What ghost jobs are
Ghost jobs are published vacancy advertisements that the organisation does not intend, at the time of publishing, to fill with anyone: either because the seat is not open, or because it is open but the appointment has already settled on a particular person.
The name suggests deliberate deception, and most of what falls under it is not deliberate. The useful questions are therefore what produces an advertisement with no vacancy behind it, and what it does to the numbers, rather than a verdict on an intention nobody can know.
Four situations that produce the same result
- An advertisement that was never closed. The role was filled and the advertisement was not withdrawn from the job boards. This is the most frequent case and the least intentional. It happens especially where an advertisement is posted in several places and closed in only one.
- An advertisement to build a talent pool. Its purpose is to gather CVs for roles expected to open later. That is useful in itself; the fault is not in doing it but in presenting it as a live vacancy.
- An advertisement as a signal. The organisation keeps advertisements open to appear to be growing, or to keep a presence in its market. The seat does not exist at all.
- An advertisement to satisfy a procedure. The role has been decided for an internal candidate or a named person, and the advertisement is published because the procedure requires publication. The decision precedes the first application received.
The first situation is a procedural error, fixed by procedure. The other three are deliberate decisions, and the question about them is not whether they are legitimate but what they do when they cannot be told apart from real vacancies in the system that reads the numbers.
All four share one feature, and it is the source of all the damage: they are not distinguished in the record. Most recruitment systems have no field saying that an advertisement is not a vacancy anyone intends to fill, so it enters every calculation built on the number of open requisitions or the number of applicants alongside everything else. The remedy therefore lies in the structure of the record before it lies in anyone’s behaviour.
The effect on recruitment metrics can be calculated
This is where the overlooked damage sits. An advertisement that leads to no hire enters the calculation of recruitment key performance indicators without ever producing the hire they are meant to count, so it distorts every figure it enters.
The age of open vacancies
Take an organisation with 14 open requisitions today: 10 genuine ones with an average age of 30 days, and 4 that nobody intends to fill, each open for 180 days. The figures are assumed to show the structure.
- The reported average: 10 times 30 plus 4 times 180, divided by 14, which is 1,020 divided by 14, or about 72.9 days.
- The average over genuine vacancies only: 30 days.
The reported figure is about 2.43 times the real one, and four advertisements out of fourteen did that. An organisation reading this figure concludes its process is slow and works on the process, when the cause is a record that was never cleaned.
Time to fill
Time to fill is the period between opening a requisition and the start date of the person hired, and both of its ends can be placed in more than one position. An advertisement that is never filled has no second end at all. Either it stays outside the calculation and never appears, or it is eventually closed as cancelled and everything spent on it disappears with it. In both cases nobody is held to account for it, and that is precisely why it survives.
Selection ratio and cost per hire
A selection ratio quoted without its denominator cannot be compared with anything, and counting people who never entered selection in that denominator shrinks the ratio in a way that says more about the advertising than about the selecting. Ghost jobs are an extreme case of that distortion: applicants who entered the denominator with no numerator ever waiting for them. The ratio falls, and is read as a wider choice and a more selective process, when in fact it describes no selection that took place.
Cost per hire moves the other way. What was spent on the advertisement and the platform fees enters the numerator, and nobody enters the denominator. The cost is spread across the other hires, and every hire looks more expensive than it was. The largest item in cost per hire is usually internal time that appears on no invoice, and here it is even more visible: hours of screening and interviewing spent in a process that had no exit.
How to count them in a given organisation
Because these advertisements are not recorded as such, the question of how many there are cannot be answered by running a report. What answers it is three reconciliations between records that already exist:
- Match advertisements live on job boards against approved requisitions. Every advertisement without an open requisition behind it today falls into this category, whatever the reason. This reconciliation surfaces the whole of the first situation, the one nobody knows about.
- Read the distribution of open requisition ages, not their average. A requisition open many times longer than its counterparts in the same role either has a reason that can be stated or belongs to this category. The average alone hides both possibilities at once, as the calculation above shows.
- Compare the number of applicants with the number of CVs actually read. An advertisement that drew hundreds of applications, only a few of which were opened, was not a working process, whether or not that was intended.
What the last two produce is not a verdict on anyone but a list of cases, each put to a single question: does anyone intend to fill this seat? Most answers correct themselves as soon as the question is asked.
What they cost beyond the numbers
Candidate experience is what an applicant meets along the way. An unanswered application to an advertisement with no seat behind it is its worst form, because it cannot be corrected: there is no decision to communicate, so the application stays suspended indefinitely.
The effect reaches the employer brand as well, and the link is arithmetical rather than a matter of impressions. Each time the organisation advertises without hiring, fewer people apply the next time, the denominator of the selection ratio shrinks, and the choice for the next real vacancy narrows. The damage returns to the same process one cycle later.
What removes the cause
- Tie every advertisement to an approved job requisition, with its approved pay range and authorisation route. What has no requisition is not published, and the third situation falls away at the root.
- Give every advertisement an expiry date at which it closes automatically and reopens only by decision. That alone ends the first situation, which is the most common.
- Separate two requisition states in the system: open, funded and intended to be filled; and open to receive CVs. Only the first feeds the metrics.
- Tell every applicant of some decision, even if the decision is to close the requisition. The cost is close to nothing, and it is the difference between a process that ended and one that did not.
- Give every advertisement a named owner. The first situation survives because withdrawing an advertisement is nobody’s particular job, while every other step in the process has an owner. The cheapest fix for work that does not get done is to assign it to a person.
None of this fixes the fourth situation, publishing an advertisement for a role already settled on its holder. It cannot be fixed by tidying the record because it is not a fault in the record. Two things can be said about it: where the procedure requires publication, the advertisement text can state that an existing candidate is under consideration; and the advertisement should be removed from every metric denominator. Leaving it in makes the organisation measure a selection that never happened and credit its process with a choice that was not exercised.
The figures in the calculations above are assumed to show how the structure moves the metrics. In the sources we reviewed we found no estimate of how widespread ghost jobs are in any particular market, and we reach no conclusion on the legal position of publishing a job advertisement or on what follows from it, which is a matter for the rules that govern it.
Where the line falls
The line used throughout is whether an approved seat existed, with an intention to fill it, at the moment of publishing. Cases that pass that test are not ghost jobs even when they look like them.
A vacancy opened and then cancelled passes it: cancelling after a need changes is a legitimate decision, because the purpose existed when the advertisement went out. What it calls for is telling the applicants, not refusing to cancel. A vacancy with a strong internal candidate passes too, so long as the decision is still open. An internal candidate enters the same process and is measured by the same criteria, and preferring them when candidates are equal is acceptable. The dividing line is that the decision can still be changed by what arrives from outside.
An advertisement for a scarce role that stays open for a long time can pass as well, since its length may describe the scarcity of the profession rather than the absence of a seat. Where a profession is scarce, the selection ratio can reach one, and a long time open there signals a narrow choice, not a missing intention. One general advertisement covering several seats in the same role has simply gathered real vacancies into one text.
A talent pool advertisement fails the test honestly and leaves the category by saying so. Once its text states that it is collecting CVs rather than filling a live vacancy, it is no longer a ghost job and no longer belongs in the denominator of any metric.
A standalone Saudi HR system
One employee file holding the contract, the documents and their expiry dates, the attendance record, leave, salary and end-of-service entitlements. End-of-service, overtime and leave-balance calculations are built into the system.
A standalone system on its own subscription. The connection to Qoyod Accounting is now available.