Work-sample assessment

Assessment by doing a piece of the job: a take-home exercise, a pairing session on real code, a review of a codebase, a paid trial sprint. The third assessment regime this spoke holds, after the trivia-and-recall interview behind interview-preparation and the scored competition behind capture-the-flag — and the first one defined by what it refuses rather than by what it measures.

Entered with hiring-without-whiteboards, which lists 762 companies claiming to hire this way.

The claim

The argument is a resemblance argument: an interview question should look like the work, because a candidate who can do the work should pass and one who cannot should not. Its target is the prepared answer — the CS-trivia question you get right by having studied for it rather than by being able to do the job. hiring-without-whiteboards puts the whole test in two lines: a real-world problem is fine with or without a whiteboard; puzzles, riddles and brainteasers are not, with or without one.

Note what that does to the room. The whiteboard is a prop in the argument, not the subject — which is why the best-known list of companies avoiding whiteboards contains entries that happily use whiteboards.

Why this spoke cares

Its founding thesis is that syllabus follows assessment (interview-preparation). Every curriculum here trains for the recall regime: coding-interview-university excludes frontend and SQL and spends months on algorithms; llm-interview-questions-hoang is 50 questions with model answers and no learning path at all. You cannot cram for a work sample the same way. If work-sample hiring spread widely, the thesis predicts the curricula would have to change shape — toward the build-from-scratch and clone-project end, where the artifact is the point.

The corpus cannot yet say whether that is happening. What it can say is that 762 companies is not enough to move a market of curricula with 350k-star audiences, and the spread is unmeasured anyway (see below).

What it costs, and who pays it

The regime moves work from the interview room onto the candidate's evenings. That has a distribution: someone with caring responsibilities or a current job has less of that time than a new graduate. hiring-without-whiteboards's own RECOMMENDATIONS.md names the mitigations — pay for the time, time-box it under four hours, offer a shorter pairing session as an alternative for candidates who cannot do a take-home — which is a fair account of the cost.

Almost nobody on the list does it. Of the 762 entries, 27 mention paying and 4 mention a time box. So the regime as advocated and the regime as practised differ on exactly the point that decides who can afford to be assessed by it, and the gap is visible in the same file that states the recommendation.

What is not established

Whether work-sample assessment predicts job performance better than the thing it replaces. The claim is intuitive and this corpus has no evidence for it — the standing content of growth edge 4 (synthesis). hiring-without-whiteboards is advocacy with no outcome data; the industrial and organizational psychology literature on selection validity would settle it, and no source here touches that literature.

Stated plainly here because the intuition is strong enough to pass for a finding.

Related

hiring-without-whiteboards · interview-preparation · capture-the-flag · problem-archive · build-from-scratch · clone-project · lauren-tan