Methodology and sources
Every specification on this site comes from one public dataset, and every calculation is arithmetic on either that dataset or a number you typed. This page lists both, plus the four places where a judgement is ours rather than the source's.
1. Specifications: ENERGY STAR Certified Heat Pumps
| Dataset | ENERGY STAR Certified Heat Pumps, U.S. Environmental Protection Agency |
|---|---|
| Access method | Socrata SoQL queries against the JSON endpoint, in batches filtered by outdoor_unit_brand_name and model_number prefix, each batch limited to a handful of records. The rows as returned are kept verbatim in scripts/energystar-rows.psv; this file is generated from them by scripts/build-systems.mjs. |
| API endpoint | https://data.energystar.gov/resource/83eb-xbyy.json |
| Dataset last updated | 2026-10-16 (the rowsUpdatedAt field in the dataset's own metadata) |
| Retrieved | 2026-10-16 |
| Records here | 101 certified pairings across 92 outdoor units |
| Licence | Published by the U.S. EPA ENERGY STAR program under the EPA data licence (https://edg.epa.gov/EPA_Data_License.html). ENERGY STAR is a registered mark of the U.S. EPA; this site is independent and is not affiliated with or endorsed by the EPA. |
| Verification | Every value is copied from the API response for the pd_id shown beside it. No value is estimated, interpolated, rounded or carried over from another record. Where the dataset has no value, this file stores null and the pages print 'not published'. |
The raw rows as the API returned them are kept in the repository and the processed records are published at /data/systems.json, with this provenance block embedded in the file itself.
What a record is
ENERGY STAR certifies a combination: one outdoor unit matched with one indoor unit. The ratings belong to the pair, not to the outdoor box alone, and the same outdoor unit paired with a different head often publishes a different SEER2 and HSPF2. So one record here is one certified pairing, and a page covers one outdoor unit with every pairing the dataset lists for it. Where two pairings publish identical figures, they appear as one row with the other indoor model numbers named beside it, rather than as duplicate tables.
Model numbers are reproduced exactly, asterisks included. ENERGY STAR carries asterisks as the manufacturer's wildcard for cosmetic or trim variants, so MUZ-FS12NA*** is the published identifier and not a typo on this site.
Coverage, and what is left out
A deliberate sample, not the whole dataset. The dataset held 17,874 records of product_type 'HP - Mini or Multi Split' at retrieval. Sampling is by brand and model series so that each page is a distinct outdoor unit rather than a cosmetic variant. Records whose indoor_unit_model_number is empty are excluded: those are multi-zone outdoor units certified without one matched indoor unit, and a matched-system spec table for them would be incomplete.
Records whose indoor_unit_model_number field is empty are excluded. Those are usually multi-zone outdoor units certified without a single matched head, and a matched-system spec table for one would be incomplete in a way the page could not honestly label. Ducted split systems and single-package units are out of scope entirely: this site covers ductless mini and multi splits.
2. The field definitions, and one gap in them
SEER2, EER2 and HSPF2 are defined in the dataset's own column metadata: SEER2 is seasonal cooling efficiency in Btu/Wh, EER2 is cooling efficiency at the 95°F test point, and HSPF2 is total seasonal heating output divided by total electric input over the heating season, also in Btu/Wh. Cold-climate status marks models meeting ENERGY STAR's cold-climate criteria, which the program describes as better low-temperature heating performance, suitable for use as a home's sole heating source.
The gap: The dataset's own column metadata carries no description for heating_capacity_at_47_f_btu_h, heating_capacity_at_17_f_btu_h, heating_capacity_at_5_f_btu_h or cop_at_5_f. They are therefore presented as the published capacity and COP figures at those outdoor temperatures, with no assumption about which rating point (maximum or rated) each represents. Some records publish a higher capacity at 5°F than at 47°F; such values are shown as published and not corrected.
This matters for one visible oddity. 77 of the cold-climate units here, and 27 of all 92, publish a higher capacity at 5°F than at 47°F. A plotted curve that rises as it gets colder is not physically how a heat pump behaves across one operating mode, so the most likely explanation is that the two rows describe different operating points — a maximum output at low temperature against a rated output at mild temperature. Because the dataset does not say, this site does not assume. The figures are printed as published, the curve joins the published points, and the caption on every plate says the lines between the points are drawn to join them and are not measurements.
3. The four assumptions that are ours
| Assumption | What it does |
|---|---|
| 1 | capacityClassBtu is ours, not a dataset field: the nearest standard nominal size (6k, 9k, 12k, 15k, 18k, 24k, 30k, 36k, 48k, 60k BTU/h) to the highest rated cooling capacity among a unit's pairings. It is used only to group comparable systems. |
| 2 | retention5 is ours, not a dataset field: the 5°F heating capacity divided by the 47°F heating capacity, as a percentage to one decimal place. |
| 3 | A page leads with the pairing that has the highest published HSPF2, then the highest SEER2. Every other certified pairing is listed on the same page. |
| 4 | Two pairings whose published figures are identical are shown as one row, with the other indoor model numbers listed beside it. |
None of the four invents a measurement. The capacity class is a grouping label, the 5°F share is a division of two published figures, and the other two are editorial choices about which pairing a page leads with and how identical rows are collapsed. All four are labelled as ours wherever they appear on a page.
4. The running-cost calculator
The running-cost calculator contains no data of its own. It converts one annual heating load, which you supply, into a gas cost and an electricity cost at prices you supply. The formulas are definitions:
A therm is 100,000 BTU by definition and a kilowatt-hour is 3,412.14 BTU by energy equivalence. The arithmetic is exact; the inputs are where the uncertainty lives, which is why the calculator prints each step with your numbers in it and labels the output a scenario.
The one modelling claim the calculator makes is that dividing an annual load by HSPF2 approximates a season's electricity use. That is what HSPF2 is constructed to represent, but it is measured under a standard rating procedure at fixed conditions, not in your climate. A winter colder than the rating condition will push real consumption above the figure; a milder one will push it below. The calculator says this on the page, next to the result, and it is the reason nothing on this site presents a cost output as a prediction.
The formulas live in one file, src/js/running-cost.js, used by both the build and the browser, so a page and the tool cannot drift apart.
5. What this site does not claim
- No author testing. Nothing here has been measured, installed or operated by the author. Every specification is a published figure with its record ID printed beside it.
- No installed prices. The dataset has none, and no verified regional price data is in this site. Cost queries about equipment or installation stay unanswered rather than estimated.
- No location predictions. SEER2, EER2, HSPF2 and COP at 5°F are rating-condition metrics. They compare systems with each other; they do not forecast a bill or a performance in any named city.
- No sizing advice. Choosing a capacity needs a room-by-room heat-loss calculation, which this site does not perform.
- No affiliation. ENERGY STAR is a registered mark of the U.S. EPA. This site uses the EPA's public data and is independent of the program; nothing here is endorsed by it.
- Not professional engineering review. The formulas and the cold-climate wording on this site are awaiting review by a qualified HVAC professional, which is recorded as an open item in the repository.
6. Corrections
If a figure does not match its ENERGY STAR record, that is a bug and not a judgement call. Each spec table prints the record's ENERGY STAR ID and AHRI reference number; the data file carries version 2026-10-16 and the retrieval date 2026-10-16. Those three things are enough to trace any number on this site back to the row it came from.