Candidate Scoring

Rubric-based fit scoring — same questions, same scale, fair comparison.

Recall bias is real. By the 10th candidate you can't remember what the 3rd one said. Raffi scores every candidate against the same rubric so when you compare, you're comparing the same thing.

What you see in the shortlist

  • Overall fit score (0–100) against your hiring bar
  • Per-criterion breakdown (technical, communication, culture, experience)
  • Top 3 strengths + top 3 gaps in plain English
  • Recommendation: strong fit / fit with caveat / not a fit
  • Side-by-side comparison view across the shortlist
What it does

The capability, broken down.

You set the rubric — Raffi scores against it

When you create a role, you define 4–6 criteria that matter. 'Cold outbound sales experience.' '3+ years of TypeScript.' 'Bilingual EN/AR.' Raffi turns each into questions, scores answers on a 1–5 scale per criterion, and produces the weighted overall fit score.

Same questions across every candidate

The Hire Bar of Bias problem: when you ask candidates different questions, you can't compare them fairly. Raffi asks the same core behavioral questions to every applicant for a role — then probes the answers individually. Same anchor, fair comparison.

Strengths + gaps, in plain English

Numbers are useful but rarely actionable. Every candidate scorecard ends with 3 strengths (with specific quotes from the interview) and 3 gaps (with the exact moments they showed up). You can read the scorecard in 90 seconds and know what to dig into in your final round.

Side-by-side compare across the shortlist

Select 3 candidates from your ranked list, hit Compare. You see all three scorecards side-by-side — same criteria, same scale. The decision is faster, the bias is lower, the hire is better.

Under the hood

Technical specifics.

Score scalePer-criterion 1–5, weighted to overall 0–100
Criteria per roleConfigurable, default 4–6
RecommendationStrong fit / Fit with caveat / Not a fit
Evidence linkingEvery score links to the specific transcript moment
CalibrationAnchored against ground-truth ratings on 100k+ interviews
Override flowRecruiter can adjust scores; model learns from overrides
FAQ

Candidate Scoring — questions, answered.

How is this different from generic resume-screening AI?

Resume screening rates the CV — what the candidate claims. Raffi rates the interview — what the candidate demonstrated. The CV gets you to the interview; the scorecard tells you whether to extend an offer. Both matter, but only one tells you what actually happens when you ask hard questions.

Can I see the prompts Raffi uses to score?

Yes — on every scorecard, you can expand the per-criterion view to see exactly what Raffi was looking for in the answer and what the candidate said. No black box.

What if I disagree with a score?

You can adjust it. The transcript is right there; you can see exactly what was said. Adjusted scores feed back into Raffi's calibration so the model gets sharper for your specific role over time.

Does Raffi remove bias completely?

No, and we don't claim to. What it does: it removes recall bias (you don't have to remember 30 calls), it removes ordering bias (the candidate you interviewed first doesn't get more attention than the 30th), and it makes the criteria explicit so you can audit them. The remaining bias is in the rubric you set — which is a problem worth having, because at least it's visible.

Get started

Hire your next role with Raffi.

Start free with $25 in credits — no card required.

Get started free