The ATS scores we deleted, and what was wrong with them

Four vendor scores sat on the analysis screen, each with a percentage beside a recognisable name. They looked like the most valuable thing on the page, because they answered the question everybody actually has: will this survive the specific system this employer uses?

They were one internal score multiplied by a different constant per vendor. Workday was not measured against Workday. It was our own formatting score times 0.9.

No number is attributed to a third-party product, because we have never run one.

What replaced them is smaller and real: the rules that failed, and which systems each failed rule is known to affect. If nothing failed that is known to matter for a given system, that system is simply absent from the list. An absence is honest. A percentage we assembled from a constant is not, however confident it looks.

Nobody outside that one file had ever read the vendor scores, which is its own finding. A number can be the most prominent thing on a screen and still be load-bearing for nothing.