How we review

Every verdict box on this site links here, because a score is meaningless unless you know how it was arrived at.

Every verdict box on this site links here, because a score without a method behind it is decoration. This is how we review, in full.

What we actually do

We are a research-driven publication. That phrase gets used loosely, so here is precisely what it means in our case.

Specification analysis. We read the manufacturer's full specification sheet, not the marketing bullets, and we convert every figure into something comparable. A snow load quoted as "1,500 lb" and one quoted as "66 lb/m²" describe the same property in different units, and putting them on the same scale is frequently where the interesting findings are.

Installation manuals. These are published as PDFs before you buy, and almost nobody reads them. They contain the clearances, the slope limits, the torque specifications and the electrical requirements — the numbers that decide whether a product fits your house rather than whether it is any good.

Warranty documents in full. Not the headline term. We check what is covered, for how long, and which component carries which period. A "five-year warranty" that covers roof panels for five years and the structural frame for one is a materially different offer, and that distinction has changed our score on more than one product.

Regulatory and code research. Where a product's use is governed by rules — NEC circuit requirements for electric tankless heaters, federal LSV rules for utility vehicles, ASCE snow loads for outdoor structures — we go to the governing document rather than to a summary of it.

Owner-feedback aggregation. We read retail reviews, forum threads and manufacturer support documentation looking for patterns, not anecdotes. A single one-star review is noise. Twelve people reporting the same failure at the same component is a finding, and it is the closest thing to durability data available for products this new.

What we do not do

We do not claim hands-on testing. We do not receive review units and we do not buy every product we cover, so we do not write "we ran it for six months" or "we measured 0.3 inches of deflection at 45 mph". Publications do write those sentences without having done the work, and it is both dishonest and, under FTC guidance on endorsements and testimonials, a legal exposure.

Where the site's founder has genuine first-hand experience with a product from their own property, that is stated explicitly as first-hand and distinguished from the specification work.

We do not accept payment for coverage. No sponsored posts, no paid placement, no advertiser input into what gets covered or how it scores.

We do not use aggregate ratings. You will not find a star average on this site, and our structured data deliberately omits aggregateRating. A single-editor publication has no aggregate to report, and manufacturing one is the fastest route to a structured-data penalty.

The scoring rubric

Scores are out of 10, to one decimal place, and they are set before any affiliate link is attached to the article.

Score Band What it means
9.0–10 Exceptional Best in its category on the specifications that matter, with no significant caveat. Rare.
8.0–8.9 Strong buy Clearly good, with limitations that are real but easy to work around or plan for.
7.0–7.9 Good with caveats Sound product, with at least one issue serious enough that a specific group of buyers should look elsewhere.
6.0–6.9 Situational Works well for a narrow use case and poorly outside it.
5.0–5.9 Hard to recommend Competent alternatives exist at similar money.
Below 5.0 Avoid A defect, a safety concern, or a value proposition that does not hold up.

Four factors move a score, weighted roughly equally:

  • Does the product do what it claims, at the conditions you will actually use it in? Headline figures measured at flattering conditions are noted rather than accepted.
  • Is the installation requirement honestly represented? A cheap appliance with an expensive installation is not a cheap appliance.
  • What does the warranty actually cover, on the components most likely to fail?
  • Is there a competitor doing the same job better at similar money?

We do not grade on a curve, and we do not adjust a score to match a conclusion we had already reached.

How we handle contradictions

Manufacturers publish contradictory specifications more often than you would expect. One product we cover publishes a snow rating of 8.2 psf in its listing bullets and 66 lb/m² — about 6.1 psf — in its description, with no note reconciling them.

We print both figures and say which one we would design around. Silently choosing the flattering number is easier and it is exactly the behaviour that makes product reviews worthless.

Prices and dates

Prices are stored in structured data files with a last-checked date, and that date is shown wherever a price appears. Prices in this category move constantly — a gazebo can swing $500 across a season — so a price without a date attached is not information.

The updated date on an article changes only when the content meaningfully changes. We do not bump dates to look fresh to a crawler; if you see a recent date, something was genuinely revised.

Corrections

If we get something wrong, we fix it and say so in the article. Our full corrections policy is on the our editorial policy page, and the fastest way to tell us about an error is through get in touch.

How this site is funded

Affiliate commissions, disclosed above the first affiliate link on every article that carries one, and explained in full on how we are funded.

Every affiliate link on the site is generated from a single central configuration file rather than written into individual articles. That is an engineering decision with an editorial purpose: there is no per-product commission data anywhere in the content for a score to be influenced by, and adding or removing a merchant relationship cannot quietly change what an article says.

Questions people actually ask

Do you physically test the products you review?

No, and we say so on every review. We are a research-driven publication: our work is specification analysis, manual and warranty review, regulatory research and owner-feedback aggregation. Claiming hands-on testing we have not done would be both dishonest and, under FTC rules on endorsements, a legal problem. Where the site's founder has genuine first-hand experience with a product, that is stated explicitly as first-hand.

Do manufacturers pay for coverage?

No. We accept no payment for coverage, no paid placement, and no sponsored content. We do not receive review units, which is part of why we do not claim hands-on testing. Our funding is affiliate commissions, disclosed on every page that carries a link.

Does a higher commission rate influence a score?

No, and the structure of the site makes that verifiable. Scores are set in the content file before any affiliate link is added, and every affiliate link on the site is generated from one central configuration file rather than written into the article. There is no per-product commission data anywhere in the codebase for a score to be influenced by.

What does a score of 8.4 actually mean?

It means a strong buy with known limitations. The bands are set out on this page: 9.0 and above is exceptional, 8.0–8.9 is a strong buy, 7.0–7.9 is good with real caveats, 6.0–6.9 is situational, 5.0–5.9 is hard to recommend and below 5.0 is avoid. We do not grade on a curve and we do not adjust a score to fit a conclusion.

How often are reviews updated?

Prices are re-checked quarterly and the date shown on each review reflects the last substantive update. Specifications are re-checked when a manufacturer revises a listing or a model changes. The updated date on a page changes only when the content meaningfully changes — we do not bump dates to look fresh.

What do you do when a manufacturer publishes contradictory specifications?

We print both figures and say which one we would design around. This happens more than you would expect — one product we cover publishes two snow ratings in the same listing that differ by 25%. Picking the flattering number silently would be the easy option and it would be the wrong one.