Score Architecture

The methodology
behind the Hiring Reliability Index™.

The HRI is not an opinion score or a weighted average of fields a recruiter fills in. It is a model-driven composite that ingests real operational signals and updates continuously. Here’s how it’s built.

Three tiers of input.

The HRI model ingests signals at three tiers: real-time events (candidate stage transitions, outreach sends/replies), daily aggregates (pipeline velocity, response rate trends), and weekly composites (recruiter load trend, shortlist approval pattern).

Each tier feeds into the composite at a different latency. Real-time events update the score immediately. Daily aggregates update nightly. Weekly composites update on the mandate’s weekly review cycle.

Real-Time Events

Candidate stage transition, outreach reply logged, interview scheduled, offer extended. Each event updates the relevant sub-signal immediately. Score updates within minutes.

Daily Aggregates

Pipeline velocity (stage transitions per day), outreach response rate (trailing 7 days), recruiter activity volume. Aggregated nightly and fed into the model.

Weekly Composites

Recruiter load trend, shortlist approval rate, close velocity. Slower-moving signals that require weekly context to be meaningful. Updated on mandate review cycle.

The model gets smarter with every mandate.

The HRI weighting model was initialised from observed mandate outcomes across Majhi OS’s operational history. It is not static. Every mandate that closes — success or failure — feeds back into the model.

When a mandate with HRI 62 closes in 8 weeks because of a specific recovery action, the model registers that action-plus-signal combination as a positive outcome. When a mandate with HRI 71 stalls because of an undetected signal, the model registers the gap.

The HRI is a compounding intelligence asset. Each mandate makes the model more precise. The more mandates Majhi OS runs, the more predictive the score becomes.

Important distinctions.

The HRI is not an AI black box. Every sub-signal is inspectable. When the score drops, Majhi OS surfaces the specific signal driving the drop — not just the composite number.

It is not a survey. It does not ask recruiters to self-report mandate health. It derives health from operational events, not from human judgment.

Not a Survey

Recruiters don’t rate mandate health. The system infers it from operational signals. Removes reporting bias and lag entirely.

Not a Black Box

Every HRI score is fully decomposable. Clicking the score shows you the five sub-signals and which one caused the change.

Not Static

The model updates continuously. Signal weights evolve as more mandate outcome data is ingested. The HRI of year two is more precise than the HRI of year one.

See the methodology applied to your mandate.

In 45 minutes, Majhi OS will decompose the HRI on your active search and show you exactly which signal is driving your score.

Book Your Mission Walkthrough →