A well-designed work sample is the most predictive assessment available to you. It measures what someone can do rather than how they present, and it favours competence over interview polish.
A practical overview of mouse jiggler detection software can help teams connect this process with day-to-day operational data.
A badly designed one loses you exactly the candidates you most want — the ones with other options, who decline rather than spend a weekend on it.
Scope it honestly
Two hours, three at the outside. State the limit and mean it.
The candidates who will absorb an eight-hour task are disproportionately those without competing offers, caring responsibilities, or a demanding current job. That is a selection effect working directly against you, and it is not a neutral one — it correlates with things you should not be selecting on.
Test your own estimate. Have someone on the team do the task and time it. Estimates written by the person who designed it are systematically optimistic.
Position it late
After at least one substantive conversation, when the candidate knows the role is real and you know they are a serious prospect.
A task sent at screening stage to thirty applicants generates goodwill from none of them and gets declined by the strongest.
Make it relevant, and make it hypothetical
Relevant: it should look like the work. A writing sample for a writing role, an analysis for an analytical one, a debugging exercise for an engineering one.
Hypothetical: anonymised or invented data, not a live business problem. Asking candidates to solve something you currently need solved is unpaid work, it is recognised as such, and it damages your reputation among exactly the people you are trying to recruit.
The test: would the output be useful to us tomorrow if we do not hire this person? If yes, either redesign it or pay for it.
Tell them what you are assessing
Publish the criteria with the task. It is not a trick, and telling candidates what matters produces cleaner signal — you learn whether they can do the thing rather than whether they guessed what you wanted.
Also tell them: how long it should take, who will review it, whether there will be a discussion afterwards, and what stage of the process this is.
Consider paying
For anything over about three hours, or for senior roles, payment is increasingly standard and it materially improves acceptance rates.
It also disciplines your own scoping. Employers who pay for work samples design shorter ones.
Score it like an interview
Define what a weak, adequate, and strong submission looks like, in writing, before any arrive.
Score against those criteria, with a short evidence note. Two reviewers where the decision is close. Same process as structured interviews, for the same reasons.
Score what you said you would score. If you published three criteria and then reject someone for a fourth thing you noticed, the exercise was not what you said it was.
Always discuss it afterwards
A twenty-minute conversation about the submission is often more informative than the submission itself: what they assumed, what they deprioritised, what they would do with more time.
It also catches the one thing a take-home cannot: whether they actually did it themselves.
Give feedback
Every candidate who completed the task gets a substantive response, whatever the outcome. Two or three sentences on what was strong and what was not.
They spent hours on it. A one-line rejection after that is the sort of thing people mention publicly, and it is entirely avoidable.
What this costs you if you get it wrong
Drop-off before submission. Strong candidates decline. You never know, because they say they withdrew for other reasons.
Selection bias toward candidates with the most free time.
Reputation. Candidates discuss hiring processes publicly, and unpaid multi-day tasks are a recognisable complaint.
Wasted internal time scoring submissions inconsistently against criteria nobody wrote down.
Alternatives worth considering
A working session. Sixty to ninety minutes, live, on a realistic problem. Costs the candidate less, shows you how they think, and cannot be outsourced.
A portfolio review. For roles where candidates have existing work, a structured discussion of something they already did is often better evidence than something they produced under artificial conditions.
A short structured exercise during an interview, timeboxed to thirty minutes.
For many roles one of these produces the same signal at a fraction of the candidate cost. Use a take-home where the work genuinely requires uninterrupted time.
Before you send one
- [ ] Someone on the team has done it and timed it
- [ ] Under three hours, stated honestly
- [ ] Hypothetical, not a live problem
- [ ] Positioned after a real conversation
- [ ] Criteria published with the task
- [ ] Scoring rubric written before submissions arrive
- [ ] A discussion scheduled afterwards
- [ ] Feedback committed to for everyone who submits
- [ ] Payment considered for anything substantial
PJP designs assessment processes as part of HR consulting engagements. HR consulting → For broader reference, consult U.S. Department of Labor hiring resources.