We do not claim a laboratory.
We claim a method.
There is a temptation in this category to describe every review as hands-on testing. We do not, because it would not be true at the scale we publish. What we do instead is state where every figure comes from, publish the weighting behind every ranking, and name the limits of the estimate. This page is the complete account of that.
The four-stage protocol
Every product review passes through the same four stages in the same order. The order matters: constraint mapping is deliberately last, so that a strong specification cannot carry a product past a situation it does not actually solve.
Specification Auditing
Manufacturer claims are extracted first and treated as claims rather than facts. We check whether the stated motor type is consistent with the claimed output, whether the stated runtime is plausible against battery capacity and motor draw, and whether the claimed ingress protection matches the charging architecture.
Where a specification is internally inconsistent, that inconsistency is published.
Aggregated Owner-Report Synthesis
A single unit tells you about one unit. We work through large volumes of verified-purchase reports and enthusiast-community threads looking for pattern failures: charging faults that appear after a specific number of cycles, seams that separate, coating that degrades. Pattern detection requires volume, and volume is the advantage of the aggregation model.
A failure that appears in a handful of reports is a note, not a finding. We distinguish the two.
Material and Documentation Review
We review safety documentation, material certifications and supply-chain claims. Where a product is described as body-safe, we check which specific material and which certification that description rests on, because the phrase itself carries no regulatory weight.
Certification references are reported as the manufacturer states them. We do not verify certificates we cannot obtain.
Constraint Mapping
The product is then mapped against the situations where it will and will not work: shared walls, reduced hand strength, heightened sensitivity, post-surgical recovery, long-distance use, budget ceilings. This is the stage that determines whether a product appears in a ranked guide at all.
A technically strong product can be excluded from every ranking if it fails the constraint map.
Where each figure comes from
This table is the honest answer to the question "how do you know that". If a number appears in a review without a stated basis, that is a defect in the review.
| Field | Source | How we present it | Confidence |
|---|---|---|---|
| Noise level | Manufacturer figure plus aggregated owner reports | Published as a range at a stated distance and setting | Moderate - construction and unit variance are both real |
| Battery runtime | Manufacturer figure, checked against reported real-world use | Stated at mid intensity, with the setting named | Moderate - depends heavily on intensity and age |
| Materials | Manufacturer specification and documentation review | Named material rather than marketing terminology | Higher - this is a documented attribute, not a measurement |
| Ingress protection | Manufacturer rating | Reported as the stated IP class without interpretation | Higher - though not independently re-tested |
| Editorial rating | Our own judgement against our own criteria | A single 0.0 to 5.0 figure with the reasoning published | Opinion, and labelled as such |
How rankings are weighted
Every ranked guide on this site publishes three things: the factors that were scored, the relative weight given to each, and the products that were excluded with the reason for exclusion.
Weights are not adjusted per article to produce a preferred outcome. A category in which the manufacturer's strongest product wins on its merits produces a ranking where it wins. A category in which the strongest product is beaten by something cheaper produces a ranking where the cheaper product wins.
The reason weighting is published rather than hidden is that it lets you disagree with us productively. If noise matters far more to you than it does to our default weighting, you can read the comparison table, apply your own weights to the published data, and get a different order. That is the intended use.
Limitations we accept
- Unit variance is real. Two units of the same model can differ audibly. This is why we publish ranges rather than single values, and it is why a complaint in one owner report is a note rather than a finding.
- Field performance differs from laboratory performance. Published acoustic ratings come from controlled assemblies. Real buildings have gaps, penetrations and shared structure. Our acoustic model applies a standard field allowance, and it can still be wrong about your specific building.
- We do not provide medical guidance. Nothing on this site is medical advice. Product selection where a diagnosis, medication or surgical history is involved is a conversation with a clinician, not with a review site.
- Our catalogue is deliberately incomplete. Recommendations are limited to products we have reviewed. A large part of the market is absent rather than ranked low, and absence is not a negative judgement.
- Ratings are opinions with published bases. A rating is a judgement against criteria we chose. The measured data underneath it is verifiable; the judgement on top of it is ours, and you are entitled to reject it.
Corrections policy
If a figure or a conclusion on this site is wrong, we want it reported. Corrections are made in the body of the affected page with the date of the change, not appended as a footnote and not made silently. Where a correction changes a verdict or a ranking position, the verdict or position changes too.
Send corrections to [email protected]. We also do not respond to link exchange requests, paid placement enquiries or SEO solicitation, and mentioning those in the same message as a correction will not change whether the correction is made.
Methodology questions
Do you physically test every product you review?
No, and we say so on every page rather than implying otherwise. We purchase some products, and we also work from manufacturer specification sheets, safety documentation and aggregated owner-report corpora. What we consistently do is state the source of every figure and flag where our confidence is lower.
How can you publish decibel figures if you do not measure every unit?
We publish ranges rather than single values, and we describe the measurement distance and setting. Where a manufacturer claim diverges from what owners consistently report, we print both instead of choosing the flattering one. A range with a stated basis is more honest than a precise number with an unstated one.
How are affiliate commissions handled?
Outbound commercial links route through an internal redirect and carry sponsored and nofollow attributes. Commission rates are not an input to any rating, ranking or inclusion decision, and a product without an affiliate programme can appear on a recommended list.
What happens when you get something wrong?
We correct it and note the correction rather than quietly editing the page. Where a correction changes a conclusion, the conclusion changes in the body copy rather than only in a footnote.
Do you accept free products from manufacturers?
Where a sample is provided it is disclosed in the review. Samples do not carry any commitment to coverage, to a rating floor, or to an editorial angle. We do not publish sponsored placements or accept payment for a position in a ranking.
Why is the ranking methodology published in full?
Because a ranking is an argument about priorities, and the writer priorities are the part readers cannot see. Publishing the weighting lets a reader who cares about something else entirely re-run the logic against their own values.