Work Samples

A work sample asks the candidate to perform a task resembling the role. It predicts better than almost anything else, for the obvious reason.

Once the criteria described in “Work Samples” are fixed, the practical challenge is running them consistently across assessors. A coordinator can use time tracker for projects to plan review time, compare the effort spent at each stage and spot avoidable delays, without confusing administrative efficiency with the quality of the evidence collected from candidates.

For an independent benchmark, compare the process with CIPD selection methods guidance; the useful test is whether the local method remains job-related, proportionate and explainable.

Why it works

It has content validity by construction: you are watching the thing you want to predict.

It requires less inference than any other method. An interview produces claims about capability; a work sample produces capability.

And candidates generally find it fair, which reduces dropout and complaints.

What a good one looks like

Drawn from a real task the role involves, simplified rather than invented.

Short: one to three hours for most roles, and stated in advance.

Scored against criteria written before anybody attempts it.

Done under conditions resembling the job — with reference material available, if the job has reference material available.

And the same for every candidate, which sounds obvious and is frequently not the case.

The common design failures

Too long. A task requiring a weekend selects on available time, not capability.

Unpaid real work. Its own note covers where the line sits and it matters more than most organisations admit.

Artificial constraints: no internet, no documentation, a whiteboard instead of a computer. If the job allows these, the assessment should.

Assessing presentation rather than substance, where the scoring drifts toward polish.

And no rubric, which turns the exercise into an impression with extra steps.

Live versus take-home

Live shows you the process and how somebody handles being stuck, which is frequently the informative part.

Take-home is less stressful and closer to real conditions, and selects on available time.

A short live exercise with discussion afterwards gets most of the benefit of both and is the arrangement worth defaulting to.

Scoring

Three to five criteria, each with a description of what good and inadequate look like.

Scored independently by two people where the decision matters.

And scored on what the job requires rather than on how you would have done it, which is the commonest assessor failure and is addressed by training.

Where it does not fit

Roles where the work cannot be sampled meaningfully in hours: strategic, relational, long-cycle work.

There, a structured interview about specific past situations is the better instrument, and its own note covers how.

Do not invent a fake exercise to preserve the format. A bad work sample is worse than a good interview.

Accessibility

Work samples can disadvantage candidates with disabilities in ways the job would not, because the compressed time and unfamiliar setting amplify barriers.

Offer adjustments, ask in advance, and be flexible about format where the format is not what you are assessing.

What to check

Does your work sample resemble something the role actually does?

How long does it take, and is that stated before candidates commit?

Is there a written rubric, applied by more than one person?

And could a candidate with an access requirement complete it?

The point

A work sample has content validity by construction: you are watching the thing you want to predict, which is why it outperforms methods requiring inference..

Underlying all of this

Almost everything in this collection reduces to one discipline: write down what the job requires, assess that thing directly, record the evidence, and look at your own outcomes afterwards. None of it requires buying anything, and organisations that do those four things consistently outperform ones running longer processes built from instruments chosen before the requirements were known.