A match score is not a reason
Every ATS on the market will hand you a number between 0 and 100. Almost none of them will tell you what it is made of — and that is the whole problem.
Ask a recruiter why a candidate scored 78 and you will usually get an honest answer: they do not know. The number came out of a system that weighted some keywords against some other keywords, and the weighting is not visible, and even if it were it would not correspond to anything a hiring manager could argue with.
This is fine right up until the moment it matters. A candidate asks why they were passed over. A hiring manager asks why the shortlist looks like that. Someone from legal asks whether the process can be described. At that point a score is worse than nothing, because it looks like a justification while containing none.
What a reason looks like
A reason has three parts: a requirement that was agreed in advance, a finding about whether the candidate meets it, and the evidence the finding rests on. Take any one of those away and you are back to a number.
The requirement has to come first, and it has to be written down before anyone reads an application. Otherwise the criteria drift to fit whoever was read most recently, which is how a process that feels rigorous produces a shortlist nobody can defend.
Why we show the quote
When YoHe says a candidate has shipped a payments integration, it shows the sentence in their own documents that says so, with the file and page it came from. This is a constraint on the system, not a feature bolted onto it: if the passage is not there, the requirement is marked unevidenced rather than estimated.
That makes the output longer and less tidy than a score. It also makes it something you can disagree with, which is the point. A finding you can argue with is a finding you can rely on.
The uncomfortable part
Showing the evidence also shows when the evidence is thin. Plenty of good candidates write applications that do not demonstrate what they can do. An evidence-led system will say so, plainly, and the temptation is to smooth that over with a confidence estimate.
We think the opposite is right. A visible gap is an interview question. A gap averaged into a score is a person quietly dropped. So gaps get their own list, and next to each one, the question that would settle it.