Energy-Savings Proposals: Compare Baselines Before Comparing Percent Savings

Short answer: two energy-saving proposals are not comparable just because both promise a percentage reduction. A factory owner must first align the measurement boundary, baseline period, production variables, operating hours, weather treatment, excluded events, reporting interval and savings equation. The useful question is not “which percentage is larger?” but “would both bidders calculate savings from the same measured facts?”

This guide is written for an FDI factory owner, finance team, facility manager and EPC/procurement team preparing or reviewing an efficiency proposal in Vietnam. It is an engineering and procurement framework, not a guaranteed savings claim, financial offer or substitute for an approved measurement and verification plan.

Why headline savings percentages can mislead

Energy savings cannot be read directly from a meter because the meter records consumption after many conditions have changed. Production volume, product mix, shift pattern, ambient temperature, compressed-air demand, cleanroom pressure, occupancy, maintenance outages and added equipment can all move consumption independently of an efficiency measure.

A proposal that compares this July with last July may be reasonable for one stable facility and unsuitable for another factory that added a line, changed operating hours or experienced a long shutdown. A bidder may also report electrical energy saved at one machine while another bidder reports net imported energy at the point of common coupling. Both calculations may be internally correct and still answer different questions.

ISO 50006:2023 provides guidance on energy performance indicators and energy baselines, while ISO 50015:2014 provides general principles for measurement and verification of organizational energy performance. The 2014 edition of ISO 50015 was confirmed current by ISO in 2025. These standards guide the process; they do not supply a universal model, threshold or guaranteed result for a particular factory.

Freeze eight items before comparing bidders

If these items are not common, ask bidders to recalculate against an owner-issued baseline specification. Do not average incompatible percentages.

Use one transparent savings equation

For a reporting period, a practical high-level structure is:

Energy savings = adjusted baseline energy − reporting-period energy

The adjusted baseline is the energy the agreed model estimates would have been used during the reporting period under the reporting-period conditions, before the efficiency measure. Depending on the agreed convention, documented non-routine adjustments are then incorporated in the baseline or reported separately. The contract must state the sign convention so the same change cannot be added twice.

Cost savings require an additional calculation:

Cost savings = avoided energy charges + avoided demand charges − new operating costs

Energy and cost are not interchangeable. A tariff change can increase cost savings without increasing kWh savings. Rooftop solar can reduce imported energy while total facility use remains similar. A battery can reduce peak demand while increasing total imported kWh because of conversion losses. Keep each quantity visible.

Worked example: the model choice changes the answer

Consider a hypothetical factory. The inputs below are teaching assumptions, not actual QuangAnhcons project data:

• Baseline production: 100,000 units/month.
• Baseline electricity: 1,200 MWh/month.
• Reporting-period production: 112,000 units/month.
• Reporting-period electricity after the measure: 1,080 MWh/month.
• A physically reviewed two-part baseline model: 600 MWh/month fixed load plus 0.006 MWh per unit.

Check 1 — raw comparison

The unadjusted difference is 1,200 − 1,080 = 120 MWh. Dividing by the raw baseline gives 120 / 1,200 = 10.0%. This calculation ignores the 12% production increase.

Check 2 — two-part adjusted baseline

The variable term for 112,000 units is:

0.006 MWh/unit × 112,000 units = 672 MWh.

The adjusted baseline is therefore:

600 MWh + 672 MWh = 1,272 MWh.

Adjusted savings are:

1,272 − 1,080 = 192 MWh, or 192 / 1,272 = 15.1% of the adjusted baseline.

Independent unit and arithmetic check

0.006 MWh/unit equals 6 kWh/unit. At baseline production, the variable component is 6 × 100,000 = 600,000 kWh = 600 MWh; adding the 600 MWh fixed term reproduces the 1,200 MWh baseline. At reporting production, 6 × 112,000 = 672,000 kWh = 672 MWh. The units and both endpoints reconcile.

Why a simple specific-energy comparison gives a different result

Baseline specific energy is 1,200,000 kWh / 100,000 units = 12.0 kWh/unit. Applying that ratio to 112,000 units predicts 1,344 MWh. Compared with 1,080 MWh, this implies 264 MWh or 19.6% savings. That is 72 MWh more than the two-part model.

The difference is not a rounding error. The simple ratio assumes all baseline energy varies directly with production. The two-part model assumes 600 MWh remains fixed. An owner should not choose the model that produces the largest claim. The bidder must demonstrate the physical relationship, data quality, valid operating range and predictive performance before the model is accepted.

Baseline structures to compare

1. Fixed historical baseline

A fixed baseline may suit a stable load with little change in output or operating hours. It is simple to audit but weak when relevant variables move materially. Define the allowed operating range; outside that range, require review rather than silently extrapolating.

2. Intensity baseline

Metrics such as kWh per tonne, kWh per unit or kWh per operating hour are easy to communicate. They can mislead when base loads are large, product mix changes or yield losses shift. Always show both total energy and the intensity indicator.

3. Regression or engineering model

A model may include production, weather, operating time or another justified variable. The proposal should disclose the equation, coefficient units, training period, excluded records and model diagnostics. No single statistical threshold is universally acceptable for every process; the acceptance criteria should reflect the decision risk and available data.

4. Retrofit-isolation measurement

For a defined compressor, chiller, motor system or process cell, direct measurement around the affected equipment can reduce interference from unrelated factory changes. The owner must still define interactions such as upstream transformer losses, cooling loads or production effects.

5. Whole-facility measurement

Whole-facility metering can capture the net outcome that finance cares about, but small project savings may be difficult to distinguish from normal facility variability. The US Department of Energy describes retrofit-isolation, whole-building and calibrated-simulation approaches in its public M&V guidance. The selected approach should match the project boundary and materiality, not a bidder’s preferred template.

Require a baseline data dictionary

Each input should have a controlled definition. “Production” could mean good units, gross units, tonnes, machine cycles or shipped product. “Operating hours” could mean scheduled hours, motor run hours or hours above a load threshold. A data dictionary should record:

• variable name, unit and physical meaning;
• source meter, historian, MES, BMS or manual record;
• timestamp basis and time zone;
• sampling interval and aggregation rule;
• responsible data owner;
• missing, duplicated and estimated-data flags;
• calibration or validation evidence;
• valid range and change-control process.

For electrical data, reconcile the meter hierarchy first. The separate guide on factory electrical metering boundaries explains why grid exchange, onsite generation, battery flows and submeter sums must not be double-counted.

Separate routine and non-routine adjustments

Routine adjustments cover variables expected to change regularly, such as production volume, operating hours or weather. Their treatment should be built into the approved model.

Non-routine adjustments address structural changes: a new production line, changed floor area, a major shutdown, a different product recipe, altered compressed-air pressure or a new onsite generator. The contract should define who raises the change, what evidence is required, who approves it and whether prior reports are restated.

A useful change log records the date, affected boundary, expected direction, calculation, attachments and approver. Without this log, discussions become retrospective negotiations rather than auditable M&V.

Decision matrix for comparing two proposals

Owner deliverables for the RFQ or contract

Ask each bidder to return a complete M&V package, not only a spreadsheet:

1. Boundary diagram and meter list.
2. Raw baseline dataset and data-quality report.
3. Baseline period selection note.
4. Variable dictionary and source ownership.
5. Model equation with coefficient units and valid range.
6. Routine and non-routine adjustment rules.
7. Worked example using owner-supplied test data.
8. Meter accuracy, CT/VT ratio and time-synchronization evidence where applicable.
9. Commissioning and data-continuity plan.
10. Monthly reporting template with energy, demand, cost and uncertainty kept separate.
11. Exception, change-control and dispute workflow.
12. Editable model plus version-controlled calculation record.

The broader electrical ITP guide is useful for assigning review, witness and hold points. For operational datasets and recurring evidence, the factory maintenance tender guide shows how to link scope, access and records.

Metering QA, commissioning and HSE

An M&V model cannot repair incorrect field measurements. Before the performance period, the commissioning plan should check meter identity, phase association, CT/VT ratio, polarity, multiplier, import/export direction, time synchronization, communications loss, interval alignment and reconciliation with upstream totals. Record test equipment, calibration status, expected result, actual result, witness and punch-list closure.

Any inspection or correction involving energized switchboards, CT secondary circuits or measurement wiring must be planned by competent authorized personnel. Apply the site electrical safety procedure, isolation, lockout/tagout, test-before-touch and arc-flash controls as applicable. Never open a CT secondary under load. This article does not provide a live-work method statement.

Commercial risks to put in the risk register

How this applies to solar, BESS and factory electrical projects

For rooftop solar, separate facility consumption, PV generation, export and curtailment. The rooftop solar feasibility data guide provides the owner-input context. For BESS, distinguish peak-demand reduction, energy shifting, backup readiness and conversion losses; the industrial BESS procurement brief explains how to state the AC boundary and duty.

For conventional electrical upgrades, do not claim energy savings merely from replacing equipment with a newer model. Record the pre- and post-condition, loading, operating schedule, control sequence and interactions. Reliability improvements and avoided downtime may be valuable, but they are separate from measured electrical energy savings.

Questions an owner should ask at bid clarification

• Which meter and exact energy flow define the savings boundary?
• Why is the proposed baseline period representative?
• Which variables materially affect consumption, and who owns those data?
• What happens when production leaves the model’s valid range?
• How are line additions, shutdowns and product-mix changes approved?
• Can the owner reproduce the result from exported raw data?
• Which part of the benefit is kWh, kW demand, tariff or avoided operating cost?
• What meter or data failure makes the result invalid?
• What evidence triggers acceptance or payment?
• Who retains the model, scripts and calculation history after handover?

FAQ

Should an owner always require 12 baseline months?

No universal duration fits every facility. The period should capture relevant production and seasonal variation while excluding unjustified structural changes. The bidder must explain why the selected period is representative.

Is kWh per unit enough?

It is useful but may not be sufficient. Fixed loads, product mix, yield, operating hours and weather can change the ratio. Show total energy and test whether the normalization has a defensible physical basis.

Can a guaranteed savings percentage be accepted before the baseline is frozen?

It can be treated as a commercial assumption, not a verified result. The contract should define how the baseline and acceptance equation are finalized before that percentage controls payment or performance liability.

Should avoided cost use the current electricity tariff?

Use an owner-approved tariff scenario and show it separately from energy savings. Demand charges, time-of-use periods, taxes, export rules and future tariff changes can alter financial results without changing measured kWh savings.

What if two bidders use different M&V approaches?

Require both to recalculate a common owner scenario, or score the approaches separately for boundary, data burden, uncertainty and commercial suitability. Do not rank only the largest percentage.

Conclusion

A credible energy-saving proposal lets the owner reproduce the result. Freeze the boundary, baseline data, variables, adjustment rules, reporting period and acceptance equation before comparing percentages. Then compare technical scope, data burden, uncertainty, risks and commercial terms on the same basis.

For a factory-specific review, prepare the single-line diagram, meter list, 12–24 months of interval energy data where available, production/shift records, equipment change log and proposed project scope. QuangAnhcons can use that owner-supplied information to structure a preliminary electrical interface and M&V requirement for tendering; final design, contractual guarantees and regulatory compliance remain subject to project-specific review and approval.

Primary references

• ISO 50006:2023 — guidance on EnPIs and energy baselines.
• ISO 50015:2014 — general M&V principles; ISO confirms the edition remained current after its 2025 review.
• EVO IPMVP Core Concepts 2022 — current public protocol resource.
• US DOE FEMP M&V options — public explanation of retrofit-isolation, whole-building and calibrated-simulation approaches.

No quick estimate yet. You can still submit a project quotation request.