Appendix: baseline and verification worksheets

Use this appendix to assemble an applied evidence packet: service reconciliation, observation eligibility, tariff comparison and adjustment design. The instruments request data and tests; they contain no populated project answers. The fictional Aster Works names, deadlines and stop authority remain in The committee needs balanced evidence. The retained numerical reference identifies the teaching calculation, while these worksheets investigate the field propositions needed for applied use.

Start the packet with a service coordinate rather than an energy total. Define accepted output, its quality condition and the relevant operating period, then map each carrier and meter to that service. Energy per accepted unit is a useful decomposition only when numerator and denominator share a perimeter and time window. Lower total consumption alongside proportionately lower accepted output can leave specific consumption unchanged. Counting rejected output as accepted would create a different apparent result. Record both total energy and the service quantity so that a change in demand can be decomposed into changed activity and changed input requirement.

Make the affected-load map concrete enough to test an overall percentage. Each measure row identifies the loads it acts on, the supporting loads it can increase and any new load that arrives before or after it. Compare the sum of the mapped loads with the meter boundary. An unmapped residual is a question to investigate, not a load silently assigned the package percentage. If service depends on a supporting system outside the proposed meter, record that system in the interaction perimeter. This packet tests whether the annual model's scope represents the intervention before refining its parameter precision.

The observation packet separates raw data, eligibility rules and comparison variables. Retain raw readings without overwriting them; create a separate eligibility view that marks outages, missing intervals, unusual operation and changed meter boundaries. For every excluded interval, retain the reason and the proposed treatment. A useful comparison can then be reproduced with and without a disputed eligibility choice. The audit object is the difference those choices make to the proposed saving, rather than a general statement that the data has been cleaned. A missing interval remains an unknown quantity until its treatment is supported.

Design the comparison before reviewing performance. Identify the operating variables that plausibly change the reference input requirement, specify the domain over which their relation will be used and distinguish routine variation from a structural boundary change. A proposed adjustment based on production can be tested on retained reference periods that were not used to choose the relation. Ask whether its errors still track production, operating mode or time. A pattern tied to one of those variables suggests a missing mechanism or an inappropriate comparison domain.

Matched operating periods provide a useful challenge to the adjustment design. Compare like service and mode where the reference and project records overlap, then identify the part of the proposed claim that requires extrapolation beyond that overlap. If the intervention runs in a new mode, a historical relation can be precisely fitted and still lack support there. The worksheet should name the unsupported domain explicitly. Restricting the initial claim to supported operation, obtaining further evidence or using a suitable physical model are distinct responses. Choosing the most favourable extrapolation without an operating explanation would not resolve the gap.

Data gaps can also be challenged through bounds. Split an energy total into retained observations and the missing portion. If an accepted operating envelope supports lower and upper energy bounds for the missing intervals, carry those bounds through the same reference and project comparison. Do not create bounds from a convenient average merely because it preserves the original saving. Jointly affected gaps need a stated dependence treatment: common clock loss or a shared meter error should not be treated as unrelated uncertainties by default. The resulting range exposes whether a conclusion depends on an unobserved part of the year.

Build the commercial packet as a bill comparison. Map each physical quantity to the charge it changes and retain a separate list of fixed, minimum, demand-related and contingent payments. For a flat unit price, the responsive energy quantity has a direct bill relation. For a minimum purchase or delivery obligation, used energy and paid energy can differ. Annotate the record with the charging quantity, the price basis, the persistence condition and any overlap with listed capital or continuing expenditure. A quoted unit rate without those terms cannot identify the annual cash release.

The retained surplus exhibit gives a concrete entry point for a symbolic comparison. Let H be an additional constant annual payment for unused supply and K an omitted upfront interface cost. Hold the physical package, other costs, horizon and discount convention fixed. Applied NPV would then be screening NPV less the annuity factor times H, less K. Under a nonnegative-NPV rule, the admissible H is no greater than screening NPV less K, divided by the annuity factor. If that upper bound is negative, no nonnegative H can satisfy the rule. The applicable bill basis supplies the terms for this algebraic test.

If the price applies to all nominal supply, the extra unused-supply payment in that comparison is unused quantity times the applicable unit price, provided the same charge has not already been included elsewhere. A contract may have another quantity basis, credits or timing, in which case its own bill relation replaces that expression. The original annual engine charges only used energy and gives the excess no value. Its physical result cannot answer the charging question. The exhibit preserves the held proposition at reliance, while the symbolic test identifies the particular commercial term capable of consuming the value margin.

Exact decimal subtraction and credible measurement answer different questions. Let D be the estimated adjusted reference energy minus project energy, with service and periods reconciled. Suppose the receiving method can justify conservative absolute error bounds for the adjusted reference and project quantities, including adjustment error. Adding those bounds gives a worst-case bound U for their difference. The admissible energy difference lies between D minus U and D plus U. A positive lower bound supports the sign under those stated bounds; an interval spanning zero cannot settle it. No probability or statistical confidence follows from this deterministic construction.

Record where each bound comes from before using that sign test. Instrument specifications, missing-data treatment and model error can have different units and periods; convert them to the same energy comparison before combining them. A bound that excludes a material uncertainty is not conservative for the whole claim. If dependence is known, a justified joint bound can replace the simple sum, but an assumed cancellation should not. The small-positive computational study remains useful: its ratio is mathematically defined even when this applied evidence packet could not establish an observable reduction of that size.

The same discipline applies when transferring an energy interval into an emissions statement. Factors and boundary choices can add another uncertainty or change the proposition entirely. Retain the energy interval separately, name the factor basis and identify whether the intended statement concerns electricity attribution or a wider intervention. Do not select a factor merely because it makes the lower bound favourable. A cash conclusion can meanwhile remain meaningful under a supported bill comparison. The final evidence packet should state which proposition has passed its own test, avoiding a single accepted label that conflates service, savings, cash and emissions.