Every score comes with its evidence

The situation

A sourcing tool hands you three candidates: 92, 88 and 84. You ask why the third one is an 84, and there is no answer available anywhere in the product. So you open all three profiles and read them yourself, which is the work the ranking was supposed to save you.

Nothing is attached to that number. You cannot show a hiring manager why a candidate scored what they scored. You cannot correct a score that is wrong, because you cannot see what produced it. And a candidate who genuinely fails a requirement looks identical to one whose CV never mentions it, which are opposite facts that a single number collapses into one.

What sits behind a score

When you confirm the brief, we write out what a weak answer and an exceptional answer look like for each requirement, in the language of the role you are hiring for. Candidates are judged against that description. Whether a keyword appears on the page has no effect.

Open a candidate and every requirement carries the line from their profile that decided it. You read the evidence, then the score.

Each requirement can come back three ways: met, clearly not met, or unanswered by the profile. An unanswered must-have is never a reject. Until it is known to be clean, though, the candidate is never moved forward automatically, and Zia asks them. A CV that says nothing about team leadership has not told you the person cannot lead, and good candidates fall off shortlists on that confusion every day without anyone noticing, because the number looks the same either way.

Why every must-have has a cost

A score you can trust also needs a brief that can produce one.

A must-have the profile contradicts rules the candidate out, while one the profile is silent on becomes a question, not a reject. Rejections still multiply rather than add. A requirement that excludes three candidates in ten looks harmless on its own. Five of them together leave you a sixth of the pool you started with, and the people they remove are not the weakest ones, just the ones whose profile happened to be quiet on one line.

Each extra must-have removes people who would have cleared your bar.

Every must-have you state is enforced, so state only the ones you mean. Checks the AI adds on its own are capped at four, and whatever you do not mark as a must-have stays a preference: preferences still rank candidates and still show up in the score, they simply do not reject anybody.

What this changes for your week

When a hiring manager pushes back on a candidate, you send the line from the profile instead of arguing from memory. Requirements the CV never addressed reach the candidate as questions, so people stop dropping out of your process over silence. And an over-specified brief surfaces while you are still writing it, rather than three weeks into a search that was never going to work.

How to use it

1. Read the standard written for each requirement at the brief stage, and change anything that does not match how you would judge it yourself.
2. Mark only genuine dealbreakers as must-haves, because each one you state is enforced. Everything else still counts toward the score as a preference.
3. On a contested candidate, read the evidence column before the number. It is usually one line that settles the argument.
4. Leave the unanswered requirements for screening. Zia asks them per candidate in the conversation, the answers come back to the same record, and the candidate is not moved forward automatically until they are known clean.