Aller au contenu
Toutes les missions
Veille marché · Dev-to Nouvelle Publiée il y a 11 heures

One in four new job postings is gone within two weeks: 131,119 closures on 11,600 career sites

A careers API tells you a job is open. It never tells you it closed: the posting is in the list one day and missing the next. We read about 11,600 employer job boards every day, on seven app...

Entreprise left. By
Mode g is in the list one day and missing the next

Résumé

A careers API tells you a job is open. It never tells you it closed: the posting is in the list one day and missing the next. We read about 11,600 employer job boards every day, on seven applicant tracking systems (Greenhouse, Ashby, Lever, Personio, SmartRecruiters, Teamtailor and Recruitee), and since 29 October 2026 we have written down every posting that left. By 16 November that was 131,119 closures. This is what they say. 1. Jobs close on working days, and you can see it the next morning Our refresh runs at about 02:15 UTC and records what was removed since the one before. In the fourteen refreshes from 3 to 16 November, 99,998 postings closed: Refresh on a Refreshes Closures, lowest to highest Average Tuesday to Saturday 10 8,059 to 10,264 9,446 Sunday or Monday 6 (from 1 November) 1,188 to 1,695 1,477 A Tuesday-to-Saturday refresh sees the removals of a working day; a Sunday or Monday refresh sees a weekend day. Employers take down six times as many postings on a working day. If you clean a job board once a week, do it on Saturday: by then the week's closures are all in. Against an index of 505,077 open jobs on 16 November, a working day removes about 2% of everything listed. 2. One in four new postings is gone within two weeks To measure how long a posting lives you need to have seen it arrive. We took every posting that first appeared between 30 October and 2 November on a board we were already reading, on the five systems we have read since October, and followed it to 16 November. Four boards had to be set aside first. Between them they produced 9,275 of the 21,771 "new" postings of those four days, and they behave like nothing else: two recruiting-agency boards on Lever listed 1,024 and 1,056 new postings and had removed every one of them within fourteen days, one restaurant chain's board listed 6,628, and a fourth board 567. Leave them in and they are the whole result. On the other 2,406 boards, 12,496 new postings appeared in those four days: Gone within Share of new postings 3 days 3.9% 7 days 12.8% 14 days 23.3% By system, the fourteen-day figure is close to the same everywhere we had enough to count: Greenhouse 20.5% (4,928 postings on 1,256 boards), Lever 22.1% (380 on 128), Ashby 23.7% (1,249 on 490), SmartRecruiters 26.1% (5,776 on 425). Personio was lower, 11.0%, on only 163 postings. So a list of jobs scraped two weeks ago is wrong about roughly one new posting in four, before counting the ones that were edited. 3. What "closed" means here, and what it does not A posting counts as closed when its board was read in full at a refresh and no longer lists it. A board that could not be read closes nothing that day. A posting that comes back under a new id with the same title, place and requisition number is treated as the same job, not as a closure and a new job; we wrote up why in an earlier piece. A posting re-listed with a changed title still counts as closed and new. Closed is not filled. An employer removes a posting when the role is filled, cancelled, paused or moved to another board, and nothing public says which. Two limits of the measurement. Our records start on 22 October, so nothing here speaks about postings that live for months: the figures are about the first two weeks only. And "first appeared" is the day we first saw the posting, at most a day after the employer published it. The tool The closures are published daily as Closed Jobs Feed on Apify Store: one row per removed posting with its id, employer, title, link and the dates it was first and last listed, from the same index as Career Site Jobs Search, so the ids match and you can delete exactly the jobs you took. $0.02 a run plus $0.10 per 1,000 closed jobs: a full working day's closures cost about a dollar. A sample to look at first, with no account: 1,000 of the closures of 13 November as CSV.