How we score
How AI interview scoring works
AI rates your requirements. Your importance settings weight the overall. Try the calculation with this sample interview.
Methodology checked 30 Sep 2026.
Illustrative sample. Change an importance or drag a rating.
A plain average of the 4 ratings would be 7.3.
(16.0 + 18.0 + 18.0 + 10.0) ÷ (4 + 3 + 2 + 1) = 6.2
From interview to scored report
Read the evidence behind each rating before making a hiring decision.
- 01
Your criteria
Requirements and importance set by your team.
- 02
AI ratings
Interview evidence assessed against each requirement.
- 03
Weighted overall
Ratings multiplied by importance, then averaged.
- 04
Your review
Check the recording, reasoning and individual ratings.
What the numbers mean
A rating against the role's criteria.
The overall summarises the requirements you set. It gives more influence to the ones you marked most important.
6.2 is not a 62% chance of success or a hiring recommendation. Review the individual ratings and evidence.
Compare candidates within the same role and criteria. A different requirement list changes what the number means.
Your priorities determine the weights
A Must Have contributes 4 times as much as a Good To Have.
The weight changes the calculation. Marking a requirement Must Have does not create an automatic rejection rule.
The sample starts at 6.2. Giving all 4 requirements equal weight produces 7.3, rounded to 1 decimal place.
When there is not enough evidence
An unevaluated requirement is recorded separately from a normal rating.
- Some requirements cannot be evaluated
- A cannot-evaluate marker is stored as 11. In a partly evaluated interview, it contributes 3.0 with its full weight. 3.0
- Missing or unusable ratings
- They also contribute 3.0 when other requirements have usable results. They are not treated as evidence of a candidate's ability.
- Every requirement is marked cannot-evaluate
- The overall remains cannot-evaluate and is excluded from the fit score. 11 is a status, not 11 out of 10.
- The weighted calculation cannot run
- If there are no requirements or no usable matched results, the engine keeps the model's overall as a fallback. The source is recorded.
Check missing evidence before shortlisting. An unanswered requirement may need a follow-up interview.
Read the answer behind the score
Separate answer scores help you find relevant moments in the transcript.
Illustrative transcript
"We had a customer whose order was 3 weeks late. I called her before she called us, explained the delay, and refunded the delivery fee."
"The candidate is generally proactive with customers."
An exact transcript phrase can appear as a highlight. A paraphrase is not accepted as a quote.
- Little relevant or concrete detail
- Answer score 1-3
- Relevant, but generic
- Answer score 4-6
- Specific, relevant examples
- Answer score 7-8
- Exceptional, quantified detail
- Answer score 9-10
Answer scores are separate from requirement ratings; they are not averaged into the weighted overall. Empty or untranscribed answers receive no answer score.
Your team makes the hiring decision
Use the score to guide your review. Read the requirement-level reasoning and check the relevant recording before deciding.
AI assessments can be wrong. Candidates can request human review through our privacy process.
Read about AI Ethics and security and privacy.
Need a demo? Have a question?
3 ways to reach a person.