1. Find the promise
Product claims come from named primary sources. The record preserves the wording or labels a paraphrase, its source, observation date, scope, and any known version or plan. Unknown metadata stays unknown. Sources are not automatically fetched or executed.
2. Read firsthand experience
Worked, partly worked, and did not work describe one contributor’s task. One active experience per person and scope is counted. Product and claim experiences remain separate. Affiliated contributors can participate, but are excluded from independent summaries. A small sample is not a representative probability.
3. Follow the evidence trail
Evaluations describe the task, expected and observed results, repeatable method, environment, interventions, and limitations. Trials, cost, and time are recorded only when known. Links and files support inspection; they do not prove authenticity or independent reproduction. Private files are visible only to their contributor and authorized moderators.
4. Assess a scoped finding
An unaffiliated moderator can record support, mixed support, contradiction, or insufficient evidence. Findings need a specific scope, limitations, a reason, and published evaluation references unless evidence is insufficient. They are editorial judgments about the tested conditions, never universal product scores.
5. Keep the record correct
Pending revisions do not replace approved content. Earlier findings retain their history. Withdrawal or restriction removes the affected public contribution. A finding whose references are no longer available must be read as needing reassessment. Corrections and appeals go to the review queue.