A data-center announcement can put several impressive numbers beside one another: megawatts, investment dollars, jobs, gallons, and renewable-energy percentages. They may all be accurate while describing different projects, dates, boundaries, or kinds of commitment. Understanding the proposed facility requires joining the records behind those numbers, not merely collecting more headlines about artificial intelligence.
The national context is substantial but dated estimates must stay dated. In December 2024, the Department of Energy summarized a Berkeley Lab report estimating U.S. data-center electricity use at 176 terawatt-hours in 2023 and projecting 325–580 terawatt-hours in 2028. Those are historical estimates and forward scenarios from that report, not a meter reading for October 2026 or a forecast for any particular neighborhood. DOE's December 2024 report announcement
This article develops a site-level reading method. It does not investigate an actual local development, recommend an investment, or claim that a particular company concealed costs. Every project, invoice, meter series, contract, and numerical exercise introduced below is fictional unless expressly identified as a dated primary-source fact. The examples are deliberately constructed to expose different errors a reader could otherwise miss.
The central object is a fictional development called Parcel Q. Its records are not a simulated official approval. They are an analytical case file: a land-use drawing, utility correspondence, an environmental worksheet, proposed commercial terms, and later operating reports. We will ask which statements that file supports, which it contradicts, and which remain unanswered.
1. Identify the decision before interpreting the document
Minnesota's Pollution Control Agency provides one real example of divided responsibilities. Its data-center page describes environmental review and permits covering such matters as backup-generator air pollution, cooling-water and wastewater discharges, and fuel storage. It also says the agency does not decide whether a project is built. This is a description of Minnesota's process, not a universal rule for every state or an approval of our fictional project. MPCA data-center responsibilities
Parcel Q's file contains four documents. Drawing Q-01 shows space for three buildings. Utility letter Q-02 says a study will examine an initial electrical request. Water worksheet Q-03 estimates demand under a particular cooling design. Council resolution Q-04 approves a road agreement. None is called an operating report, and none includes a completed year's consumption.
A careless summary combines them into a claim that the council approved three operating buildings with guaranteed power and water. The documents do not establish that. The drawing establishes a proposed physical layout. The study letter establishes the scope of a study. The worksheet supplies assumptions. The road resolution records a decision about a road agreement.
The first useful product is therefore a decision map rather than a verdict. Beside each document, write the issuing body, the action it actually took, the project phase, and the conditions still visible in the record. A signed document can be authoritative for one narrow proposition without settling all the other questions a reader cares about.
Now add a fifth item: a later drawing Q-05 removes one building but keeps the project name. The name alone cannot identify which layout a subsequent water estimate used. The relevant link is the drawing version or another unambiguous description of the configuration. If the worksheet cites Q-01, it should not silently be presented as an estimate for Q-05.
This distinction also changes how missing information is described. If a public file lacks a final utility agreement, the supported statement is that the inspected file did not contain it. That is not automatically proof that no agreement exists anywhere. Conversely, a developer's statement that an agreement exists is not a substitute for examining the terms relevant to a public claim.
The result is a more useful question for a meeting: which exact decision established this claimed right or obligation, and does it apply to the same version of the project? That question can be answered with records rather than with competing adjectives about whether a development is innovative or dangerous.
2. Separate a power request from energy consumed
The Energy Information Administration distinguishes power at a moment from energy over a period: watts measure power, while watt-hours measure electricity used through time. The same distinction holds when the prefixes become kilo, mega, or tera. EIA's explanation of electricity measurement
Parcel Q requests a connection described as 120 megawatts. A later operating example, which we invent independently, averages 40 megawatts for a non-leap year of 8,760 hours. That produces 350,400 megawatt-hours. Multiplying the requested 120 megawatts by every hour instead gives 1,051,200 megawatt-hours. The second number is three times larger because it assumes continuous use at a different level; it is not another way of writing the first observation.
An article that treats the larger calculation as actual annual consumption has replaced a request with a utilization assumption. An article that treats the smaller observed average as proof that a 120-megawatt connection could never be needed makes a different leap. Neither the annual average nor the requested limit alone describes every operating interval.
Consider a simpler fictional day. The facility draws 60 megawatts for sixteen hours and 20 megawatts for eight hours. Its daily energy is 1,120 megawatt-hours, and its average draw is approximately 46.67 megawatts. Its peak in this deliberately simplified series is 60 megawatts. The three numbers describe the same day without being interchangeable.
The practical record question is whether a quoted megawatt figure is a design rating, an application, an agreed service limit, a measured interval peak, or an average inferred from energy. That classification belongs beside the number. Without it, even correct arithmetic may answer a question the underlying document never addressed.
Meter boundaries require similar care. Suppose an operating report measures electricity entering the entire campus, while a technical presentation reports only the server rooms. Their totals can differ without a faulty meter. To compare them, first identify which equipment and buildings fall inside each measurement. The comparison should not assume that every meter with the project name covers the same footprint.
For Parcel Q, our provisional conclusion is narrow: the application requests 120 megawatts, while a separate hypothetical operating series averages 40. The file needs an interval record and the applicable service terms before someone can characterize actual peaks or contractual exposure. There is no need to decide which number is the real one by discarding the other.
3. A project queue is not a list of additive commitments
Create a fictional utility spreadsheet with three entries: Parcel Q, 100 megawatts; Q North Campus, 100 megawatts; and Project R, 200 megawatts. Summing the visible rows gives 400 megawatts. An attached correspondence file reveals that Q North Campus is an alternative name for the same initial Parcel Q application. The unique proposed projects total 300 megawatts under those assumptions.
A second attachment says the initial signed service commitment for Parcel Q is only 40 megawatts, with later phases conditional. Project R has an 80-megawatt signed initial commitment. The signed initial total is therefore 120 megawatts. That number answers another question again. It is neither the gross row sum nor the unique eventual requested capacity.
This example does not suggest that any real utility counts duplicates. It demonstrates why a reader cannot infer a firm construction program from a spreadsheet of expressions of interest. The status field and entity links matter as much as the capacity column.
The same problem appears inside a single development. Imagine that Parcel Q's 100-megawatt master request includes a 40-megawatt first phase. Adding the master request and the phase creates 140 megawatts only by counting the first phase twice. A document register should identify parent projects, subprojects, alternatives, and replacements before attempting a regional total.
Dates can make an apparent contradiction disappear. The 100-megawatt request may be accurate as of January, and a 70-megawatt revised request accurate as of September. Neither should be treated as a second facility unless the record says it is one. For a time series, preserve both versions with their effective dates instead of mixing them into a single current snapshot.
There is also a difference between an unknown status and a canceled project. Suppose two rows lack a final answer from the applicant. Assigning both a zero probability of construction would be an unsupported judgment. Assigning both certainty would be equally unsupported. A public summary can retain an unresolved category rather than hide uncertainty in a total that looks precise.
A useful Parcel Q summary now has several separate lines: gross spreadsheet requests, deduplicated project requests, signed initial commitments, and unresolved later phases. Readers can see exactly where the figures diverge. They can also see which additional record would move a proposal from one category to another.
That is more informative than a dramatic announcement that hundreds of megawatts either appeared or vanished. In our case, some of the change came from naming, some from phase definitions, and some from the distinction between a request and a commitment. Those are different developments with different implications.
4. Efficiency ratios do not replace absolute totals
DOE's Federal Energy Management Program defines annual power usage effectiveness as total facility energy divided by IT-equipment energy. Its site water usage effectiveness divides annual site water use in liters by IT-equipment energy in kilowatt-hours. These are different ratios with different numerators. The page supplies definitions, not a current performance rating for Parcel Q. FEMP's data-center energy and water metrics
Give our fictional facility an initial annual ledger containing 100 million kilowatt-hours for IT equipment and 130 million for the whole facility. Its calculated energy ratio is 1.30. In a later invented year, the IT total rises to 200 million and the facility total to 240 million. The ratio falls to 1.20, while the absolute facility energy rises by approximately 84.62 percent.
Both statements are true in the example. The overhead relationship improved, and total energy increased. Calling the lower ratio proof of lower total consumption would omit the growth in the underlying workload and equipment footprint. Calling the increased total proof that the ratio deteriorated would ignore the actual division.
Add site water records of 40 million liters in the first year and 60 million in the second. Against the respective IT-energy totals, the ratios are 0.40 and 0.30 liters per kilowatt-hour. Absolute site water use increased by fifty percent while the reported ratio decreased by twenty-five percent.
The reader now has four meaningful observations rather than a single score: the two ratios improved, and the two absolute totals grew. Whether those changes are acceptable in a particular community requires evidence beyond these arithmetic relationships. The calculations alone do not establish a water shortage, an adequate supply, or the environmental value of the computing performed.
A further trap appears when two reports use different denominators. Dividing our first year's 40 million liters by total facility energy yields approximately 0.308 liters per kilowatt-hour, rather than 0.40. That is not a newly discovered efficiency improvement. It is a different calculation. A public table must identify which denominator it uses before comparing sites or years.
To preserve the evidence, keep the numerator and denominator beside every ratio. If only a rounded ratio is published, recovering exact totals may be impossible. Many combinations of water and energy can produce the same displayed value. Asking for those underlying totals is not an accusation; it is the information needed to understand the claimed performance.
5. Build a water balance before using the word consumption
USGS distinguishes water withdrawn for thermoelectric cooling from water consumed, including through evaporation. It describes once-through and recirculating systems and notes that recirculating systems need replacement water for several losses. The distinction matters when examining electricity-related water, but it does not establish which cooling arrangement a particular data center uses. USGS thermoelectric water-use explanation
For a separate fictional day at Parcel Q, assume the inspected records show one million liters delivered from a municipal system and another half-million delivered as reclaimed water. The campus receives 1.5 million liters in total. Its records also show 400,000 liters leaving through an identified wastewater outlet, one million evaporating, and a 100,000-liter increase in stored water. Those values balance exactly under our deliberately complete assumptions.
Subtracting wastewater from incoming water gives 1.1 million liters. Calling all of that evaporation would be wrong here because 100,000 liters remained in storage. Conversely, adding the stored water to the wastewater and calling both a return to the environment would describe an event that did not occur in the example.
The exercise also distinguishes the source of a delivery from its eventual disposition. Reclaimed water still enters our fictional campus water balance. Its availability and alternative uses would need their own local evidence. The label alone does not make a quantity physically disappear, nor does it tell us how much drinking-water demand the arrangement avoided compared with a specific alternative design.
Suppose a promotional summary reports only the one million liters from the municipal system. That can accurately describe that supply stream, but it is not the total input in our case. A critic who adds 1.5 million liters of site deliveries to the municipal supplier's one-million-liter delivery entry would make the opposite mistake, counting the same municipal flow twice.
The useful check is to draw a boundary and follow each stream across it once. For this exercise, the boundary is the campus fence over one specified day. A supplier's system-wide withdrawal, the campus delivery meter, and a wastewater outlet are observations at different locations. They should not be added merely because they all contain the project's name.
If one flow is missing, retain an unexplained residual instead of forcing the balance to close. A residual can arise from incomplete records, mismatched periods, meter problems, or an omitted flow. Without further evidence, it is not a diagnosis of which cause occurred. A transparent unresolved number is preferable to a fabricated explanation.
6. Annual water totals conceal the days that test a system
Invent two possible annual operating profiles, each totaling 395 million liters over 365 days. Profile A uses one million liters on 355 days and four million on ten days. Profile B uses approximately 1.082 million liters every day. Their annual totals match, but Profile A's peak day is nearly 3.70 times Profile B's constant daily quantity.
These profiles are not forecasts of actual cooling equipment. They are a controlled comparison showing what an annual total omits. A statement about annual volume cannot, by itself, answer a question about peak-day delivery. A statement about peak delivery cannot determine the annual volume without a time profile.
Now give our fictional water agreement a documented daily limit of three million liters. Profile A exceeds that invented limit on ten days; Profile B does not. This is a conclusion about our assumed contract and data, not about an actual water right or permit. If the real agreement instead used an hourly limit, a monthly allocation, or an emergency exception, the comparison would need to change.
A further scenario exposes a different problem. Suppose the ten high-use days fall in the same period as a hypothetical supply restriction. The annual total still has not changed, but the overlap becomes central to the project question. To evaluate that overlap in reality, the records would need dates, available supply, applicable restrictions, and the facility's permitted response. Annual averages cannot supply those missing facts.
This is where a request for public information can become precise without demanding proprietary server details. A reader can ask for the time resolution of the published water estimate, the assumed peak, the conditions associated with it, and the document that authorizes the proposed supply. Those questions concern the project as proposed; they do not require an invented claim about the operator's internal workload.
Parcel Q's annual worksheet is therefore not rejected as useless. It answers an annual-volume question. We add a separate peak-day question and record that it remains unresolved until a compatible time series or design estimate appears. The problem was not that the worksheet lied. It was that an annual answer had been stretched to cover a daily question.
The same discipline applies when comparing alternatives. If one design is represented by its hottest day and another by its annual average, the apparent difference may be mostly a mismatch in measurement periods. A fair comparison specifies the same period and operating assumptions before drawing a conclusion about either design.
7. Keep off-site estimates separate from local water records
Consider another explicitly hypothetical calculation. Parcel Q has 130 million kilowatt-hours of purchased electricity in a year, and an analyst assumes an off-site water-consumption factor of half a liter per kilowatt-hour for a defined electricity-supply scenario. The resulting estimate is 65 million liters. This factor is invented for the exercise; it is not a measured regional average or a claim about any current grid.
Assume a separate, compatible site record establishes 30 million liters of direct consumption for that same year. Adding the two gives 95 million liters within the analyst's stated combined boundary. That does not mean the local water utility delivered 95 million liters to the campus. The off-site estimate describes a different location and method of calculation.
The difference matters to a neighborhood question. A combined footprint can be useful for one analysis, while a local intake or delivery record answers another. Presenting the combined number as a local withdrawal would misplace the off-site quantity. Presenting the local number as a complete electricity-related footprint would omit the assumed off-site component.
Test the sensitivity rather than conceal it. If the invented factor is changed to 0.2 liters per kilowatt-hour, the off-site estimate becomes 26 million liters and the combined total becomes 56 million. At 0.8, the corresponding results are 104 million and 134 million. The change comes entirely from an assumption we varied, not from a new site measurement.
Those endpoints are not a statistical confidence interval. We have supplied no probability model and no evidence that the chosen range captures actual uncertainty. They are scenario results. A graph should label them accordingly, especially if readers might otherwise mistake a shaded band for a measured error bar.
To turn this into a real analysis, the missing evidence would include the source and year of the electricity factor, its geographic and technology coverage, and whether it represents consumption or another water quantity. It would also need compatible treatment of imports, time periods, and any electricity already counted elsewhere in the boundary. Without that information, a precise multiplication can still rest on an undefined input.
For Parcel Q, the publication-ready result of this exercise is therefore conditional: under the explicitly stated half-liter assumption, the combined calculation is 95 million liters. It is not an assertion that a real campus used that amount. Keeping the conditional language attached to the result protects both the usefulness of the calculation and the limits of the evidence.
8. Cost allocation begins with a ledger, not a slogan
On June 18, 2026, FERC announced show-cause orders requiring six regional grid operators and relevant transmission owners to justify or reform large-load tariff provisions. Its identified issues included cost shifting, transmission-cost transparency, co-location, and flexible loads. That was a specific procedural action; it should not be rewritten as proof that every later regional rule was finalized or that every data-center cost is now assigned in the same way. FERC's June 2026 large-load action
Build a fictional project ledger with a 100-million-dollar infrastructure cost. The developer contributes 40 million upfront, leaving 60 million to be recovered under an assumed agreement. For simplicity, suppose the agreement provides six million dollars of principal recovery annually for ten years. This toy ledger excludes financing costs, taxes, operating expenses, and discounting. It is not a utility rate model.
After five annual payments, thirty million of that original remaining principal has been recovered and thirty million remains. Now imagine the project closes. A separate fifteen-million-dollar security instrument is fully collectible under our assumptions, and the equipment has ten million dollars of recoverable value. The residual is five million dollars, before any omitted costs or other contractual recovery.
The headline that the developer paid forty percent of construction is true but does not answer the closure question. The statement that the agreement contemplated full recovery is also true in our setup but does not establish that the scheduled future payments arrived. The closure scenario depends on the remaining balance, collectible security, reusable asset value, and the terms governing any residual.
We cannot assign that five-million-dollar residual to households merely by observing that it exists. The payer would depend on the applicable agreement and regulatory treatment, which our toy ledger has not specified. Nor can we announce that households are protected simply because a contract mentioned security. We assumed full collection to make the arithmetic explicit; a real evaluation would need evidence about enforceability and the available amount.
The original exercise clarifies what a public claim should identify. If someone says the developer pays all costs, ask which costs, over what period, with what treatment if the project changes. If someone says other customers pay, ask for the allocation record and the particular charge. An infrastructure price tag alone does not establish either conclusion.
This method leaves room for legitimate disagreement about policy. It simply requires the disagreement to attach to a defined balance and a proposed payer rather than to an undefined promise that everything is covered or an unsupported claim that every dollar will be socialized.
9. A demand charge is not an energy-meter reading
Ohio Power's public data-center tariff guide identifies a July 2025 effective tariff and an October 2026 study tranche. Its minimum-demand discussion includes a historical billing-demand comparison and a capacity schedule, rather than one flat percentage applied to every quantity in a press release. The guide points readers to Schedule DCT. This is a real example of why the applicable terms must be inspected, not a legal interpretation of an individual customer's bill. Ohio Power's data-center tariff guide
To avoid substituting a simplified calculation for that actual tariff, use a wholly invented Parcel Q contract. It sets a 50-megawatt contract quantity and, for this example only, a minimum billable demand of eighty percent of that quantity. The minimum is 40 megawatts, or 40,000 kilowatts. At an invented ten dollars per kilowatt per month, that component is 400,000 dollars.
Separately, suppose the month contains ten million kilowatt-hours of metered electricity charged at an invented six cents per kilowatt-hour. The energy component is 600,000 dollars. The two stated components total one million dollars. We have deliberately excluded all other charges and adjustments; this is not a quoted rate or a prediction of a real bill.
The minimum-demand component does not establish that the facility consumed forty megawatts continuously. It is a billing input under the assumed contract. Conversely, knowing the month's energy quantity does not reveal which minimum-demand clause applied. The calculation needs two distinct quantities and two distinct prices.
Change only the monthly energy to eight million kilowatt-hours. The energy component falls to 480,000 dollars, but our stipulated minimum-demand component remains 400,000. The two-component total becomes 880,000 dollars, a twelve-percent reduction rather than the twenty-percent reduction in energy. The difference follows from the deliberately fixed component, not from a billing error.
A real analysis would inspect the actual tariff version, contract, billing period, and charge descriptions before constructing such a comparison. A public-facing explanation can illustrate the distinction while refusing to pretend it has reproduced an actual bill. That restraint is especially important when a broadly worded promise to pay for power is being used to answer a narrower question about infrastructure recovery.
For Parcel Q, the next evidence request is now concrete: provide the clause defining billable demand and identify which construction costs that revenue is intended to recover. Without those links, neither a consumption estimate nor a hypothetical monthly payment proves who ultimately carries the project's financial risk.
10. Annual renewable matching and time-matched supply answer different questions
EPA explains that unbundled retail renewable energy certificates represent attributes of renewable generation and can be purchased separately from electricity. The certificates themselves do not physically deliver electricity to the buyer. EPA's page also describes generation-period and other certificate attributes. Those distinctions are relevant to a renewable-energy claim, but do not determine the reliability of a particular campus connection. EPA's explanation of retail renewable energy certificates
Create an intentionally simple two-period example. Parcel Q consumes 500 megawatt-hours in Period One and another 500 in Period Two. A specified renewable generator produces 900 megawatt-hours in the first period and 100 in the second. The annual totals both equal 1,000, assuming these are the year's only two reporting periods.
Under a separate, deliberately defined same-period matching test, only 500 of the first period's generation matches that period's load, and only 100 matches the second period's load. The matched total is 600 megawatt-hours, or sixty percent. The remaining first-period generation does not erase the second-period shortfall under this particular test, which includes no storage or transfer between periods.
This exercise does not pronounce an annual renewable claim fraudulent. It shows that annual equality and our stricter timing test answer different questions. Actual claims also require the relevant ownership, retirement, eligibility, and accounting evidence. The example does not establish any company's compliance with a reporting standard.
Now consider what the public record would need to evaluate a statement that the campus has continuous dedicated renewable supply. Two annual totals cannot establish that proposition. Our invented dataset is a counterexample: equal annual quantities coexist with a mismatch between periods. A reviewer needs evidence at the resolution of the claim rather than a larger total that happens to share the same unit.
A separate reliability claim needs still other evidence. Nothing in our two-period arithmetic establishes available backup, transmission capability, or how a disruption would be handled. A certificate record, a generation record, and a service agreement may all be useful, but they are not interchangeable exhibits.
For Parcel Q, the proper summary distinguishes an annual quantity comparison from a timing comparison and from a physical-service claim. This is not a choice between praising or dismissing renewable procurement. It is a way to describe exactly what the supplied records demonstrate, so that the claimed benefit does not expand beyond its evidence.
11. Backup equipment needs an operating record as well as a nameplate
Return to the fictional site drawing. It lists twenty generator units rated at three megawatts each. The sum of their ratings is sixty megawatts. That number describes the equipment list in our example; it does not establish how many units operated together, their loading, their runtime, or their emissions during a particular month.
An invented test log says five units each ran at three megawatts for two hours. The calculated output is thirty megawatt-hours. Multiplying the full site's sixty-megawatt rating by those same two hours would produce 120 megawatt-hours, four times the logged exercise, because it silently assumes all twenty units operated.
The reverse mistake is possible too. If a public report supplies only total runtime across all engines, it may not establish the number of hours during which the neighborhood experienced simultaneous operation. Ten engine-hours can mean ten engines for one hour or one engine for ten hours. Those arrangements differ even though the summed engine-hours agree.
To make the distinction concrete, invent two schedules. In Schedule A, ten units run together from noon to one. In Schedule B, one unit runs during each of ten successive hours. Both contain ten engine-hours. Under a stipulated identical output of three megawatts per running engine, both produce thirty megawatt-hours, but their simultaneous output is thirty megawatts and three megawatts respectively.
These schedules do not predict noise, air quality, or compliance. Those questions require additional parameters and the relevant applicable conditions. The point is that total energy or engine-hours alone cannot reproduce the temporal pattern. A reader who wants to discuss that pattern should ask for the operating interval evidence rather than reverse-engineer it from an annual sum.
Parcel Q's file also contains a draft operating limit and a later issued document. Before comparing the test log with a limit, the reader must identify which version applied and how it defined the counted activity. A numerical difference from a draft is not automatically a violation of the issued condition. Equally, calling an activity a test does not, by itself, establish that it falls outside whatever conditions actually govern it.
A responsible public account can state the listed capacity, the documented operation, and the unresolved compliance question separately. It should not convert a large equipment rating into an invented pollution event, or convert a short test log into a blanket assurance that no relevant impact could occur.
12. Freeze a common project version before combining conclusions
Imagine that Parcel Q's record set changes halfway through review. Layout Version A contains three buildings. Layout Version B contains two. The initial water estimate describes Version A, the later electricity study describes Version B, and an economic presentation still uses the original construction schedule. All three files are authentic within the fictional case, yet combining them creates a project that no document actually describes.
The remedy is not to select the smallest number from every file and call that the revised proposal. Nor is it to select the largest and call that the worst case. First create a version crosswalk identifying the scope each figure covers. A comparison can then mark a value as compatible, obsolete for the current layout, or unresolved.
Suppose Version A's three equal buildings were assigned a combined annual water estimate of ninety million liters. It is tempting to multiply by two-thirds and assign Version B sixty million. That calculation would be valid only under the additional assumption that the relevant water use scales proportionally and all other design conditions remain fixed. The smaller building count alone does not establish those conditions.
If the revised worksheet instead supplies seventy million liters, the arithmetic difference is not proof of an error. Perhaps the configuration changed in ways our simplified scaling omitted. The appropriate response is to inspect the revised assumptions, not to overwrite the authored estimate with an unsupported proportional guess. If the assumptions remain unavailable, report that limitation.
This version problem can affect promises as well as impacts. An economic presentation might count 1,200 construction job-years while a later summary calls the same number permanent jobs. In a separate invented example, 400 full-time-equivalent construction positions maintained for three years produce 1,200 job-years. That unit does not mean 1,200 permanent positions after opening. The original measure and period should remain attached to the number.
For the final Parcel Q briefing, we can now classify six proposed headlines. The 120-megawatt application is supported as a request, not annual energy. The claim of declining absolute water use is contradicted by our annual ledger despite a better ratio. The annual renewable quantity matches, but our timing test reaches only sixty percent. The hypothetical closure ledger has a five-million-dollar residual with no payer specified. Backup operation requires interval evidence. The revised project's combined resource estimate remains unresolved until the documents refer to the same layout.
Those are substantive findings, not a refusal to reach conclusions. Some claims pass, some fail, and some need a particular missing record. The method makes that distinction visible without manufacturing a local controversy, treating a proposal as an operating facility, or accepting a numerical slogan as a complete account of what the project will require.