Six numbers tell a hiring team whether its interviews are working: the spread between interviewers scoring the same competency, the relationship between interview scores and later performance, pass-through rates by stage and by group, the time from last interview to decision, candidate drop-off per stage, and interviews per hire. Everything else an analytics dashboard can show, from total interviews conducted to minutes saved by AI, is decoration. This page defines the six, explains where each comes from, names the vanity metrics to ignore, and shows how a team turns the numbers into changes to its rubrics and training.
The six numbers
| Metric | Definition | What it tells you | Review |
|---|---|---|---|
| Score spread | Range of scores different interviewers give the same answer or candidate on one competency | Whether rubric anchors are clear | Quarterly |
| Score-to-outcome | Correlation between interview scores and 6- or 12-month performance ratings | Whether the rubric measures what matters | Twice a year |
| Pass-through by stage and group | Share of candidates advancing at each stage, split by demographic group where lawful | Adverse impact and bottlenecks | Monthly |
| Interview-to-decision time | Days from the final interview to an offer or rejection | Debrief and decision discipline | Weekly |
| Drop-off per stage | Candidates who withdraw or go silent at each step | Process friction and candidate experience | Monthly |
| Interviews per hire | Total interviews conducted divided by hires made | Screening quality and loop length | Quarterly |
The first two require structured scores with evidence; the rest need timestamps and stage data most applicant tracking systems already hold. Benchmarks for several of them are collected in AI in hiring statistics for 2026.
Vanity metrics to ignore
- Total interviews run with AI. Volume says nothing about quality.
- Minutes saved on notes. Real but small, and it does not change who you hire.
- AI accuracy. Accuracy against what? If the vendor cannot name the ground truth, the number is marketing.
- Average candidate score. Drifts with the rubric, the role mix and the season.
- Dashboard logins. Adoption of the dashboard is not adoption of structured interviewing.
Where the numbers come from
Score spread and score-to-outcome exist only if every interviewer scores the same competencies on the same scale with evidence notes. That is what an interviewer copilot produces during the call; CoPilot Interview's Interviewer Mode writes rubric-based scores with evidence notes into a report per interview, and team plans add a centralized evaluation dashboard that collects those reports across interviewers and roles. Pass-through, drop-off and timing come from the ATS. Interviews per hire is a division you can do in a spreadsheet. The team and enterprise page describes the central view; most teams start with the free app and a shared folder.
Turning numbers into changes
- Spread above two points on a ten-point scale: rewrite the anchors for that competency and run a calibration session; the method is in how to calibrate interviewers.
- No score-to-outcome relationship: the rubric is measuring interview skill, not job skill. Change the competencies before changing the tool.
- Pass-through differs by group at one stage: audit that stage's questions and scoring inputs first, then the interviewers.
- Decision time above a week: schedule the debrief when you schedule the final interview.
- Drop-off spikes at one stage: ask the next ten candidates who withdraw why; the answer is usually scheduling or silence.
- Interviews per hire climbing: the screen is passing people the loop rejects; tighten the first round.
Where CoPilot Interview fits
CoPilot Interview is the AI interview assistant candidates use to prepare; Interviewer Mode is the same desktop app on the hiring side, producing the structured scores and evidence notes that make the first two metrics possible, free on Windows and macOS with no signup, with a centralized evaluation dashboard on team plans. See Interviewer Mode. Whatever you measure, disclose scoring to candidates and follow your employer's rules on retention and demographic data.
Disclosure: CoPilot Interview is our product, so weigh what we say about it accordingly.
Frequently asked questions
What should hiring teams measure about interviews?
Score spread between interviewers, score-to-outcome correlation, pass-through by stage and group, interview-to-decision time, drop-off per stage and interviews per hire.
Which AI interview metrics are vanity metrics?
Total interviews run, minutes saved, unexplained accuracy figures, average candidate score and dashboard logins. None of them changes who you hire.
Do I need an enterprise dashboard to track these?
No. The two score metrics need structured per-interview reports; the rest come from your ATS. A central dashboard saves collation time once you pass a few interviewers.
How often should we review interview analytics?
Decision time weekly, pass-through and drop-off monthly, score spread and interviews per hire quarterly, score-to-outcome twice a year.
Does Interviewer Mode include analytics?
The free app produces a structured report per interview. The centralized evaluation dashboard that aggregates them across a team is part of team plans.
Try Interviewer Mode on your next screen
Role-specific questions, live rubric scoring with evidence notes and a structured report at the end of the call. Free in the desktop app for Windows and macOS, no signup.
Get the desktop app