The Record Now Names the Gate a Product Failed

A product has to clear three separate things to be published here: a composite score, a floor on trust, and a floor on how confident we are in the evidence behind it. Miss any one and the verdict is no.

The list where those verdicts live has always said which one. The record page you click into from that list never did. It showed the number and it showed the bar, and it left the reader to assume the number was the reason.

Often it was not. A product can sit a hundredth under the bar and be turned down for something else entirely, and on that record the number was pointing at the wrong cause while being perfectly accurate.

A rejection record now carries a labelled line naming the gate it fell short on, wherever a gate was recorded against it. The words are not written twice: the list and the record read the same three labels out of one place, so the two can never end up describing the same decision differently.

The line matters most where the page has nothing else to offer. Some products are stopped before a score is ever reached, and those records show no measurement at all. There the labelled line is the only thing on the page that says why, and it is the reason we put it in the record rather than inside the measurement strip, where it could never have reached them.

What it does not do is repeat. The strip above it already carries the score and the distance from the bar, so the line carries neither. It adds the one thing the page was missing and nothing the page already had.

These records also gained a way out. Until now a rejection ended with a single link back to the whole lab, and almost nothing else: this is a family of pages that rarely links anywhere. Each one now offers the rejections in its own category alongside it, built from the category the record already holds, with no list kept by hand anywhere.

We were wrong about why that link was worth building, and the reason is worth publishing. The plan argued the closing row looked broken with one link in it. Reading the stylesheet before building showed it does not: a rule already in the stylesheet centres a lone link precisely so it reads as a choice. The cosmetic argument was false and the real one survived it, which is that a reader who finishes one of these had nowhere of their own to go.

One thing is deliberately absent. The new link carries no count of what is behind it. Our own count for a category includes the record you are standing on, so any number on that link would have promised one more than it delivers, and a number that is almost right is the kind we do not publish.