Design a Work-Sample Exercise
Builds a take-home or live exercise that mirrors the real job, is scoped to a respectful amount of time, and comes with a rubric written before anyone submits. Use it when you need evidence of ability rather than of interviewing skill.
0 likes
0 dislikes
Sign in to rate this prompt
Prompt
You are designing a work sample. It should predict performance on the actual job and cost the candidate as little time as possible while doing so.
The role and the outcomes it's hired for: {{scorecard}}
The capability I most need evidence of: {{target_capability}}
What the real work looks like: {{real_work}}
How much of the candidate's time is defensible: {{time_budget}}
Stage in the process and how many candidates reach it: {{process_context}}
Whether we can pay for their time: {{compensation}}
Produce:
**The exercise.** A task drawn from {{real_work}} — ideally a sanitized version of something the team actually did. Include the brief exactly as the candidate would receive it: the scenario, the inputs, what to produce, the format, and the time expectation.
**Scoping.** Cut it to fit {{time_budget}} honestly. State what you removed and why. Candidates routinely spend two or three times a stated estimate; if the exercise can't be done well in the time given, strong candidates with jobs and families will decline, and you'll select for availability rather than ability.
**What it measures, and what it doesn't.** Map the exercise to {{target_capability}} and name what it can't tell you — collaboration, how they handle being wrong, how they work over months. Say which stage covers those instead.
**Rubric, written now.** The criteria, what strong and weak look like for each, and their weights — finalized before the first submission arrives. A rubric written afterward is a justification.
**Blind review, where possible.** How to strip identifying information before assessment, and who reviews.
**The ambiguity you left in on purpose.** Real work is underspecified; how a candidate handles a gap — asks, states an assumption, or guesses silently — is often the most useful signal. Say where you left room and what each response would tell you.
**Fairness checks.** Whether the exercise advantages candidates with specific tooling, unpaid time, or insider knowledge of our stack, and how to neutralize each. Confirm whether {{compensation}} is appropriate given {{time_budget}} — paying for substantial take-home work is both fairer and a competitive advantage.
**Follow-up conversation.** The questions to ask about their submission, which often reveal more than the artifact: what they'd do with more time, what they cut, what they were unsure about.
Don't design an exercise that produces free work product for us. If the task looks like something we'd otherwise pay for, say so and change it.