Every review on this site scores the brand on the same six things, with the same weights, and the overall score is the weighted average. I do it that way so you can put two reviews side by side and trust that a 7.8 means the same thing on both pages. The number is not a feeling. It comes out of the table below.

The six criteria

CriterionWeightWhat it measures
Material Integrity25%Thickness, seam technology, UV resistance, fire rating.
Warranty Terms20%Length, coverage, exclusions, workmanship warranty requirements.
Price Transparency15%Published pricing, ability to audit a quote, hidden costs.
Customer Service15%Direct manufacturer access, responsiveness, post-install support.
Real-World Longevity15%Field track record, industry data, installer interviews.
Installer Network10%Dealer coverage, installer training, regional availability.

Each criterion is scored from 0 to 10. Multiply by the weight, add them up, round to one decimal, and that is the overall score at the top of every review. The same six numbers drive the score breakdown you see on each review page, so the weights here and the weights there cannot disagree.

What the scale means

ScoreWhat it means
9-10Leads the category and is hard to fault on this point
7-8Strong, with a minor gap I can name
5-6Adequate. Real weaknesses exist and I say what they are
3-4Below the industry baseline. Hard to recommend on this criterion
1-2Failing or effectively absent

If a brand publishes nothing on a criterion, I score the absence, usually a 3 or a 4. I don't skip it and I don't average around it. A warranty document I can't read is a warranty I can't vouch for.

Where the evidence comes from

I read the manufacturer's actual warranty document, not the marketing page that summarises it. The two often say different things, and the document is the one that decides a claim. Where a brand holds a code evaluation (ICC-ES, CCMC, Intertek), I read the report and note its conditions, because a Class A rating over concrete is not the same thing as one over plywood. Dealer counts come from the brand's own locator. Pricing comes from published price lists where they exist, and from installer quotes I have seen where they don't, and the review says which.

Where I quote an installer, it is usually anonymous. The ones who tell you the truth about a product tend to prefer it that way.

If I don't have a number, the review says "varies" or "industry estimate." It does not invent one.

The rules I hold myself to

  • No paid, referral, or sponsored relationship with any manufacturer. Not Valordek, not Duradek, not anyone. No brand pays for placement, no link on this site earns me a fee, and if that ever changed it would be stated plainly on every page it touched.
  • No manufacturer sees a draft before it publishes.
  • No free product in exchange for a review.
  • The same six criteria for every brand. When a brand I like scores badly on a criterion, the score stands and the prose explains why I might still pick it for a specific reader. When a brand I'm sceptical of earns an 8, it gets the 8.
  • A wide spread, on purpose. If every brand scored 7 or better the scale would mean nothing. Across the site you'll find overall scores from about 5 to 8.5, and that spread is the point.
  • Corrections stay visible. When I get something wrong, the page says it was corrected rather than quietly changing.

What a best-of list adds

The ranked "best for" lists use these same scores, re-weighted for the job in the title. A rooftop list leans on roofing classification and on what the warranty pays when the room below gets wet. A coastal list leans on salt tolerance and appearance coverage. Every list names the criterion that decided each rank, and when a lower-scoring brand places above a higher-scoring one, the list says which fact justifies it. If a brand shouldn't be used for the application at all, the list says so and cites the brand's own documentation.

Questions

If a score looks wrong to you, tell me. The contact page reaches me, and so does the box at the bottom of every post. I have changed scores when a reader showed me a document I hadn't read. That is how this is supposed to work.

— Craig