One recommendation. Not fifty.
That's the whole idea.
Most gift sites show you the winners. We show you the work.
This is our research log. Products we rejected. Decisions we debated. Systems we changed. The messy middle that usually stays hidden.
Why? Because trust isn't built by saying "trust us." It's built by showing you how we think.
Eight product pages printed a confidence label their own score contradicted, and one rejection record gave a single audit factor 25 points out of a possible 20.
Source Correction - The confidence label is now derived from the score, and the factor score above its cap is named as wrong.
Our rejection records said products are turned down for confidence below 60, gave a floor of 40 where it is 35, and in one case printed no reason at all.
Source Correction - Rejection reasons cited gates we do not use. Every verdict stands; every reason now matches what actually happened.
The Bar's machine readable dataset description stamped itself as revised today on every day, publishing a daily revision nobody made.
Source Correction - The Bar's dataset stamp was the render date; it now carries the real recount date or nothing.
Our statistics page published a "Last updated" date that was simply the current date, on every page load, so it appeared to have been revised the day you read it and every day before.
Source Correction - The statistics page stamped itself with the current date on every load, publishing a daily revision it had not made
Products must pass minimum thresholds in each dimension AND achieve 8.0+ weighted total.
Source Correction - Our Lexicon said the score needs a minimum in every dimension; three gates reject products and none is a dimension floor
Below a Confidence Score of 35 we publish no verdict on a product. That floor decided real rejections and appeared on no page, and it is not the same rule as the 60 mark for the Labs Verified badge.
Decision Log - The Confidence floor of 35 becomes a published gate, and is separated from the 60 mark that governs the Labs Verified badge
The Bar listed four reasons a product fails verification. Two of them, source diversity and unverifiable claims, had never caused a rejection, and a real gate was missing from the page.
Source Correction - Two published rejection reasons had caused none, the real third gate was unpublished, and the counter counted notes not products
Our For Brands page told brands that every product passes a five-gate evaluation, Trust, Glow, Wow, Moment and Fit, and that failing any one gate ends it.
Source Correction - Our For Brands page published a five-gate pass/fail standard we never used; it now states the four weighted dimensions we do
The example database entry on our trust page, displayed under the words Not close. Not paraphrased. Matched. The quote it showed as matched source text existed nowhere in our database.
Source Correction - The invented example is replaced by the real database row, the superseded version line is gone, the absolute claim is withdrawn
Corrected pages showed no trace of having been corrected, corrections carried no name, and the busiest record took every slot on all four pages that display them.
System Update - On-page correction notices, signed corrections and rule changes, a guaranteed slot per record, and the lab-coat chrome removed