How HVAC Bench rates HVAC brands
A brand rating here answers one question: how well does this manufacturer support what it sells? It does not answer whether the equipment is reliable, because that is a model-level question measured by surveys this publication does not run.
What the number means
The HVAC Bench brand rating is a support rating. It measures the warranty a manufacturer writes down, how a replacement part and an authorised repair reach the equipment, whether the technical documentation can be obtained, and how much of a heating and cooling job one badge covers. Every one of those is a published fact that can be checked, argued with, and re-checked next year.
It is not a quality ranking, and reading it as one will mislead you. A manufacturer can write a long warranty and build indifferent equipment. Another can build excellent equipment and publish nothing a technician can use. The rating says which of those two positions a brand is in, and it stops there.
What a brand rating cannot tell you about your equipment
A brand makes many products across many years, and they do not fail at the same rate. A manufacturer with a strong rating here still sells entry models and premium models, and a badge that scores well on paperwork can still have shipped a bad production run. Use the rating to compare manufacturers on the commitments they make. Use the model number on the data plate for anything about the machine itself.
The four categories
Weights total 100%. They came from what the research showed matters and from what can be evidenced at brand level, not from an opinion about which category sounds most important. A test in this repository fails the build if the weights stop adding up.
What the manufacturer's own warranty document promises, what it promises when the buyer does nothing, and whether labour is covered.
This is the one place where a manufacturer states in writing, at its own cost, how long it expects its equipment to last. It carries the most weight because it is a commitment rather than a claim, because it is enforceable, and because the gap between a headline term and the term a buyer actually receives is the most consequential thing a brand page can say before somebody signs a quote.
What earns a score
How a replacement part and an authorised repair reach the equipment, and whether a part can be identified without a trade account.
A warranty that covers a part is worth what the part is worth on the day it is needed. Access is structural and published: a factory part either sits on a wholesaler's shelf or it is ordered through one channel. This weighs almost as much as the warranty because it decides what happens on the day a system stops, and because equipment outlives whatever warranty arrived with it.
What earns a score
Whether the service literature, the fault code meanings, and the installation and operation manuals can actually be obtained.
A machine nobody can read the manual for gets diagnosed by guesswork, and guesswork is billed by the hour. HVAC Bench holds first-hand evidence here that it holds nowhere else: building this reference library meant obtaining documentation for every manufacturer on the site, and that record shows which of them publish their own literature and which required a copy hosted by somebody else.
What earns a score
How much of a heating and cooling job one badge covers, taken from the equipment classes and product families recorded in the HVAC Bench brand registry.
Range matters because mixing brands across a matched system splits responsibility when something fails, and because a narrow range forces a second supplier into the job. It carries the least weight because breadth is not quality: a manufacturer that makes one thing can make it well, and this category should never be able to carry a brand on its own.
What earns a score
Why reliability is not one of them
This is the question readers ask first, so it deserves a direct answer. Measuring brand reliability means asking thousands of identified owners what broke and when. Consumer Reports estimates it from survey data covering more than 13,000 installed central systems. Which? runs an annual survey of over 10,000 boiler owners alongside about 200 heating engineers. Those are real instruments with real sample frames.
HVAC Bench runs neither. We could assemble an impression from forum threads and retailer reviews, attach a decimal point to it, and publish it next to figures that came from warranty documents. That would be the single most dishonest thing this site could do, because the number would look exactly like the ones beside it and would mean nothing. So reliability is absent, and it is absent on purpose.
Family-level owner evidence is handled by the HVAC Bench Score, which is a separate instrument with its own thresholds and publishes no figure until the observations behind it clear them.
What else was left out, and why
- Reliability
Brand reliability is measured by asking thousands of identified owners what broke and when. Consumer Reports estimates it from survey data covering more than 13,000 installed central systems; Which? runs an annual survey of over 10,000 boiler owners alongside about 200 heating engineers. HVAC Bench runs no such survey. Building one from forum threads would produce a figure with the shape of evidence and none of the substance, and it would be a figure about who posts rather than about what fails.
Family-level owner evidence is handled separately by the HVAC Bench Score, which publishes no number until the observations behind it clear a stated threshold.
- Owner satisfaction
The same problem for the same reason. Satisfaction is a survey result, and the people who volunteer an opinion online are not a sample of the people who bought the equipment.
Recurring owner and technician themes appear on a rating as reported signals, labelled as testimony, and they never move a score.
- Price and value for money
Installed price is set by the contractor, the region, the season, and the state of whatever is being replaced. The same model can differ by thousands between two quotes in one city, so there is no brand-level price to rate.
Where a warranty shifts running cost, that shows up in the warranty category, which is where a documented cost commitment belongs.
- Independent performance certification
This one was drafted as a scored category and then removed. Certification directories such as the AHRI directory in the United States and MCS in the United Kingdom are the right kind of evidence: independent, published, and outside the manufacturer's control. The problem is granularity and verification. Listings are per product rather than per brand, so a brand listing proves nothing about the model somebody is being quoted, and the directories could not be queried brand by brand in a way this site could reproduce on review. Scoring it would have meant assuming listings we had not confirmed, across every brand, inside a number presented as evidence-based.
The directories are named on this page so a reader can check a specific model themselves, which is the level the certification actually operates at.
- Popularity
We looked for a defensible way to include it. AHRI publishes United States shipment data for the industry as a whole rather than split by brand, and brand-level share is sold by market research firms rather than published, so there is no free reproducible figure. Popularity also measures distribution and advertising rather than whether a machine is any good.
Nothing about how widely a brand sells is folded into its score, in either direction.
The evidence hierarchy
A category is scored from the strongest class of evidence that answers it. Lower classes are used only where nothing above them can, and the bottom class is never used to set a score at all.
- Manufacturer warranty documents. The warranty certificate or the published warranty terms. These settle what a warranty says, because the manufacturer is the party bound by it.
- Other manufacturer-published policy. Support programmes, literature libraries, parts and authorised service pages. These establish how a manufacturer has chosen to structure its support.
- Independent directories and safety regulators. Certification listings and recall notices. Outside the manufacturer's control, which is what makes them worth citing.
- The HVAC Bench sourcing record. First-hand for this publication: which documents could be obtained for every manufacturer here, and whether they came from the manufacturer or from somebody else hosting a copy.
- Trade press reporting a documented programme. Used where a manufacturer changed something and the change is reported before its own terms page catches up.
- Technician and owner reports. Read to find recurring themes worth investigating, and shown on a page as testimony carrying that label. A report can point at a question. It cannot answer one, and it never sets a band.
Technician and owner reports sit at the bottom for a reason. A forum thread describing a three-day wait for a part is worth reading: it points at a question about how a channel is built, and that question can then be answered from published material. What it cannot do is establish a fact about a manufacturer, because one person's week is not a pattern. Where such reports appear on a rating, they are labelled as testimony and they move nothing.
How a score is calculated
- Each category is assessed against the published rules above, producing sub-points out of 4.
- Sub-points are multiplied by 2.5 to give a category score out of 10.
- Every score is rounded to the nearest 0.5. Nothing in this evidence distinguishes 8.3 from 8.4, so publishing a tenth would borrow the appearance of a measurement nobody took.
- The overall figure is the weighted average of the rounded category scores, rounded the same way. That is deliberate: it means the numbers printed on a rating card add up to the number at the top of it, and a reader can check the arithmetic by hand.
- Where a category does not apply to a kind of product, its weight is removed from the total rather than counted against the brand. Range is not applied to controls brands, because a thermostat manufacturer is not trying to cover a heating job.
When no score is published
A brand shows no figure unless all of the following hold.
- Warranty is established from manufacturer documents. It carries the largest weight and rests on the firmest evidence, so a rating without it would be a rating of the periphery.
- Scored categories cover at least 80% of the weight that applies to that brand.
- Every scored category cites at least one document that is allowed to set a score. Testimony alone leaves a category unscored.
Where those are not met, the card says Not enough evidence and lists what is missing. That is a statement about our record rather than about the brand, and the difference matters: a manufacturer that publishes nothing centrally is not the same as a manufacturer that publishes something poor.
Missing, conflicting, and regional evidence
Missing
A fact that could not be established is recorded as unestablished and scores nothing. Several manufacturers publish the term they want quoted without saying what applies to a buyer who registers nothing, and in those cases the card says the fallback was not stated rather than assuming a figure. This costs those brands points they may well deserve, which is the correct direction for the error to run.
Conflicting
Where accounts disagree, the manufacturer's own document wins. Where two manufacturer statements disagree, or where no manufacturer statement settles it, the record takes the more conservative reading and says so in the limitations on that brand's card.
Regional
Warranty terms and parts channels are national. A rating names the market whose documents were read, and it describes that market only. Several manufacturers appear here twice for this reason: the badge on a heat pump in Manchester and the badge on a furnace in Ohio can belong to different companies with different obligations.
Model-level differences within one brand
Warranty length routinely varies across a manufacturer's own catalogue, sometimes by ten years between the entry range and the flagship. Where that happens, a rating records the longest term a buyer can reach and the term that arrives when nothing is done, and the limitations on the card name the range the headline figure belongs to. The product tier on a quote decides what is actually being bought, and the badge does not.
How user feedback is used
Technician and owner reports are read while researching a brand, and they earn their place by pointing at questions. A recurring complaint about part availability is a reason to go and establish how the parts channel is built. Once that is established from published material, the published material is what appears in the score, and the complaint appears, if at all, as a labelled observation about what the structure means in practice. No count of posts, no sentiment average, and no star rating from any retailer enters the arithmetic.
Certification, which is named rather than scored
Independent certification was drafted as a fifth category and removed. Listings are per product rather than per brand, so a brand listing proves nothing about the model on a quote, and the directories could not be queried brand by brand in a way this site could reproduce on review. Scoring it would have meant assuming listings we had not confirmed. The directories are named here instead, because they answer the question at the level it actually operates at.
- AHRI Directory of Certified Product Performance
Certified matched systems, with published ratings verified by independent laboratory testing under the applicable standard. Search by the outdoor and indoor model numbers on the quote, not by the brand.
Applies in: United States and Canada
- ENERGY STAR certified product finder
Efficiency recognition, applied per product. A manufacturer being an ENERGY STAR partner does not make every model it sells qualifying.
Applies in: United States
- NEEP cold climate air source heat pump product list
Performance across the operating range rather than at one rating point. Listed products hold at least 70% of rated 47F heating capacity at 5F and reach a coefficient of performance of 1.75 at 5F.
Applies in: United States, cold climates
- MCS product and installer directories
Certification for the product and for the installer. Both are conditions of the Boiler Upgrade Scheme grant, so an uncertified product or installer sits outside the main funding route.
Applies in: United Kingdom
Why these ratings show no stars in search results
Star ratings in a search result come from review structured data, and this rating does not qualify for it. A review snippet describes a rated item such as a product, and it expects a rating either from a named critic reviewing that item or aggregated from people who used it. What is published here is one editorial assessment of how a manufacturer supports its equipment, which is neither. Marking it up as an aggregate would assert a sample of user ratings that does not exist.
So the markup on these pages describes what the pages are, a manufacturer collection and a method document, and nothing more. A test in this repository fails the build if rating markup ever appears on a brand hub or on this page. Stars would earn clicks. They would also be a claim about our evidence that our evidence does not support, and that is the exact trade this publication exists to refuse.
Review and reproducibility
Every record is reviewed at least every 12 months, and sooner when something documented changes: a warranty revision, a change in how literature is published, a change of ownership, or a safety recall. Each record carries the date it was last reviewed, and each evidence entry carries the date the document was read rather than a date the document claims, because warranty pages change without announcing it.
The scores are computed rather than entered. The facts, the weights, and the rules live in the repository as data, and there is no field anywhere that lets an editor type a number. Revising a weight changes every rating on the next build. Correcting a fact changes one brand, everywhere it appears, at once.
Every rated manufacturer
48 of 50 manufacturers carry a figure, running from 3.0 to 9.0. Scores compare manufacturers on support, and they are most meaningful read within a market and a product class rather than across the whole table.
Manufacturers with no score
These carry no figure because the evidence did not reach the bar above. In both cases the missing piece is the warranty a buyer can read before purchase, which is the largest single weight in this framework.