Screening analytics
The screening funnel, ranking, quality, and skill matrix — the views that turn your screening activity into a hiring signal.
Screening analytics
The Screening analytics view is the analytics surface for the screening console. It is the answer to "is the bar working?" — across many candidates, not just one.
The funnel
The screening funnel shows the count of candidates at each verdict stage. It is the right place to see how a hiring round is going at a glance.
The funnel also shows the machine suggestion funnel, which includes
INSUFFICIENT_DATA. Comparing the two is the easiest way to see how often the
machine and the humans agree.
Completeness
The completeness panel shows how many candidates are verdictable (their evidence is complete) versus incomplete (still waiting on something). The completeness rate is the fraction of candidates in your directory that you can decide on right now.
A low completeness rate is an early signal: it means the platform is doing the right thing by not deciding on incomplete evidence. A high completeness rate on a fresh directory means the assessments are landing clean.
Detection rate
The detection panel shows the fraction of candidates with at least one integrity flag, and the distribution of flag types. A high detection rate is not a bad thing on its own — the bar is meant to surface concerns. What matters is whether the flags correlate with low quality and whether recruiters are acting on them.
Review load
The review load is the count of candidates in REVIEW — the queue waiting for a
second look. A growing review load is a sign that the bar is escalating more
candidates than your team can process. The right response is either to expand the
team, sharpen the bar, or both.
Ranking
The ranking view lists top candidates by best score, with their current verdict,
suggested verdict, effective recommendation, and the count of assessments they have
completed. Use it to find the highest-scoring candidates who are still NEW and
need a decision.
Quality distribution
The quality distribution shows how candidates cluster on quality_aggregate. A
bimodal distribution is normal: the assessment separates the strong candidates from
the weak ones, and the middle is small. A flat distribution is a sign the assessment
is not separating.
Skill matrix
The skill matrix shows, per role title, the top candidates and their best score. It is the right view to find a strong candidate for a specific role from across your full directory.
Reviewer calibration
The calibration view measures how consistently your reviewers apply the same bar to the same evidence. It compares, over a configurable window, how often reviewers diverge from one another and flags individual decisions that fall well outside the group.
| Section | What it tells you |
|---|---|
| Divergence | How far apart reviewers land on comparable candidates. |
| Reviewers | Per-reviewer decision patterns, so a single reviewer's skew stands out. |
| Outliers | Individual decisions that sit far from the rest. |
A high divergence rate is a signal worth acting on — usually it means the bar needs sharpering, or the team needs to re-align on what each verdict means — rather than a reason to distrust the data.
Compliance exports
For governance and audit needs, you can export two records of how decisions were made:
| Export | What it contains |
|---|---|
| Decision audit | The full, machine-readable trail of screening decisions — who decided, when, what changed, and the reason. |
| Bias audit | A consistency summary over a window, for reviewing that the bar was applied without skew. |
Each export is prepared as a job; when it is ready you download the artifact. These are the same underlying records the audit trail describes, packaged for a reviewer or a regulator to inspect.