Who this is for. In-house talent acquisition, agency recruiters, and hiring managers who run two or three processes a year and have never seen what they look like from the other end. Nothing here requires budget. Most of it takes minutes. Every number below carries its source and its limitations — see how we handle evidence.
What a long search looks like from the other side
Research says
Silence has a cost, and it is larger than it looks
In an Indeed survey of UK workers, 55% of job applications went unanswered, and 86% said they would rather have a clear rejection than no reply at all. A rejection can be filed and moved past. Silence has to be interpreted, and people interpret it by re-reading their own application looking for a flaw — which is slower, more painful and less useful. A two-line automated rejection at the point of decision costs almost nothing and does something that a well-meant intention to “get back to everyone eventually” does not.
Source: Indeed, UK survey of 1,000 workers and 1,000 hiring professionals, May 2024.
Research says
Ghost jobs are not a victimless piece of admin
Greenhouse, which runs an applicant tracking system used by thousands of employers, found that in any given quarter 18–22% of jobs posted on its platform are classified as ghost jobs, with no active hiring behind them. Separately, in a US survey of hiring managers, 40% said their company had posted a fake listing in the past year and around three in ten said one was live at the time of asking. Pipeline-building, roles left up after they are filled, and listings an ATS re-posts automatically all look identical from the outside to a real opportunity. Every one of them absorbs hours of unpaid work from someone who does not have hours to spare.
Source: Greenhouse, State of Job Hunting, December 2024 — the company's own analysis of postings on its platform; the classification method is not published, so this is vendor data rather than independent research. ResumeBuilder, May 2024 — a self-selected online panel of 649 US hiring managers; the “three in ten” is derived from 40% having posted a fake listing, of whom 79% said one was still live. Neither is UK data.
In practice
Long processes are not a neutral filter
A five-stage process running over ten weeks, with a presentation and a task, asks for something beyond the ability to do the job: repeated days off, unpaid preparation time, and the financial room to wait ten weeks for an answer. Those are not evenly distributed. Around three of the five million unpaid carers in England and Wales are women. Caring responsibilities, disability and low household savings all track characteristics you are not permitted to select on, and none of them appear anywhere in your process as a criterion. Nobody has measured who drops out of long processes and why — but every additional stage is a decision about who can still be in the room, and it is worth knowing you are making it.
Source: ONS, Census 2021, England and Wales: approximately 3.0 million of 5.0 million unpaid carers are women. The differential-attrition claim itself is an inference from those distributions, not a measured finding.
In practice
Unpaid tasks are work, and their scale has drifted
There is a real difference between a two-hour exercise designed to see how someone thinks and a multi-day project that produces something usable. Candidates can tell which is which, they discuss it publicly, and the second kind damages an employer brand for years. If a task would take a current employee more than about two hours, it is worth either paying for it or cutting it down.
Research says
Age bias mostly operates before anyone has made a decision
In research for the Centre for Ageing Better, 36% of people aged 50–69 said their age would put them at a disadvantage when applying for jobs, and 17% said they had experienced age discrimination during recruitment. Among those who reported experiencing it, 68% said it had undermined their confidence and 76% said it had put them off applying for further jobs — which is the part that does not show up in your data, because it consists of applications you never received. Much of this may not involve anyone consciously deciding to exclude older candidates. Bias can also enter through screening criteria, the phrase “digital native”, an implicit picture of what the team looks like, and risk-avoidance under time pressure. Those are all fixable at the process level.
Source: Centre for Ageing Better, with Demos and NIESR — Too much experience: older workers' perceptions of ageism in the recruitment process, February 2021. Nationally representative survey of 1,539 UK adults, plus 55 in-depth interviews in England with people aged 50–69. The 68% and 76% figures rest only on the sub-sample who reported experiencing discrimination.
Evidence suggests
The person in the interview may be several months into this
Sustained stress and poor sleep tend to blunt working memory and mental flexibility — small average effects in the research, but they land on precisely the moment an interview is testing. Someone six months into a search may present as less capable than they are. Sending questions or topics in advance, saying clearly what the format will be, and starting with something easy costs you nothing and gives you a better read on capability rather than composure.
Source: The controlled work here is on acute stress in laboratory conditions — Shields, Sazma & Yonelinas, Neuroscience and Biobehavioral Reviews, 2016 — where effects on working memory and cognitive flexibility are consistent in direction but modest in size, and effects on inhibition are mixed. Chronic stress cannot be manipulated experimentally, so the extension to a long job search is inference, not measurement.
In practice
Delay is expensive on your side too
Delay costs you candidates, and disproportionately the ones already in work or holding another conversation. Long processes increase drop-out, force re-runs, and leave roles unfilled while the work piles up on the team you already have. Most of what is good for candidates here is also simply faster hiring.
Source: Widely reported by recruiters; not measured in controlled research.
Eight things that cost almost nothing
In rough order of how much difference they make relative to the effort involved.
1. Reject at the point of decision, automatically, with a real send date. Two lines is enough.
2. Take down listings when they are filled, and stop posting roles you are not currently hiring for.
3. Publish the number of stages and the expected timescale in the advert, and hold to it.
4. Cap unpaid tasks at around two hours, or pay for them.
5. Send interview topics or questions in advance — you will assess capability rather than composure.
6. Give a specific reason to anyone who reached a final stage. It takes three minutes and they will remember it.
7. Review screening criteria for proxies for age: graduation years, “digital native”, “recent graduate preferred”, arbitrary years-of-experience ceilings.
8. Tell people if a process is paused or the role is pulled. Being told is vastly better than being forgotten.
The pages most worth reading from a hiring perspective
This page is not an argument that candidates should be handled gently because they are fragile. It is an argument that a process which leaves people uninformed, unpaid and unanswered produces worse hiring decisions as well as worse experiences — and that the fixes are mostly administrative.