How every number on this site is produced

The formulas

All-in effective rate   = total bill / kWh
Supply effective rate   = eligible supply charges / kWh
Delivery effective rate = (delivery + demand + fixed + riders) / kWh
Load factor             = kWh / (peak kW x hours in period)
Demand share            = demand charges / total bill

Billed demand           = max( period peak kW,
                               ratchet% x trailing peak kW,
                               contract minimum kW )

Tiered energy cost      = sum over blocks of ( kWh in block x block rate )
                          MARGINAL, never an average rate
Time-of-use energy cost = sum over periods of ( period kWh x period rate )
Power factor adjustment = billed kW x (threshold / actual PF), when below threshold

Estimating from building type and floor area

The cost estimator, the survey ranges on the industry pages, the typical-building table on each state page and the optional building comparison in the bill check all read one record: the U.S. Energy Information Administration's 2018 Commercial Buildings Energy Consumption Survey (CBECS) public-use microdata, 6,436 sampled buildings, each standing for many real ones through its survey weight. It is a screen for what a building like yours usually uses and pays. It is not a measurement of any building.

Intensity     = annual kWh / floor area, for each surveyed building that uses electricity
Range shown   = 25th to 75th percentile of the surveyed buildings of the chosen type,
                with the median, each building weighted by how many real buildings it stands for
Annual kWh    = floor area x intensity, at each of those percentiles
Price used    = the state commercial average for the 12 months to August 2026
                x size factor, or the price you enter if you enter one
Size factor   = what buildings of this size paid per kWh / what the average building paid,
                from the same survey, interpolated in log floor area between its size classes
Annual cost   = annual kWh x price used

• The survey data are checked against EIA's own published tables before anything is used. The ingest refuses to run unless the microdata reproduce EIA's published building count, floor area and total electricity, and, across 31 groups of its Table C14 (size, activity and census region), the average intensity, quartiles, price per kWh and the relative standard error of the average computed independently from the survey's 151 replicate weights. Agreement is close, with the worst average 2% from EIA's published figure, but not exact, because EIA lightly alters the public-use file to protect respondents.
• A building type with too few surveyed buildings gets no range. Under 30 surveyed buildings, as with refrigerated warehouses (23), the page says so instead of showing a figure a handful of buildings cannot support. A regional range is used only where the region holds at least 60 surveyed buildings of the type; otherwise the national range is read, and the result says which.
• Estimated use is flagged. EIA imputed or estimated rather than metered the electricity use of 27% of the buildings in the survey, and of every strip shopping centre. A type where 90% or more were estimated is marked indicative, because estimation narrows the spread between buildings.
• The size factor is a ratio, not a 2018 price. Only how much more or less each size of building paid per kWh, against the average building, is carried forward and applied to today's state price. Small buildings paid more in the survey. The survey records what was paid, not why.
• Where your own use falls is a band, not a rank. Five published percentiles cannot support a rank to the decimal, so a building is placed between them: below the lowest tenth, between the tenth and the quarter, and so on. The bill check annualises one month of kWh (times twelve), which is rough for a building with a strong summer or winter peak, and says so.
• The same building in every state holds the kWh fixed and changes only the price, so the comparison is of markets, not climates.
• It does not cover manufacturing or industrial buildings (outside the survey), buildings of 1,000 square feet or less, a tenant suite inside a larger building (the survey measures whole buildings, classed by what fills the most floor space), or any change in how buildings use electricity since 2018.

Rounding and precision

Computation runs at full precision. Rounding happens once, at display, using deterministic half-up rounding that is not subject to binary floating-point drift. We never display more precision than the inputs justify — an estimate derived from floor area is shown to two significant figures, never to the cent, because to the cent would be a false claim about accuracy, and accuracy is the only thing we are selling.

The proof ladder

Every figure we show you is labelled with its evidence level. These are not interchangeable and we never blur them together.

Controlled vocabulary

These words mean exactly one thing across the whole site, in copy, in proposals and in sales conversations.

Comparison rules we will not break

• An all-in bill rate is never compared with a supply-only quote. The engine refuses the comparison rather than producing a misleading number; it requires an explicit basis argument and throws on a mismatch.
• A benchmark is never presented as an available offer. Benchmarks are dated public context.
• Eligibility is never inferred from a ZIP code or city. It comes from the utility territory serving the specific service address, confirmed from the account identifier.
• No savings figure is published that is not computed from a specific customer’s own data, with the method shown and the assumptions named.

Known limitations, stated plainly

Our standard is 25 or more real bills across at least 5 utility territories, reproduced within 2 percent on the all-in total. The count above is read from the encoded fixtures at build time rather than typed here, so this page cannot claim more validation than exists. A bill outside tolerance is investigated as an engine bug or a missing tariff feature; the tolerance is never widened to make a test pass.

• State figures are averages, not prices. State commercial prices are U.S. Energy Information Administration averages (Form EIA-861 annual, final; Form EIA-861M monthly, preliminary), ingested by script from pinned source files and checked against EIA’s own printed prices. Each one blends every commercial customer in the state, so it is dated context for reading a bill, not a quote or an available offer.
• Utility prices are class averages. The utility tables on the state pages and the price-by-utility page are EIA’s Form EIA-861 sales to ultimate customers, final data: revenue per kWh averaged over every commercial customer of a utility and service type. Bundled and delivery-only prices are different parts of a bill and are never added or compared, a utility with larger customers averages a lower price per kWh, and retail power marketers and utilities with fewer than 500 commercial customers are left out.
• Tariff detail is published only for the utilities on the tariff index. For those 16 U.S. utilities, and the Alberta distributors in the Alberta guide, we quote schedule figures from the utility’s own filed tariff, dated and linked, never as a bill estimate, and a verifier finds every quoted figure in the document. A row is withdrawn after the last day we vouch for it. For every other utility, rate schedules, demand thresholds and ratchet terms are not published here, and nothing computes a bill from a tariff.
• Building estimates are ranges from a 2018 survey of whole buildings. They show what comparable surveyed buildings used and what that would cost at the state's average commercial price. They are not measurements of any building, they do not cover manufacturing or a suite inside a larger building.
• Market classification is a routing framework, not legal advice. It is built from public market structure and applied state by state. Whether a given account can choose a supplier is always confirmed from the account itself.
• The load factor bands are interpretive. They prioritise work; they are not a tariff rule.

We publish these gaps rather than filling them with plausible-looking numbers. A gap is recoverable. A wrong number attributed to us is not.

Corrections

When a published number is found to be wrong we correct it, date the correction, note it on the page, and tell anyone who received the wrong figure in an analysis. The corrections log is public.