Methodology
How every verdict is computed.
A verdict combines two axes. Both come from public data; we do not test products ourselves.
Owner satisfaction (Amazon)
We take up to 20 of the brand's most-rated products on Amazon.com. A is the rating average weighted by the number of ratings. L is the share of those products rated below 4.0.
- High: A of at least 4.5 and L of at most 10%
- Medium: A of at least 4.2 and L of at most 30%
- Low: everything else
A brand needs at least 5 rated products to get a verdict.
Safety record (CPSC)
Recalls published by the U.S. Consumer Product Safety Commission in the last 36 months: none is Clean, one or two is Some, three or more is Frequent. Recall counts are not adjusted for sales volume, so large brands can look worse than small ones.
Verdict table
| Owner satisfaction | Safety record | Verdict |
|---|---|---|
| High | Clean | Reliable |
| High | Some | Generally good |
| High | Frequent | Mixed |
| Medium | Clean | Generally good |
| Medium | Some | Mixed |
| Medium | Frequent | Use caution |
| Low | Clean | Mixed |
| Low | Some | Use caution |
| Low | Frequent | Use caution |
What is shown but not scored
Court cases (only product-related ones are listed; a filed case is an allegation, not a finding), Trustpilot (it rates the brand's store and service, not the products), warranty terms and company facts.
Why not review analysis?
Tools that graded individual reviews, like Fakespot, are gone. We rely on records that are hard to fake instead. More on Fakespot alternatives.
Commissions
We earn from some links. Commissions do not change the formula or the order of products.