A tester wrote that picking a food didn’t change the wine. The unit tests passed — they score a hand-built wine that is obviously right against one that is obviously wrong.
So I measured what he saw: forty cellars from the real catalog, one white among eight reds. Ten times out of forty, choosing fish kept a red.
The white had no style traits. Half the published catalog has none, and a missing trait scored zero — while a light, high-acid red collected points for having been enriched. The ranking wasn’t matching wine to food. It was ranking wines by whether we had data on them.
Fixed by gating wrong colours out, and scaling style points into the colour signal’s range — so a wine we know nothing about sits in the middle, not the bottom.
Absence has to be neutral, or the score measures your own pipeline.