Method

Every number on this site comes from two public federal datasets and some arithmetic. This page says which, how, and — the part that matters more — what we will not calculate because the data does not support it.

Sources

Certified heat pump data: U.S. EPA, ENERGY STAR Certified Ducted Heat Pumps. Retrieved 2026-09-26. Most recent certification date in the file: 2026-09-23. 266,456 rows.
Electricity prices: U.S. EIA, Form EIA-861M, residential sector average by state, period 2026-07.
Heating degree days: U.S. EIA, State Energy Data System, series ZWHDP, period 2024. Used only to scale the optional worked example.
Test procedure definitions: AHRI Standard 210/240-2023 and its Appendix M1, for what HSPF2 and SEER2 measure.

Both federal datasets are U.S. government works in the public domain. This site is independent and is not affiliated with, or endorsed by, the EPA, the EIA, AHRI or any manufacturer.

How the catalogue is grouped

The EPA publishes one row per certified pairing: an outdoor unit with a specific indoor model. We group those 266,456 rows into 5,521 outdoor models by brand and outdoor model number, merging different spellings of the same brand and duplicate filings of the same model by two ENERGY STAR partners. For each model we count the distinct certified HSPF2 values and how many certificates carry each one.

Two deliberate cleanups happen in that grouping, and both change the totals:

Brand spellings are merged. The source writes some brands two ways ("Comfort Aire" and "Comfort-Aire"). We treat these as one brand. Three longer variants are merged too, each filed by the same ENERGY STAR partner and sharing model numbers: "Bryant" with "BRYANT HEATING AND COOLING SYSTEMS", "Payne" with "PAYNE HEATING AND COOLING", and "Innova Srl" with "Innova". Other similar names are kept apart unless the data shows they are the same brand.
Duplicate partner filings are merged. The same brand and model sometimes appears twice because two ENERGY STAR partners filed certificates for it. For someone holding the equipment that is one product, so we merge them and keep the partner on each individual pairing.

Why the URLs look the way they do

Each model is identified by its brand and model number, as /heat-pumps/trane/5twr6024a1/. This is not cosmetic. An earlier version of this site identified models by their position in a sorted list, which meant that adding one new model to the catalogue would shift every identifier after it, and an old link would quietly start pointing at different equipment. Brand and model number come from the data itself and do not move when the catalogue grows.

Calculator links made with that earlier version, on 23 and 24 September 2026, carry a number instead (?model=5320). Those numbers only mean something against the catalogue as it was then, so they are never read against today's list. They go through a frozen table built from that version: 5,707 of the 5,716 numbers open the same model today. The other 9 point to model numbers the EPA has since rewritten (most had a trailing * wildcard) or to a model that has left the list. For those, the calculator says which model the link was for and lets you choose, rather than loading something that merely looks similar.

One exception, for a dull technical reason. A handful of words — aux,con, nul and a few others — are reserved device names on Windows and cannot be used as a folder name, which breaks the build and the version control on a Windows machine. Where a brand or model slug would land on one of those, we append -brand or -model. Today that affects exactly one brand, AUX, which lives at /heat-pumps/aux-brand/.

The arithmetic

HSPF2 is BTU of heat delivered per watt-hour consumed, so:

kWh = heat in MMBtu × 1,000 ÷ HSPF2

cost = kWh × rate in $/kWh

Reference case, which the test suite asserts: 10 MMBtu at HSPF2 10 requires 1,000 kWh and costs $200 at 20¢/kWh. Worked through in full.

Describing a model with several ratings

44.3% of outdoor models carry more than one certified HSPF2. Where that happens we show the whole distribution rather than one number, and the headline figure is a range across every rating the unit is certified at.

"Most frequent" means the rating carried by the most certified pairings. It is a count of certificates. It is not evidence about what is sold or installed, because nobody publishes that.
Ties are reported as ties. On 519 models (21.2% of those with several ratings) two or more ratings carry the same number of certificates, and on 427 of those no rating appears more often than any other at all. In that case we say there is no typical rating rather than naming one.
The 10th–90th percentile is weighted by certificate count. It describes how the catalogue is distributed. It is not a probability that your system carries a rating in that band, and it is not a confidence interval.
Bar widths are the share of the total, matching the printed percentage.

How the 16 model pages were chosen

These are models where the indoor coil on the quote moves the running cost more than almost anywhere else in the catalogue, so they are the ones where looking up the exact pairing is worth the trouble. Selection is made from the data only. No search volume, ranking or popularity figure is used, because none is available to us.

The filter, applied to the catalogue and nothing else:

At least 50 certified pairings, so a distribution means something.
At least 3 distinct HSPF2 ratings, so there is something to explain.
Ranked by annual cost difference between the lowest and highest certified HSPF2, at 40 MMBtu of delivered heat and 17.11 cents/kWh.
At most 2 per brand.
One outdoor model number can be sold under several brands, and two different model numbers can carry an identical set of certified ratings, capacities and SEER2 values. Both cases are deduplicated, so these are different machines rather than several badges on fewer machines.

A model can also get a page when Search Console shows people searching for its model number, provided it is certified at two or more HSPF2 ratings, so there is something to explain. Each of those is listed with its evidence in the site's repository.

Once a model has a page it keeps it for as long as it stays on the ENERGY STAR list, even if a later data refresh would rank it lower, so that links to it do not break. New pages are added only up to the target.

We do not use estimated search volume, keyword difficulty or popularity data to choose pages, because we do not have any and we are not going to guess at it. The only demand signal used is the one described above: searches for a model number that actually reached this site, as reported by Google Search Console.

How brand pages were chosen

A brand gets a page if it has at least 20 outdoor models and 1,000 certified pairings — enough data to describe a range rather than restate a single certificate — or if one of its models has a page of its own, so that every model page sits under its brand. That is 39 of 194 brands. The rest are listed with their figures in the directory and are fully searchable in the calculator. We are not generating a page per brand for its own sake.

As with model pages, a brand that has a page keeps it while it stays on the ENERGY STAR list, even if a refresh leaves it under the threshold. Every page this site has published either still works or redirects to the closest page that does, and the site will not build otherwise. Calculator links are different, because they name a model inside the address: a model that leaves the ENERGY STAR list cannot be shown, and the calculator says so instead of showing another. Links from before 24 September 2026 go through the translation described above.

The badge-engineering count

Counts distinct outdoor model numbers that appear under more than one brand name in the ENERGY STAR certified list. Each brand files its own AHRI certificate - no certificate number is shared - so the model number is what shows the equipment is the same.

An identical model number under two brands is strong evidence of the same equipment, but this data cannot prove identical manufacture, nor that the warranty, the dealer network or the installed price are the same.

The full write-up, with the numbers.

What this site will not estimate

A "typical" annual heat demand. The calculator's demand field starts empty and stays empty unless you fill it. Published household aggregates measure energyconsumed for space heating across homes with gas, electric resistance and heat pump systems mixed together. That cannot be converted into heat delivered by applying any single efficiency, so putting such a figure in a field that means delivered heat would be putting the wrong number in the box.

The optional example is a round hypothetical — 40 MMBtu in a 4,000 heating-degree-day climate, scaled by your state's degree days and rounded to the nearest 5 — and it is labelled as an illustration everywhere it appears.

Your bill. HSPF2 is a laboratory measurement at reference conditions. Sizing, duct leakage, commissioning and how you run the thermostat all move real consumption, and none of them is in this data.
Which pairing you will get. We can tell you what is certified and how often. We cannot tell you what your installer will fit.
Sales or market share. Certificate counts are not sales. Those figures are not published.
Whether a model is good. We have not tested any equipment and we do not rate or recommend products.
Rebate eligibility. ENERGY STAR certification is not automatic qualification for any particular incentive program. Check the program's own rules.

Missing data

Where the source does not publish a value, we show "not published" rather than a zero or a "No". This matters for cold-climate status in particular: an absent value is not a negative one, and treating it as one would invent a fact the EPA never stated.

How this is paid for

At the moment, it is not. There is no advertising, no affiliate link and nothing for sale. The site costs about twelve euros a year in domain registration and is hosted on a free tier.

Display advertising is the intended way to cover that, and the code for it is written but switched off: no ad script is loaded, no space is reserved and no ads.txt is published. If that changes, the privacy policy will be updated before a single ad is served. Advertising will not be placed inside the calculator's form or between your inputs and your result, because that would break the one thing the site is for.

Nothing here is influenced by any manufacturer. No brand has paid for placement, and the 16 model pages were selected by the rule described above, applied to the data.

Corrections

If a figure here is wrong, tell us. We will correct it and say what changed and when. The dataset is refreshed from the EPA source and every page carries the retrieval date it was built from.

26 September 2026, badge-engineering count. Published as 849 outdoor model numbers sold under more than one brand; the correct figure is 846. Some brands appear in the source under more than one spelling, and each spelling was counted as a separate brand: three Comfort Aire model numbers counted as shared only for that reason. The spellings are now merged as described under grouping, which also brings the brand count from 197 to 194 and changes the breakdown by number of brands. The largest case, 17 brands on one model number, did not change.

Who is doing this

Adrián Criado. Not an HVAC contractor, and nothing here is installation advice — that is exactly why this site sticks to what the certified data says and hands off sizing and installation questions to people licensed to answer them. What I do is build and maintain the pipeline that turns the EPA's raw certification file into something a homeowner can check a quote against. More about that.

Last built from data retrieved 2026-09-26.