How our methodology works

Start with the question

We define the product question and the use case before drawing conclusions. That keeps the work focused on practical needs instead of a universal ranking.

How we choose evaluation criteria

We identify the factors that can realistically change a buying decision before comparing products.

ChoiceCheckLab does not give equal weight to every specification simply because it appears on a product page. Criteria should be relevant to the buying question, applicable consistently across the compared products, understandable to the reader, supported by available evidence, and meaningful enough to affect the recommendation.

For a clothing-based underarm sweat pad comparison, factors such as absorption, hold, comfort, size, visibility, and clean removal may matter more than minor packaging or marketing differences.

Criteria can vary by category because different product decisions depend on different factors. We do not imply that every comparison uses the same fixed scoring model.

Our evidence standards

Recommendations should make it clear what was directly observed, what came from documentation or external sources, and what remains uncertain.

• Original hands-on testing — used only when ChoiceCheckLab directly examined or used the product or relevant feature.
• Original measurement or documentation — used when ChoiceCheckLab measured, photographed, documented, or systematically compared a product without claiming a full use test.
• Research-based analysis — used when conclusions rely on specifications, manufacturer documentation, credible external sources, and structured analysis.
• Customer-feedback analysis — used to identify recurring patterns in customer experiences; isolated reviews are not treated as proof.
• Editorial inference — used when a reasonable conclusion is drawn from available evidence and never presented as independently verified fact.

Seller or manufacturer claims are not automatically treated as verified facts. Missing or uncertain information should remain visible, and direct testing must never be implied when it did not occur. Recommendation confidence should reflect the available evidence.

How we research a comparison

We are rebuilding this site's comparisons around one idea: the useful information is usually what the product listing leaves out. This section describes that procedure, and says plainly where it has and has not been applied.

The procedure. We read the most recent customer reviews for each product, and then every one-, two- and three-star review the listing holds, as far back as they go. The recent pass is mostly people who have just opened the box, so it cannot tell you what fails. The low-star history can, and reading it year by year is what separates one bad production run from a design that has been failing the same way for a decade. We record only what appears independently three times or more: a single angry review is noise, while three separate people describing the same failure is a finding. Where a reviewer gives a timescale — after four trips, three months in — we keep it, because knowing when something fails matters more than knowing that it can.

We read the questions buyers ask on the product page. Those questions are a direct record of what the listing failed to explain.

We look beyond one retailer. The same product sold elsewhere attracts a different set of buyers, and a different set of complaints.

We read the manufacturer's warranty terms, and specifically what they exclude. A company knows what fails in its own product, and the exclusions are where it says so.

We check public safety and certification records: the CPSC recall database for anything with a battery, a heating element, or a child in the room, and the relevant certification register whenever a listing claims a certification.

Where independent laboratory testing exists, we cite it and name who carried it out. Where it does not exist, we say so.

Where it has been applied. A comparison researched this way ends with a section headed "What we looked at", listing what was consulted and on what date. If a source was checked and turned up nothing, that is listed too.

If a comparison carries no such section, it has not been researched this way yet. Those pages rest on manufacturer documentation and published independent testing, as the evidence labels on each page set out. We are working through them in order. We would rather tell you which is which than let you assume.

How we evaluate trade-offs

A useful recommendation should explain not only what a product does well, but also what the user gives up in return.

ChoiceCheckLab compares trade-offs such as performance versus simplicity, price versus long-term value, convenience versus flexibility, size versus portability, coverage versus discretion, features versus ease of use, durability versus weight or complexity, and maintenance versus convenience.

The best product is not automatically the strongest product in every individual criterion. A recommendation should identify who benefits most from the strengths, which limitations matter, which users may prefer an alternative, and what conditions could change the recommendation.

Review meaningful changes

When relevant evidence or product context changes, conclusions may need review. Corrections should have a clear path and material limitations should remain close to the claim.