never interpolate through shutdowns as if production continued. Keep raw, cleaned and model-ready datasets separate, with scripts or formulas under version control. Each edited value needs a reason code so another analyst can reproduce the treatment.
Distinguish routine from non-routine adjustments. Routine adjustments apply the frozen baseline relationship to normal changes in drivers such as weather or production. Non-routine adjustments address lasting changes to the facility, boundary or service level, such as an added production line, changed opening hours or major space expansion. Define detection thresholds, required evidence, approval authority and calculation method. Do not use an adjustment to absorb poor model performance or manufacture an expected saving; show results before and after material adjustments.
Calculate, review and freeze the result
Calculate savings through a controlled workbook or script that preserves units and sign conventions. Trace each input to the approved dataset, apply stated adjustments, quantify excluded periods, and aggregate only after interval-level checks. Address uncertainty in the manner required by the applicable methodology, including measurement, sampling and model components where relevant. Test sensitivity to influential assumptions and ensure savings from overlapping actions are not counted twice. Independent recalculation should reproduce the reported value from the frozen package.
The final M&V package should contain the approved plan, boundary diagram, baseline dataset and equation, meter register, calibrations, raw and treated reporting data, adjustment log, calculation and review record. Explain deviations from the pre-action plan and their effect rather than retrofitting the plan to match available data. Verification can examine whether quantified savings are supported, but it does not guarantee administrative validation or issuance. Retain enough provenance for a later reviewer to rerun the analysis.