calhire
Capability

Hiring analytics, benchmarks and talent rediscovery

Three related things: knowing where your funnel loses people, knowing whether that is normal, and remembering the good candidates you already met.

CalHire’s analytics cover three areas. Funnel metrics show where candidates enter, stall and drop out of your own hiring process. Benchmarks compare those metrics against aggregated data from other employers, published only where the cohort is large enough that no individual employer or candidate can be identified. Rediscovery surfaces strong candidates already in your pipeline whose verified skills match a newly opened role.

Last reviewed

The short version

  • Funnel metrics show drop-off by stage, which is usually where a hiring problem actually lives.
  • Benchmarks are only published above a minimum cohort size, because a benchmark computed from three employers describes three employers.
  • Rediscovery reuses assessments you already have, so a strong candidate for a closed role is not lost.
  • Analytics never expose another tenant’s data, and never re-identify a candidate.

The metric most teams track, and the one that matters

Time to hire is the number almost every hiring team reports, and on its own it is close to useless. You can improve it by lowering your bar, by hiring the first adequate candidate, or by counting from a later starting point. All three look like progress on a dashboard.

The useful version is drop-off by stage. A funnel that loses 80 percent of candidates at the assessment invitation is telling you something specific: either the invitation is unclear, the assessment looks too long, or the role attracted people who were never going to complete it. A funnel that loses candidates at offer is telling you something entirely different and considerably more expensive.

Stage-level data also makes fairness measurable. Adverse impact usually concentrates at one stage rather than spreading evenly, and you cannot find that stage from an aggregate. This is the same data the compliance console uses for selection-rate analysis, viewed for a different purpose.

What you can see

  • Stage-by-stage drop-off

    Where candidates enter, where they stall, and where they leave. The shape of the funnel rather than a single duration.

  • Time in stage

    How long candidates wait at each step. Waiting is the most common cause of losing a good candidate and the easiest to fix once it is visible.

  • Score distributions

    How your applicant pool actually scores against your requirements. A role where everyone clears the bar is a role whose bar is not doing anything.

  • Selection rates by stage

    The same data the adverse-impact analysis reads, so a fairness question and a performance question are answered from one source.

  • Benchmarks against other employers

    Aggregated, anonymised, and published only above a minimum cohort size. Useful for answering whether a number is bad or simply unfamiliar.

  • Rediscovery

    Candidates already assessed, matched against a newly opened role. Their verified score is still valid, so the expensive part is already done.

How benchmarks stay safe to publish

Cross-tenant benchmarks are genuinely useful and genuinely risky. Aggregate carelessly and you can reconstruct an individual employer’s hiring data, or in a small enough segment, an individual candidate’s result.

The protection is a minimum cohort size. A benchmark is only computed and shown once enough employers and candidates sit behind it that no single one materially moves the number. Below that floor the benchmark is suppressed rather than shown with a caveat, because a caveat next to a number does not stop anyone reading the number.

This means some segments have no benchmark, particularly narrow roles in small markets. That is the correct outcome. A comparison drawn from four companies is not a market rate, and presenting it as one would be worse than presenting nothing.

How rediscovery works

  1. 1

    A role opens

    With its required skills and weights, the same as any other.

  2. 2

    Existing verified profiles are matched against it

    Candidates who applied to other roles, or who are in your pool, and whose verified scores are still within validity.

  3. 3

    You review matches anonymously

    Same anonymous surface as any other candidate. Previous history with your organisation does not identify them to a reviewer.

  4. 4

    You invite, they choose

    An invitation, not an automatic re-entry into a pipeline. A candidate who declined you once is entitled to decline again without explanation.

What this does not do

Analytics do not explain causes. A drop-off at one stage tells you where to look, not what is wrong. The answer usually requires talking to candidates, and the report cards give you a reason they might be willing to reply.

Benchmarks are not available for every segment, by design. Narrow roles and small markets fall below the cohort floor and show nothing rather than something unreliable.

Rediscovery only reaches candidates with a valid verified profile. Someone whose score has expired needs to retake before their profile carries a current result, and no amount of matching can substitute for that.

None of this measures the candidates who never applied. The most consequential bias in most hiring processes happens before anyone reaches your funnel, and no funnel analytic can see it. Where the job was posted and how it was written matter more than anything on this page.

Questions people actually ask

Can we see other employers’ hiring data?
No. Benchmarks are aggregated across employers and only published above a minimum cohort size. No view exposes an individual tenant’s data.
Why is a benchmark missing for our role?
Because the cohort behind it is too small to publish safely. Suppression is deliberate: a number drawn from a handful of employers would be read as a market rate while being nothing of the kind.
Does rediscovery contact candidates automatically?
No. It surfaces matches for you to review, and any approach is an invitation the candidate can decline. Automatically re-entering people into pipelines they left is not something we build.
Can analytics identify individual candidates?
Not in benchmarks, which are aggregated above a cohort floor. Within your own tenant you see your own pipeline, subject to the same anonymity rules as everywhere else: no identity before a consented reveal.
Do these metrics feed the compliance reporting?
They read the same underlying stage and selection data. Adverse-impact analysis applies its own thresholds on top, including the minimum disclosed cohort of 30 and the minimum group size of 5.

Where this connects to the rest of the platform.

Read the reasoning

The evidence and the argument behind what is on this page.

Hiring Playbook8 min read

Time-to-hire is a throughput metric. Quality-of-hire is the one that matters

Speed metrics are easy to collect and easy to game. How to measure whether your hiring decisions were any good, and which metrics to pair with which.

Read
Hiring Playbook7 min read

Why candidates get ghosted, and how to actually stop

Ghosting is almost never malice. It is what happens when closing the loop is optional, unowned and manual. The four structural causes, and the fix for each.

Read

Browse all topics on the blog

See a verified pipeline for one of your roles

Post a role free and review anonymous, skill-ranked candidates. No card, no sales call to get started.