Methodology
How Mentari scores an interview
Mentari scores every interview on six competencies, each out of 10, alongside an overall score out of 10. The rule that makes the scorecard trustworthy is simple: no quote, no score. Every competency score must cite verbatim quotes from your own answers, and each quote is automatically verified against the transcript before you see it.
The six competencies
| Competency | What it measures |
|---|---|
| Communication | Whether your answers were clear, direct and easy to follow — not your accent. |
| Technical depth | How well you understand the substance of the role, beyond definitions. |
| Problem solving | How you break down an unfamiliar problem and reason towards an answer. |
| Ownership | What you personally did, decided and were accountable for. |
| Structure | Whether your answers had a shape: context, action, result, and a point. |
| Company fit | Whether you showed the signals this particular company looks for in this role. |
The evidence rule: no quote, no score
A score such as “communication: 6/10” on its own tells you nothing and cannot be argued with. So the evaluator is not allowed to produce one. For each competency it must point to between one and three things you actually said.
Those quotes are then checked by code, not by another opinion: each one has to appear in one of your own turns in the transcript. The check tolerates the punctuation and casing noise that speech-to-text introduces, and nothing else. A quote that cannot be found in what you said is removed — the evaluator cannot support a score with words you never spoke, or with the interviewer’s words.
What is on the scorecard
| Part | What it is |
|---|---|
| Overall score | Out of 10. |
| Six competency scores | Each out of 10, each with a short rationale and the quotes behind it. |
| Strengths | Up to three — and none at all if the interview did not show any. The evaluator is not required to invent praise. |
| Risks | Up to three things a real interviewer would flag. |
| Drill list | What to work on before the next attempt. |
| Progress | The trend across attempts. The change between a first and third interview is usually the most useful number. |
What the interviewer is doing while you answer
The round is planned before it starts, from your résumé, the job description and research on the company. During the interview the AI tracks which planned topics have really been asked and answered — only a question that was actually put to you counts, and when in doubt the bookkeeping errs towards treating a topic as not asked, so that you are not marked down for a question nobody put to you.
It follows up when an answer is vague, asks for the number, and backs off with a hint when you are stuck. If you say you have not worked on something, it moves to a different thread instead of scoring you on it.
Consent and visibility
You consent before anything is recorded, and you always see your own report. If a college enrolled you, its placement cell sees readiness and scores for the interviews it arranged, scoped to its own students. Anything you buy yourself is yours alone. Data can be exported or deleted on request.
For placement cells
The same evidence rule is what makes cohort data usable: when a student disputes a score or a recruiter asks why someone was flagged, you open the transcript at the quoted moment instead of describing an impression. Mentari for placement cells.
Last updated · Written by the Mentari team