Loading Page
Loading Page
We use cookies to keep you signed in and to remember your settings. We don't use analytics or advertising cookies.
Report structure
Reviewers know where to look, and two candidates for the same role can be read side by side.
Candidate, position, interview level, date, length and the panel. Taken from the interview record, not written by the AI.
What happened and what was shown, in at most 120 words.
Five areas scored 1 to 5, each with a reason of at most 60 words and up to 3 quotes.
Every question the candidate was asked, including ones asked out loud, with a summary of the answer, how well it was answered and the quotes behind it.
What the interviewers' scorecards say, where they disagree with each other or with the transcript, and every scorecard side by side.
Specific to the job and traceable to the transcript, the notes, the feedback or the code.
In technical rounds: what the final code and its runs show about correctness, approach and testing.
The integrity signals, stated as facts with their times, and a reminder that they can be false alarms.
What a follow-up round should probe, given the gaps or doubts in this one.
A suggested recommendation, shown last and labelled as a suggestion.
Kept apart
Three different things, shown in three separate places. Integrity signals never lower a score, and the AI's suggestion never becomes the decision.
Scored from what the candidate said, with quotes.
“I'd return an empty list first.” 13:05
Listed with times, for a person to check in the recording.
Automatic signals can be false alarms: lighting, glasses, a second screen for notes. Check the recording before drawing any conclusion.
Not scored. Not part of the average.
Made by the reviewer, next to the AI's suggestion.
Finalising locks the report and the panel's feedback.
Definitions
The same definitions apply to every report, scaled to the role's experience range and the interview level. An L1 screen is judged less deeply than an L3.
Every question the candidate was asked gets one of four ratings, with the quotes behind it:
Not Answered is its own rating, so a skipped question is never mistaken for a weak answer.
Strong Yes, Yes, No or Strong No, only when the evidence clearly supports it. Otherwise the suggestion is Not Enough Evidence. The performance band works the same way: it reads Not Enough Evidence when fewer than 3 areas could be scored or the candidate spoke for less than about five minutes in total.
The scorecards are an input, not the answer: the report scores from the transcript and points out where the panel and the transcript disagree.
Review and sign-off
Once the transcript is ready. No draft is made if the candidate didn't agree to AI help, if your company has AI reports turned off, or if there is no transcript.
A reviewer checks it against the recording and edits any part. The AI's own version is kept unchanged beside the edits. The report can be drafted again, keeping the edits unless the reviewer discards them.
The reviewer chooses their own final recommendation. The report and the panel's feedback are then locked, and the activity log records who finalised it and when.
Panel members allowed to read reports see it only once it is final, so nobody relies on an unreviewed draft. It can be saved as a PDF.
Analytics
Dashboards for every role, over the last 7, 30 or 90 days or any period up to 366 days. Counts are worked out in the database for your company only.
| Area | Your Ratings | Whole Panel |
|---|---|---|
| Technical Knowledge | 4.2 | 3.6 |
| Problem Solving | 3.1 | 3.4 |
| Reasoning and Decisions | 3.8 | 3.5 |
| Communication | 4.4 | 3.7 |
| Ownership and Role Fit | 3.0 | 3.6 |