Work Sample Tests vs Take-Home Assignments
Take-home assignments are open-ended projects, often requiring 4–20 hours, and they select for candidates with free time as much as for ability. Work sample tests are time-capped, rubric-scored exercises of 60–90 minutes that measure decision density instead of output volume, with far lower drop-off.
Updated 30 July 2026
Why do take-home assignments have such high drop-off?
Because the cost falls entirely on the candidate and scales with how many jobs they are applying for. Ten applications at eight hours each is a part-time job performed for free with no guarantee of a reply.
The bias this creates is specific and severe: it selects for candidates with childcare, without a second job, and without a demanding current role. Those are not correlates of ability. Strong senior candidates decline take-homes at the highest rate of any group.
What makes a good work sample test?
Time-capped and honest about it. 60–90 minutes, stated up front, enforced by design rather than by trust.
High decision density. The brief should contain a genuine trade-off in every part. What separates candidates is which trade-offs they notice, not how much they can produce.
Published rubric. Candidates see the scoring criteria before they start. Hidden criteria produce anxiety, not signal.
Realistic mess. Contradictory stakeholders, incomplete data, a tempting wrong answer. Clean puzzle problems measure puzzle-solving.
Portable result. The candidate keeps the proof and reuses it. This is the single change that most improves acceptance rates, because it converts free work into an owned asset.
Doesn't a shorter test give less signal?
Less output, not less signal — and they are not the same thing. A four-hour build tells you the candidate can execute a known task for four hours. A 90-minute design problem with three real trade-offs tells you what they would sacrifice, what they refuse to automate, and how they would verify they were right.
Seniority in particular is measured by judgment, not throughput, and judgment shows up in the first hour or not at all.
Should candidates be paid for assessments?
For long take-homes, yes — if you are asking for a working day of output, that is work. For a 90-minute time-capped sample the more useful compensation is portability: the candidate does it once and the proof counts with every other employer.
That reframing removes the extractive dynamic without introducing per-candidate payments that most teams will not administer consistently.
Try it on one role you’ve struggled to fill.
Define it as four to six skills with required levels, send one challenge link, and count how many of your top ten the old screen would have rejected.
Build a skill specQuestions people ask
- How long should a work sample be?
- 60–90 minutes for most roles. Beyond two hours, drop-off rises sharply and the additional signal is mostly stamina and availability.
- Should the test relate to our actual product?
- It should resemble your problem shape without requiring your domain knowledge, which would advantage industry insiders for reasons unrelated to ability. Supply whatever context is needed inside the brief.
- How do we score consistently across reviewers?
- A weighted rubric with per-criterion descriptions, applied blind, with periodic calibration where two reviewers score the same submission and reconcile differences.