Appearance
How scoring works
Available onEvery plan
Infyva scores what the candidate said against the competencies the job asks for. The scoring model, its depth, the integrity checks and question writing are the same on every plan. Plans only add ways to view the results.
AI scores are decision support. A person at the hiring company makes the hiring decision.
How each answer is graded
- Each answer is graded per competency against behavioural anchors, levels 1 to 5. Level 3 is the bar for the job's seniority.
- The grader quotes the answer word for word first, so every level points to evidence you can read.
- Every answer is graded three times and the median is used. Rows where the three grades disagreed are marked Graders disagreed on the scorecard.
The grader never sees the candidate's name or resume.
How the overall is worked out
- A competency's level is the mean of the levels of the answers that gave evidence for it.
- The overall is the weighted mean of the competencies that had evidence. The weights are the ones your team set on the scorecard, or equal weights if you set none. A model never chooses the weights.
- Competencies with no evidence are left out. Their weight is shared among the other competencies. They never count as zero. The scorecard shows No evidence in this interview. Not scored; its weight is shared among the other competencies.
Open How this score was calculated on any scorecard to see this for a candidate.
From level to score
The overall level is shown as a score out of 100 with this mapping:
| Level | Score |
|---|---|
| 1 | 10 |
| 2 | 30 |
| 3 | 55 |
| 4 | 75 |
| 5 | 92 |
The bands
Every report gets one band. The band is set on the overall level:
| Band | Rule | About this score | What it means |
|---|---|---|---|
| Strong | Level 4.0 or more, and no assessed competency below level 3 | 75 to 100 | At or above level 4 overall, with no competency below level 3. |
| Meets | Level 3.0 or more, and at most one assessed competency below level 3 | 55 to 74 | At or above level 3 overall, with at most one competency below level 3. |
| Mixed | Level 2.5 or more | 43 to 54 | Just under the bar: some solid answers and some clear gaps. |
| Below | Under level 2.5 | 0 to 42 | Well under the bar: the answers fell short on most of what was assessed. |
| Borderline | Within 0.15 of a band line (2.5, 3.0 or 4.0), or the graders disagreed on competencies carrying more than 20% of the weight | Too close to a band line to call. Review the evidence before deciding. | |
| Limited evidence | Fewer than half the competencies had evidence | Too few competencies had evidence to judge. Treat the score as provisional. |
When the graders disagreed on more than a quarter of the weight, the scorecard is marked Provisional.
Reports scored before anchor levels existed keep their older bands: Strong 75 and up, Meets 60 to 74, Mixed 50 to 59, Below under 50.
Suggested next step
Each band suggests a next step on the report:
| Band | Suggested next step |
|---|---|
| Strong | Advance to the next round |
| Meets, Mixed, Borderline, Limited evidence | Follow up before deciding |
| Below | Unlikely to advance on this evidence |
| Any band with integrity signals | Review the integrity signals before deciding |
| Held report | Insufficient evidence, needs human review |
Bands are frozen
The band, next step and verdict are saved on the report when it is graded. If Infyva later changes where the band lines sit, past reports don't move. Only a re-grade writes a new band. See Retry and re-grading.
Every screen reads the same saved verdict: the report, the job's candidate list, the hiring manager page and email, Compare and the PDF. A candidate never shows as Strong in one place and something else in another.
What is not evaluated
- Scoring does not consider appearance, voice tone, accent, fluency or emotions.
- Nothing reads or infers personality or demographic data.
- The grader never sees the candidate's name or resume.
- An answer that gives instructions to the grader instead of answering earns no credit for that part. The report shows it as Instructions to the grader so you can listen to the recording.
- In a Live AI report, Behavioral signals are shown for context and are not part of the overall score.
For the full explanation, read How Infyva AI scores interviews.