Jan 21, 2026 · 6 min read
What an explainable match score looks like
A percentage on its own is useless. Here's what a match score should actually tell you.
A bare “87% match” is a magic trick, not information. You can't act on it, you can't trust it, and you can't tell whether it's counting a keyword collision or real fit. An explainable score does the opposite: it hands you the evidence.
AmazApply runs deterministic eligibility checks first, so a hard blocker — a missing license, a work-authorization mismatch — always wins over any semantic similarity. Then it evaluates the posting requirement by requirement (met, partial, missing, unknown), blends that with your preferences using published weights, and reports a confidence level separately from the score.
The result is a breakdown you can reason about: where you're strongest, what's missing, what the posting left unknown, and what the agent would emphasize in your application. The model can explain the number — it can't quietly change it.