The Ghost Job Report

How long do postings really stay up?

A posting counts as “still advertised” only while it keeps appearing in its source feed — not merely because a page for it still exists somewhere.

• Still advertised 30+ days later23.5%
• Still advertised 45+ days later22.3%
• Still advertised 60+ days later0.1%*

*Floor, not a market rate: our dataset removes postings after 60 days, so we can only observe “60+ days” for postings that crossed that line before removal. The true share is higher.

Where postings linger: by field

Share of postings still being advertised 45+ days after going up. Only fields with at least 150 analyzed postings are shown.

• Retail35.9% of 256
• Sales31.3% of 297
• IT30.9% of 379
• Accounting & Finance28.8% of 250
• Admin27.7% of 260
• Logistics & Warehouse26.1% of 440
• Hospitality & Catering24.6% of 589
• Engineering24.6% of 779

Where postings linger: by city

Same measure, for cities with at least 50 analyzed postings.

• San Francisco, CA8.6% of 93

Methodology — and what we refuse to claim

1. Dataset. Job postings aggregated from major job feeds (Adzuna, Greenhouse boards, and others), deduplicated so the same role syndicated across several boards counts once.
2. “Still advertised” is measured, not assumed. We record the last date each posting appeared in its source feed. The advertised span is that date minus the posting date. A page merely existing doesn't count — the posting has to keep being distributed.
3. Our window is censored at 60 days. We remove stale postings from our own index, so long-tail figures are floors. We say so wherever it applies.
4. Reposting is same employer, same title, 21+ days apart. Cross-board syndication of one posting is excluded by deduplication — this counts genuine re-advertising only.
5. Minimum sample sizes. No category under 150 postings and no city under 50 appears — small samples make dramatic, unfair percentages.
6. No employer is named. A long-open posting has innocent explanations: hard-to-fill roles, evergreen pipelines, forgotten feeds. This report describes the market, not individual companies. Employer-level signals live in our Response Scores, which are based on what happens to real applications and carry a right-to-respond.