The work sample: how long, what to ask for and how to score it

Echipa HR 365 · reviewed 2026-09-07 · 5 min read

A good work sample takes under two hours, resembles real work, and is scored against a rubric written before it is sent out. The three conditions look obvious and are usually broken all three at once.

Length: the rule and why it matters so much

Past two hours, the task selects for availability rather than competence. Someone with a job, a family and two other processes running does not do it — or does it in a rush, badly. Whoever does it well is often whoever has the most free time, which is not the criterion you were after.

The practical rule: the length of the task is proportional to how far the candidate is through the process. A task before the first interview does not exceed thirty minutes; one at the final stage can reach two hours, because the investment is mutual by then.

What you ask for, to learn something useful

The best task is a reduced version of a real piece of work, with the data changed. The worst is an invented exercise testing an ability the role does not require — usually speed or memory.

1. Take a piece of work somebody on the team did last month.
2. Cut it down until it fits the allotted time, keeping the part where a decision is made.
3. Change the data and remove anything confidential.
4. Add a deliberate ambiguity — something that should prompt a question.
5. Write the scoring rubric before you send the task.

The fourth step yields the most information. A candidate who asks before assuming tells you more about how they will work than the final result does. And if nobody asks, the ambiguity was not visible — which is information about the task, not about the candidates.

The rubric, written beforehand

A task scored “on impression” is no better than an unstructured interview and is far more expensive for the candidate. The rubric has three or four criteria, each with what counts as acceptable and what counts as above expectations, and it is written before anybody sees a single submission.

Feedback is mandatory

Someone who invested two hours deserves to know what you saw in their work — whatever the outcome. It is not politeness: it is the only thing that stops the task being unpaid work. Two good things, named precisely, and one that was missing, framed against the brief.

The rule worth keeping: if you are not willing to give feedback, do not set a task. A process that consumes candidates’ time and returns nothing becomes known, and in a local market it becomes known quickly.

Do I give the same task to every candidate?

Yes, otherwise the results cannot be compared. If you want to stop it circulating, keep two or three equivalent versions, scored against the same rubric.

What if I suspect they did not work alone?

Discuss the work with them, for fifteen minutes. Whoever did it can explain why they chose each thing; whoever did not is immediately obvious — and it is a fair conversation, not an accusation.

Where to start

Time your own task: ask somebody on the team to do it and measure. It usually takes twice as long as you estimated.

Then write the rubric, even if the task has existed for a year. It is the only way two different assessors reach the same conclusion about the same submission.

When you set no task at all

A work sample is not obligatory and, for some roles, produces less information than a well-structured conversation. Three situations where dropping it is right:

• When the real work cannot be reduced to an isolated task — coordination roles, where the value lies in relationships and context.
• When the candidate has public work or a verifiable portfolio. Ask them to discuss one piece of it.
• When the role is a volume hire and process speed matters more than selection precision.

The second situation is the most often missed. A thirty-minute conversation about a piece of work the candidate has already done usually produces more information than a new task — because it can be probed indefinitely and cannot be improvised.

What you tell candidates about the task

Four things, in the invitation: how long it realistically takes, what is being assessed, by when they should send it, and when they will hear back. Missing the last point causes the most drop-outs — a candidate who worked two hours and hears nothing for three weeks assumes they were rejected and moves on.