WBS Section 6 of 7

Screening Analytics

Track AI quality, measure efficiency gains, and validate screening outcomes.
Feeds directly into the AI Evaluation Plan's post-launch dashboard.
Start / End
Process
Decision
Output / Display
Jay (Decision Maker)
Screening activity generates data Aggregate screening events Scores, overrides, timestamps, approvals, rejections, failures Three metric streams: ① Override Tracking • Override rate (%) • Avg score delta (±) • Direction: up vs. down • Distribution by score range Override > 30%? Yes ⚠ UI Warning Review rubric/prompt Pattern detection: Consistently ↑ = rubric too strict Consistently ↓ = rubric too lenient ② Screening Throughput • Resumes processed / role • Avg time: upload → shortlist • Resumes / hour rate • Parse & scoring failure rates Baseline comparison Manual: 80–100 hrs/hire vs. actual AI-assisted time ③ Funnel Conversion • % advancing each stage • Approved → Phone → Tech → Panel → Offer → Hired • Breakdown by AI score range Calibration signal: Do 9–10 candidates convert higher than 7–8? If not → Leo's Analytics Dashboard All 3 streams unified Filterable by role + time period CSV export available 🔗 Feeds into AI Evaluation Plan Post-launch 8-metric dashboard Jay's Executive Summary Accessible from main dashboard — no drill-down required Jay opens dashboard Executive Summary View Total candidates screened Estimated hours saved Override rate · Pipeline conversion