The Quiet Maker Keeps The Lower Score

Last week we published a rule change and left one question deliberately open. We are closing it today, and the answer is that nothing gets compensated.

The question was about makers nobody has written about. Before we score a product we score our own evidence about it, out of a hundred, and one of the five things that audit looks at is how many independent sources exist. A small maker with no press coverage and no reviews scores low there, through no fault of their own.

Our rulebook used to promise those makers a compensation, moving the missing points onto other factors. That promise could never run. Moving the points pushed a factor above its cap, and no factor can hold more than twenty points, not in our specification and not in the software that stores it. We retired it and said we would settle what replaces it.

Nothing replaces it. The absence is scored as it is. Every factor is answered on the evidence that actually exists, the missing coverage is written into the trust explanation on the product page, and no other factor is quietly inflated to make up for it.

Two other shapes were on the table and both were rejected for the same reason: they cost the reader more than they gain the maker. We could have scored each product only on the factors that can exist for it, but then the five factors stop adding to a hundred, and the arithmetic a reader can check by hand is the entire point of publishing the breakdown.

Or we could have changed what the first factor asks, so that a quiet brand is measured on what it can show rather than on how loud the internet is about it. That one is the trap. We spent this month finding records where a factor’s name and the question behind it had drifted apart, and changing the question again would do the same damage on purpose, to every score already stored.

What this costs is worth saying plainly, because it is a real cost. A maker the wider internet has never covered carries a lower confidence score than one it has. That score is a statement about the strength of our evidence, never a verdict on the maker, and it is published as a limited confidence label rather than as an exclusion. Since this site began, the confidence floor has decided four rejections.

Measuring that number taught us something we did not go looking for. Of the products we have turned down, a large share are exactly this case: real stockists, real history, no reviews at all. Almost none of them were turned down by the rule we just ruled on. They failed on the score that a low confidence feeds into. So the tilt we were worried about is real, and it lives in a different rule than the one we were examining. That is now an open question of its own, and it is a bigger one than this was.

We would rather publish a number we can explain than a number that flatters.