Different claims require different evidence.

Sources are selected for relevance to the defined question. Depending on scope, eligible material may include official product information, public user discussion, retailer listings, research publications, patent records and observable manufacturing or company activity. Eligibility does not make every source equally reliable.

Product names, formulas, formats, markets, sellers and versions are kept separate where possible. Signals from an LED kit, strip-only product, old formula or retailer fulfilment path should not be applied automatically to another variant.

Selection rules, date range, market and exclusions should be recorded before synthesis. Reposted, syndicated or substantively duplicate material should not be counted as independent evidence. Source volume is not published when it has not been verified.

Repetition can make a signal worth investigating, but it does not by itself establish frequency, causation or a defect rate. Isolated reports may be retained as questions or edge cases and must not be presented as category-wide conclusions.

Praise, neutral experience, conflicting outcomes and evidence that weakens an emerging interpretation should remain visible. A useful analysis explains what is inconsistent and which product, user or evidence differences could matter.

AI may help organize candidate themes, normalize labels and locate possible duplication. It is not treated as a source and does not independently verify claims, diagnose causes or decide what should be published.

Human review controls the research question, checks sources, resolves product and variant ambiguity, tests alternative explanations, writes limitations and makes the final editorial judgment.

Published conclusions should be traceable to retained notes or source records at an appropriate level of detail. Where traceability is still incomplete, the material should be labelled as preliminary, withheld from indexed lists or excluded rather than completed with invented support.

Forecast entries are rolling editorial judgments. They should be revised, narrowed or removed when technical obstacles, product evidence, patents, manufacturing signals or market activity change the underlying case.

Paid research, supplied materials, affiliate relationships, samples or other commercial context should be disclosed where relevant. Commercial access does not convert a client claim into editorial evidence or guarantee favorable coverage.