# Career-Ops — Market Calibration Benchmarks # # Directional calibration references, not targets. Sources are recruiter-industry # and aggregated survey data (not peer-reviewed); mostly US, cross-industry. # Every metric here is CANDIDATE-side. Employer-side metrics (time-to-fill, # per-process interview pass rates) are deliberately absent — their denominators # don't match anything a candidate can compute from their own tracker, and # comparing across denominators fabricates precision. # # Users can override this file without losing changes on system updates: # 1. --benchmarks flag on funnel-velocity.mjs # 2. config/benchmarks.yml (user layer — copy this file there and edit) # 3. this file (shipped default) # Non-US markets: copy to config/benchmarks.yml and localize ranges/sources. benchmarks: response_rate: denominator: per-application # any employer response / applications sent range_pct: [2, 13] # 2-3% cross-platform typical; platform spread up to ~13% (LinkedIn) typical_pct: 3 source: "HiringThing 2025-2026 application statistics; scale.jobs 2025" year: 2025 caveat: "Mass-application platform data; hand-targeted applications should beat this." application_to_interview: denominator: per-application range_pct: [2, 4] # ~3% (2024), down from 8.4% (2023); ≈30-40 applications per interview typical_pct: 3 source: "HiringThing / resutrack 2026 aggregate" year: 2026 caveat: "Mass-application platform data; hand-targeted applications should beat this." days_first_response: denominator: per-application-with-response # censored: ghosted applications never enter this metric range_days: [5, 14] # typical ~5 days; 40% of recruiters take >2 weeks typical_days: 7 source: "estimate (no citable published aggregate for this window; conservative midpoint of recruiter-survey reporting)" year: 2025 caveat: "Among applications that got a response at all (61% of candidates report post-interview ghosting)." days_interview_to_offer: denominator: per-successful-process range_days: [20, 28] typical_days: 24 source: "NACE Recruiting Benchmarks (2019 — stale, directional only)" year: 2019 caveat: "Old data; treat as order-of-magnitude."