What it is
The scoring system is what turns a finished interview transcript into a structured result. It runs on GPT-4o with a calibration layer added specifically to keep the grading consistent, since a model grading conversations without that layer tends to drift toward generous scores over time.
Each of the five dimensions gets its own 0-10 number and its own written reasoning explaining what specifically drove that number, not a single blended score with no explanation behind it. Overall reflects the interview as a whole; Communication, Technical, Behavioral, and Cultural Fit each isolate one part of the conversation.
How it works
- 1
The interview finishes
A completed session, across any of Intervieux's seven interview types, gets passed to the grading system automatically.
- 2
Each dimension gets scored
Overall, Communication, Technical, Behavioral, and Cultural Fit are each scored 0-10 based on the transcript.
- 3
Calibration floors and caps apply
Each dimension's score is checked against guardrails that prevent an unusually confident or unusually flat delivery from pushing a number past what the actual content supports.
- 4
Penalties get applied where they're earned
Specific deductions apply for things like an abrupt ending, skipped STAR structure, a filler-word ratio over 10%, no candidate questions asked, or type-specific issues.
- 5
Written reasoning gets attached
Each dimension's number comes with an explanation of what drove it, not just the score by itself.
- 6
The result becomes the report
The finished scores and reasoning form the interview report a candidate reviews, and for job-board screening interviews, the data an employer's rules engine can act on.
Why calibration exists at all
A grading model left alone tends to score more generously the longer it runs, or inflate a score because an answer sounded fluent even when the substance was thin.
Intervieux builds per-dimension floors and caps into the grading step specifically to fight that drift, so a 9 on Communication means roughly the same thing in one candidate's report as it does in another's, rather than depending on how the scoring happened to trend that day.
Penalties are specific, not vague
Rather than a generic deduction for a bad interview, the penalty system names exact issues: ending the interview abruptly costs points outright, an answer that skips STAR structure loses points, a filler-word ratio above 10% is flagged and penalized, and not asking the interviewer any questions at the end counts against the score too.
Certain interview types carry their own additional, type-specific penalties on top of those shared ones.
Five separate numbers, not one blended grade
Overall is its own dimension, not just an average of the other four. A candidate can see specifically where a session was strong, technical reasoning, say, versus where it needs work, cultural fit questions, rather than one composite score that hides which part of the conversation actually struggled.
Where the scores go next
For a job-board screening interview, an employer's rules engine can evaluate directly against all five score dimensions, alongside a challenge score, experience level, salary estimate, key skills, and the strengths and weaknesses noted in the report.
For practice interviews, a candidate's own analytics compare their scores against the platform average per dimension, and completing enough strong sessions earns achievements like Top performer at 90% or better.
Who it's for
Who it's for
Every candidate who completes an Intervieux interview gets this scoring automatically, whether it's a practice session or a real job-board screening interview. It matters most to candidates who want to know specifically what to work on rather than a single pass or fail signal, and to employers who need a scoring standard consistent enough to compare candidates against each other fairly.
Frequently asked questions
What are the five interview scoring dimensions?
Overall, Communication, Technical, Behavioral, and Cultural Fit. Each is scored 0-10 with its own written reasoning explaining what drove that specific number.
How does calibration stop scores from being inflated?
Each dimension has floors and caps built into the grading step, which keep a confident or fluent-sounding answer from scoring higher than its actual content supports, and keep one rough stretch from sinking a score that held up everywhere else.
What specific things get penalized in interview scoring?
An abrupt ending, an answer that skips STAR structure, a filler-word ratio above 10%, not asking the interviewer any questions, and certain type-specific issues depending on which of the seven interview types was run.
Can an employer see my full scoring breakdown?
For a job-board screening interview, yes: an employer's rules engine can evaluate against all five score dimensions plus a challenge score, experience level, salary estimate, key skills, and noted strengths and weaknesses. Practice interview scores stay with the candidate.
Is Overall just an average of the other four dimensions?
No. Overall is its own scored dimension with its own reasoning, not a simple average, so it can reflect the interview as a whole even where the other four dimensions vary.
Related pages
See exactly how you scored, dimension by dimension
Complete an Intervieux AI interview and get a five-dimension score with written reasoning, calibration guardrails, and a specific accounting of any penalties applied.