Pipeline
Naming the Assessment Stage in the Post Moves Drop-Off Earlier
Name the assessment stage in the job post, in three facts: how many stages there are, that one of them is an assignment, and roughly how long that assignment takes. Two lines near the end of the post. Expect applications to fall and completion to rise, which is the trade you want when the binding constraint is reviewer hours rather than applicant count. Most people who drop out at the sight of a three-hour assignment were going to drop out later, after two rounds of interviewer time.
The takeIf naming the assignment collapses the pipeline, the assignment is the problem, and the posting just told you so a quarter earlier than the withdrawal emails would have. That is the useful reading of a fall in applications: a three-hour unpaid exercise at a screening stage was always going to lose people, and hiding it until after the second interview only changes when you pay for it. Publish the number, then argue about the number.
Where Olive fits
Open a role and see what the work shows
Olive is priced per attempt rather than per seat, and one attempt returns six evidenced findings on a single candidate for the person making the call to weigh. Ten attempts a month are free, so it can run beside the assignment the post already names.
Rank your shortlistWhich three facts belong in the post?
The count, the assignment, and the clock. How many stages the loop has, that one of them is a take-home or an exercise, and its time budget in hours. Three facts, two lines, near the end of the post beside the rest of the process detail. Anything longer reads as an FAQ nobody asked for and gets skipped with the boilerplate.
A version that fits on two lines: four stages, including a 90-minute assignment done on your own clock between the screen and the panel. Total elapsed time about three weeks.
Four things stay out of the post: the brief itself, the rubric, the tools list, and any paragraph justifying why the exercise exists. Those belong in the assignment email, where the candidate has already decided to spend the time and will actually read them. The posting line exists to let somebody opt out, not to prepare them.
Naming it is rarer than it sounds, which is most of why it works. In SHRM's benchmarking survey of member organisations, 12% used work sample interviews for executive candidates, 11% for middle management and 9% for individual contributors, against 34% to 37% for structured interviews 1. That is self-reported by HR respondents about 2021 practice, before generative AI reached hiring, and organisations routinely claim more structure than they run. It still puts a described assignment in the minority of postings, which is what makes two honest lines a differentiator rather than a formality. The rest of the block belongs with the published evaluation process.
Where does drop-off actually cost you?
At the stage where somebody's calendar is involved. Application volume is nearly free to carry; a screen, a panel and a debrief are not, and a candidate who withdraws at the assignment has already spent those hours for you. SHRM's benchmarking data puts median non-executive time-to-fill at 44 days, of which the median organisation spends 5 days screening applicants and 7 days conducting interviews 1.
The cost figure in the same report makes the point sharper by what it leaves out. Median cost-per-hire for non-executive roles is $1,244, against a mean of $4,683, and SHRM's definition sums agency fees, advertising, job boards, referrals, travel, relocation, recruiter pay and system costs 1. It does not include the hiring manager's or the interviewers' time, which is usually the largest real cost of a loop. So the expense a late withdrawal creates is the one your reporting cannot see, which is why late drop-off keeps getting treated as cheap.
Top-of-funnel volume is the thing least worth protecting right now, for reasons that have nothing to do with your posting. When applications per opening have tripled, an application costs almost nothing to receive and almost nothing to lose, while a reviewer hour costs exactly what it always did.
The sequence matters more than the totals. Every candidate who withdraws before the screen is a full loop you did not run, and the posting is the only place in the process where a withdrawal is free to both sides.
How long can the assignment be before naming it backfires?
Somewhere under three hours for most roles, and the honest test is whether you would do it. A 90-minute exercise, named, reads as respect for the candidate's time. An eight-hour project, named, reads as a warning, and it is one. Disclosure does not make a long assignment acceptable; it makes the length visible earlier, which is the whole mechanism and also its limit.
Length is not buying accuracy either. The 2022 re-analysis of the selection literature puts work sample validity at .33, well below the .54 that gets quoted from the older tables, with structured interviews top ranked at .42 2. Those are corrected correlations with supervisor performance ratings rather than accuracy rates, and the credibility intervals overlap heavily, so this is not work samples losing to interviews. The relevant part for a posting is narrower: nothing in that literature is about unpaid multi-hour take-homes, and no evidence says a six-hour exercise predicts better than a ninety-minute one.
Two adjustments make a named assignment easier to publish. Cut it to the smallest task that still produces something worth discussing, then say in the post that the discussion is where the decision gets made. And consider paying for the take-home, which changes the sentence in the posting from a demand into an offer and is the only lever that reliably improves completion without shortening the work.
What to do when applications fall too far
Check completion before touching the posting. A fall in applications with completion holding or rising is the intended result rather than a failure. A fall in both means the post is deterring people you wanted as well, and that is a different repair. Read the two numbers together across a full requisition, because one week of application counts tells you nothing at all.
If both fell, work through three levers in order: shorten the assignment, move it one stage later so a person has spoken to the candidate first, or pay for it. Rewriting the posting line is the last resort, because the line is not what people are reacting to. The exercise is.
Watch the shape of the pool as well as its size. A requisition that fills its cap in two days is a different problem from one that draws nobody, and a post that closed in 48 hours on volume selected for speed of application rather than for anything you meant to measure. The assessment line pushes in the opposite direction, which is usually welcome and occasionally too strong.
One caveat worth stating plainly: no study measures what naming an assignment does to a specific applicant pool, so treat every number here as a reason to instrument your own. Run the two lines on half your open requisitions for a quarter and compare completion rates. It is one line of template text, which makes it the cheapest experiment available in a job post.
Common questions
Where in the post should the assessment line go?
Near the end, with the rest of the process detail, after the role and the requirements. A candidate reads down until something disqualifies the job for them, and process detail placed high in the post interrupts the part that is supposed to sell the role. The exception is a genuinely unusual assignment, such as one requiring specific software or an on-site day, which belongs high enough that nobody reaches the application form without seeing it.
Does naming the assignment hurt more at senior levels?
Yes, and the fix is the assignment rather than the wording. Senior candidates are the most likely to be employed, the least likely to have three spare hours, and the most likely to read an unpaid exercise as a signal about the company. Many teams replace the take-home with a paid working session or a structured conversation about work already shipped at that level, then name that instead.
Should the post say the assignment is paid?
If it is paid, say so and say the amount. It is the strongest sentence available in that part of a job post, it converts the line from a cost into an offer, and it is trivially verifiable. If it is not paid, do not imply otherwise: a candidate who discovers the exercise is unpaid after reading a vague line about compensation for your time is worse off than one who was told nothing.
What if only some candidates get the assignment?
Say that, in the same sentence. Shortlisted candidates complete a 90-minute exercise is accurate and still lets somebody decide whether to apply. The version to avoid is silence followed by a surprise, and the version to avoid even more is describing an assignment stage you skip whenever a candidate seems strong, because the two candidates who compare notes will find it.
Does a smaller applicant pool narrow who ends up in the pipeline?
It can, and nobody should assume it does not. Self-selection is not neutral: people with less time, less confidence, or less familiarity with a company's norms drop out of visible-cost stages at different rates. Measure the composition of your own pool before and after the change rather than reasoning about it, and if the pool narrows in a way you did not intend, the assignment is again the thing to adjust.
References
- 1. SHRM Benchmarking: Talent Access (Selection Criteria, Overall) shrm.org Supports how rare a work-sample stage is, the median time-to-fill and where its days go, and the cost-per-hire definition that omits interviewer time.
- 2. Revisiting Meta-Analytic Estimates of Validity in Personnel Selection: Addressing Systematic Overcorrection for Restriction of Range gwern.net Supports the corrected work sample and structured interview validity estimates behind the claim that a longer assignment is not buying accuracy.
2 sources, numbered by first appearance. How Olive sources claims
General guidance for hiring teams. What works at one company and one volume may not transfer to yours.
Olive assesses how a person works with AI. It does not detect AI-written documents, and it never produces a score, a ranking, or a match percentage for a person. Candidates read the same report the employer reads.