How long a job advertisement actually stays open

Across 54 employers, 55,942 postings and 26,042 confirmed removals, the median employer loses a quarter of its live postings within 7 days, half within 22 and three quarters within 48. Those are Kaplan-Meier estimates, so postings still open at the end of the study are counted as still open rather than quietly dropped.

The spread between employers is the more useful half of the result. The middle half of employers lose half their postings somewhere between 15 and 30 days, but the full observed range runs from 5 days to 71. An employer’s own hiring rhythm moves this figure more than the seniority or the field of the role does.

We computed the same number the ordinary way first, and it came out at 50 days rather than 22. The difference is one correction, and it applies to any study that starts watching a job board partway through its life.

If you begin counting today, the postings already on the board are the ones that survived long enough to still be there. Short-lived postings from last month are invisible to you, so the sample is selected for longevity. The fix is standard — each posting joins the analysis at the age it had when it was first seen, rather than at day zero — and here it more than halved the answer.

This matters when comparing figures. A shelf-life number quoted without saying how survivors and late arrivals were handled is not comparable with this one, and is more likely to be too long than too short.

Every night we read entire career boards — not search results — for a set of employers on Workday, SmartRecruiters and Ashby. Reading a board in full is what makes the measurement possible: when a posting is no longer on the board, the employer has taken it down. A keyword search cannot support that inference, because a posting can be missing simply because the search did not match it.

A posting’s age is measured from the date the employer’s own system reports for it, not from the date we first saw it. A posting counts as removed only after it has been absent from two consecutive full reads, because a single missed night can be a throttled server rather than a withdrawn advert.

The sample is 54 large employers whose boards we read for our own members — weighted towards pharmaceuticals, industrial manufacturing and enterprise technology, and towards Europe. It is evidence about how those employers behave. It is not a random sample of job advertisements, and it should not be quoted as one.

A removed posting is not the same as a filled role. Employers withdraw advertisements for many reasons: the requisition is filled, the search is paused, the posting is rewritten and republished, or the board is tidied. What is measured here is how long the advertisement is visible to a candidate, which is the thing a job seeker can actually act on.

Two large sources were deliberately left out. One reports a last-modified date rather than a publication date, so an employer editing a posting would have looked like an employer replacing it. The other is dominated by a single aggregator that rotates its listings on a weekly cycle, which would have published one company’s product decision as a fact about the labour market.

The practical reading is that the first week carries most of the risk. On a typical board a quarter of what you can see today will be gone within it, and on the fastest quarter of employers that happens in under a week for half the postings.

That argues for reviewing new postings on a short cycle rather than saving a long list for the weekend, and for spending the tailoring effort on fewer, better-matched roles. It does not argue for applying faster to more things: a removed posting you were never suited for cost you nothing.