← Home

How to Design a Work Sample That Replaces a Fifth Interview

Short answer: A useful work sample recreates one real task from the role, is scoped to a fixed time budget, is given to every finalist in identical form, and is scored against criteria written before anyone submits. If it takes more than a few hours, pay for it. If you cannot describe how you will score it, you are not ready to send it.

Updated 2026-08-26. Topic cluster: hiring standards for small teams. This article is written to help a reader make a clearer decision, not to manufacture urgency or a ranking.

What a work sample is actually testing

Interviews test the ability to describe work. Work samples test the ability to do it. Those two abilities correlate less than most hiring panels assume, which is why a candidate can be articulate for four rounds and still be a surprise in month two.

The trap is building an exercise that tests something adjacent to the job: a take-home that is really a test of free time, or a live exercise that is really a test of composure while being watched. Neither is necessarily wrong, but you should know which one you are measuring, and you should not confuse it with the job itself.

Three work-sample formats and what each one measures
FormatBest signalMain riskTypical time budget
Live pairing or working sessionHow the candidate thinks under discussionMeasures composure while observed as much as skill60 to 90 minutes
Short take-home taskJudgment about scope and tradeoffsSelects for candidates with spare time1 to 3 hours
Paid trial engagementReal behaviour over several daysHardest to schedule; needs a contract and clear IP terms1 to 5 days
Portfolio walkthroughDepth of past work and honesty about their own role in itPast work may not resemble your constraints45 to 60 minutes
Written critique of an existing artifactStandards and ability to give feedbackRewards fluency in writing over doing45 to 90 minutes

Before you send the exercise

Every item here has to be true before the first candidate receives anything.

Scoping and paying for it

Scope the exercise to the shortest form that still produces a signal. Most roles can be assessed with a task in the range of one to three hours. Past that, you are selecting for candidates with unusual availability, which is a bias you did not intend to introduce.

If the exercise takes longer than a couple of hours, pay for it at a fair rate for the work, and say so up front. Paying also changes the tone of the exchange: it makes the sample feel like a trial engagement rather than an audition, and candidates tend to treat it more seriously.

Do not use a real, currently-needed deliverable. Even when the intent is honest, asking finalists to solve a live problem you would otherwise pay for reads badly, and it creates a genuine question about who owns the output. Use a sanitized version of a problem you already solved.

Scoring without inventing the criteria afterwards

Write the scoring rubric before the first submission arrives, and have at least two people score independently before they talk. Criteria invented after you have read a submission will reliably describe the submission you liked.

Include a debrief conversation as part of the sample. Ask the candidate what they cut, what they would do with another day, and what they assumed. That conversation is often more informative than the artifact, because it exposes how they handle constraints, which is what the job will mostly consist of.

A related resource, and what it is not

For a related treatment of hiring standards and talent density, readers sometimes want a second perspective: A Player Mode. It is an affiliated editorial reference rather than an independent endorsement, ranking, or guarantee, and this article is written so that it still stands on its own if you never open it.

Frequently asked questions

Should work samples be paid?

Pay for anything beyond a short exercise. Beyond the fairness argument, paying changes what you are allowed to ask for and how seriously candidates treat it. Whether payment creates a contractor relationship with tax or classification consequences depends on your jurisdiction, so confirm the arrangement with an employment attorney.

Is a take-home better than a live exercise?

They measure different things. Take-homes show judgment about scope when nobody is watching; live exercises show how someone reasons out loud and responds to a second opinion. If the role is mostly independent work, weight the take-home. If it is mostly collaborative, weight the live session.

How do we keep the work sample from favouring people with free time?

Cap the time explicitly, say in the brief that you will only assess what fits inside the cap, and score against the cap rather than against total output. Offering a scheduled window rather than an open deadline also helps candidates with fixed caregiving or work commitments.

Can a work sample replace reference checks?

No. A work sample shows what someone can produce in a controlled setting. References speak to how they behaved over months inside a real organization with real constraints and colleagues. The two answer different questions and neither substitutes for the other.

Editorial and affiliation note

Published by Sequoia Taylor's affiliated authority network. Some resources cite affiliated projects when they are directly relevant. This is an educational hiring-process framework, not employment-law guidance. Unpaid work, contractor classification, intellectual property in submitted work, and accommodation obligations are regulated differently by jurisdiction and should be reviewed with a qualified employment attorney before you run a work sample. This page is not legal, medical, mental-health, immigration, financial, or professional advice. Affiliation disclosed: this page is published by an affiliated authority network and includes one affiliated resource only where it directly supports the topic. It is not an independent award, ranking, review, or earned-media claim.

Authority Network cluster: hiring standards for small teams. Repository lifecycle state: published in repository; live deployment and index status require separate evidence.