> Kepler scores your entire database against the brief and shows the evidence behind every score — criterion by criterion, sourced to the CV, call or email.

_Kepler product feature — AI Candidate Matching with Evidence. Canonical: https://keplercrm.com/features/ai-matching/_

# A ranked shortlist the moment the brief lands. With receipts.

Not a black-box percentage — a criterion-by-criterion verdict you can defend to your client.

## How it works.

### How a brief becomes a shortlist

Paste the client's brief and Kepler turns it into a scorecard, scores everyone in your database against it, and hands you the ranking — each verdict tied to the document that earned it.

**The spec becomes a scorecard** — Paste the brief and the criteria and weights are pulled out for you to tune — musts, highs and mediums, editable before a single CV is read.

**Scored against everything you know** — The judge reads CVs, notes and conversations — not just profile fields — and every verdict cites the exact source it came from.

**Ranked strongest first** — Open the tab and the best people are already at the top, each row carrying the segmented meter that shows where a score was lost.

### You stay the recruiter

The ranking arrives done, but never final — hard filters run before scoring, your pins override the order, and the whole model sits behind a standing evaluation program.

**Agree or overrule** — Accept the ranking, or drag someone up and move on — the score explains, you decide. Your pin sticks; the reasoning stays attached.

**Filter the pool first** — Hard limits — salary, location, right to work — apply before scoring, so nobody spends the day on people who could never take the job.

**It's measured** — Matching runs under a standing evaluation program: recall and precision gates clear before any change to scoring ships.

## Inbound, scored on arrival

Applicants from your hosted job board land parsed, deduped and scored against the same rubric as the rest of the pool — most turn out to be already in your database.

## Scores you can say out loud.

A match score is a claim about a person. Kepler attaches the proof, so the number survives contact with your client.

- Every number traces to evidence. Unpack any percentage and you're reading criterion verdicts, each sourced to the CV, call or email that earned it.

- Clients get the reasoning, not just the ranking. The shortlist you send carries the why — defensible in the first client call, not after it.

- The score explains; you decide. Pins, pre-filters and overrules are yours. Nothing ranks your shortlist but you.

- It admits ignorance. Nothing in the record for a criterion? The verdict is Not stated — it never guesses to fill a gap.

## Questions, answered

### How does Kepler score candidates against a job?

Paste the client’s brief and Kepler pulls the criteria and weights out of it as a scorecard — musts, highs and mediums — which you tune before a single CV is read. It then scores everyone in your database against that scorecard and ranks the strongest first.

### Can I see why someone scored what they scored?

Yes, and that is the point. Unpack any percentage and you are reading criterion-by-criterion verdicts, each one sourced to the CV, call or email that earned it — so the shortlist carries its reasoning into the first client call.

### What happens when a record says nothing about a criterion?

The verdict is "Not stated". It never guesses to fill a gap, and the segmented meter shows you exactly where a score was lost rather than hiding it inside one number.

### Can I overrule the ranking?

Always. Hard filters — salary, location, right to work — run before scoring so nobody wastes a day on people who could never take the job, and your pins override the order afterwards. The score explains; you decide.

### Is the matching quality actually measured?

Yes. Matching runs under a standing evaluation program, and recall and precision gates have to clear before any change to scoring ships.