Methodology — Water heaters

Why published running costs disagree

Search results for a water heater's running cost routinely disagree by roughly 5× on the same class of unit -- published figures from $180 to $875 a year for comparable units. The usual method is watts × hours a day × the tariff: no efficiency figure, no usage assumption stated, and for a heat pump unit it is physically wrong, because a heat pump moves several units of heat energy per unit of electricity it draws rather than converting electricity to heat one-for-one.

This site's number comes from the unit's own filed efficiency and a filed usage assumption, not an invented one. The rest of this page is that formula, written out, so it can be checked rather than trusted.

The formula

Uniform Energy Factor (UEF) -- the ratio of hot water delivered to energy used -- is filed against one of four standard daily usage profiles a manufacturer tests to, called a draw pattern. That test standard is 10 CFR 430 Subpart B Appendix E (Appendix E to Subpart B of Part 430, Title 10, effective 2025-02-05), and its own numbers are what the annual figure below is built from -- not an assumption this site adds.

The four draw patterns Appendix E to Subpart B of Part 430, Title 10 defines

Draw pattern	Daily hot water (gal)	First draw (gal)
Very Small	10	2
Low	38	15
Medium	55	15
High	84	27

Every test also assumes a 67°F temperature rise (inlet 58°F to outlet 125°F). From those two filed numbers, annual energy use is:

annual_kWh = (daily_gal × 8.24 lb/gal × 67°F × 365) / (3,412 BTU/kWh × UEF)

8.24 lb/gal is the density of water at the standard's own outlet temperature, from Appendix E to Subpart B of Part 430, Title 10; 3,412 is the fixed BTU-to-kWh conversion. Worked example: a High-draw unit filing UEF 3.75 -- (84 × 8.24 × 67 × 365) / (3,412 × 3.75) = 1,323 kWh/year. At 13.4¢/kWh (a real residential rate, see the tariff rule below), that is $177/year. A model page on this site shows this same triplet -- result, filed inputs, tariff -- for its own numbers.

Which tariff, and why

Given a ZIP code, this site resolves the utility company that actually serves it (via the EPA's own predominant-utility determination for that ZIP, not a guess) and reads that company's own residential tariff from the OpenEI URDB. Where a ZIP returns more than one active tariff, the flat or tiered rate marked as the company's default is used; time-of-use tariffs are out of scope for this category in this phase, so a rate that only applies for part of the day is not selected. The worked example above used Duke Energy Carolinas' "RS (Residential Service)" tariff. The tariff a page's figure is built from is always named on that page -- "13.4¢/kWh" alone is not a claim anyone could check; a named company and rate is.

Without a ZIP code, a page falls back to the state average from the US Energy Information Administration's retail sales data, labelled "state average" rather than presented as if it were a specific rate.

Typical installed cost -- and its limits

No energy registry publishes what installation costs. The installed-cost range shown on a model page comes from a different, still-official source: the National Residential Efficiency Measures Database (REMDB) (REMDB, version 4.0.0), a regression the US DOE's own national lab fit against retail and distributor pricing, keyed to the same UEF and tank size a model already files. It is a calibrated range, not a quote, and it is never used as an input to a 10-year cost total -- a price range and a running-cost total answer different questions and this site does not blend them.

Cost range derived from the NLR National Residential Efficiency Measures Database (REMDB) v4.0.0, retrieved 2026-09-07, escalated from its native 2023 dollars to the current year in two separate pieces -- equipment cost by the Bureau of Labor Statistics' Producer Price Index for heating equipment manufacturing, installation labor by the BLS Employment Cost Index for construction wages -- because an installation is a labor cost, and pricing it with a manufacturer price index would be the wrong instrument.

REMDB's own calibration range, by class and metric

Class	Metric	Calibrated range
Electric Instantaneous	UEF	0.84–0.998
Electric Storage	UEF	0.92–0.95
Electric Storage	Nominal volume	27–55
HP Tank	UEF	3–4.07
HP Tank	Nominal volume	40–80
HP Tank w/ Gas Capping	UEF	3–4.07
HP Tank w/ Gas Capping	Nominal volume	40–80
HP Tank w/ Gas Capping and 240V Wiring	UEF	3–4.07
HP Tank w/ Gas Capping and 240V Wiring	Nominal volume	40–80
Gas Inst	UEF	0.81–0.99
Gas Storage	UEF	0.58–0.75
Gas Storage	Nominal volume	38–98

A model outside its class's calibrated range does not get an extrapolated price range: the page says the range is not published for that model and why, rather than guessing past where the regression is calibrated.

REMDB's own published reasons an actual quote will vary from this range:

• Prevailing local wages
• Drive time
• Local code requirements
• Access
• Moisture issues present
• Condition of existing flue
• Need to bring in combustion air
• Condition of existing electrical system
• Condition of existing plumbing
• Condition of existing gas line
• Presence of hazardous materials
• Type of equipment

Where our own sources disagree with each other

This site's installed-cost range does not match two other official sources that also price installation, and that disagreement is published rather than hidden. Checked against the DOE's own rulemaking life-cycle-cost model (escalated to the same current year, same two-part method), the gap runs from 33% (Gas Storage) to 92% (Electric Storage). The reason is not an error in either source: REMDB is a regression against retail and distributor pricing for a retrofit replacement; the DOE rulemaking figure is a regulatory-impact estimate blended across replacement and new-construction installs nationally. Both call the result "installed cost"; they are pricing different things. A ranking or a total built from either number alone would inherit whichever framing it borrowed without saying so -- this page says so instead.

Ranking weights

Not published yet. A ranked list for this category needs a sorting metric that is measured to actually separate models, not assumed to; that sorting engine has not shipped for this category. When it does, the weights it uses and their relative importance will be published on this page, linked directly from the ranked list itself, not left in a footer.

---