Interviews · 5 min read
How to prepare for a skills test or work sample
A skills test or work sample asks you to do a small piece of the actual work: respond to a difficult customer email, analyse a dataset, write a brief, make a change to some code, build a short plan.
It is the assessment most worth taking seriously, because it is the one where you can show what you can do directly rather than describing it. It is also the one where candidates most commonly make an avoidable mistake, which is misjudging what is being assessed.
What they are actually assessing
Rarely a perfect answer. Usually four things.
Whether you can do the core task at a reasonable standard.
How you approach it. What you did first, what you prioritised, what you decided to leave out. For many roles this matters more than the output.
Judgement under constraint. They gave you ninety minutes for something that could absorb a week. What you chose to do with the time is the test.
Whether you followed the brief. The most common failure, and entirely avoidable. Answering the question you wish they had asked loses more candidates than any technical weakness.
Read the brief twice, and answer that one
Before you start anything, write down what they actually asked for, what the deliverable is, what the constraints are, and what the stated time is.
Then check your work against that list at the end.
Candidates lose this stage by producing something impressive that does not address the brief. If the brief asks for a one-page recommendation and you produce a fifteen-slide deck, you have demonstrated that you do not follow instructions, whatever the quality of the deck.
If something in the brief is genuinely ambiguous, ask. It is not a weakness, and for most roles asking a good clarifying question is itself positive evidence. If you cannot ask, state your assumption in your answer: "I have assumed the audience is the finance team rather than the board. If that is wrong, the emphasis would change."
Stick to the stated time
If they say ninety minutes, spend ninety minutes.
Spending six hours on a ninety-minute task is a bad idea for three reasons. It is unfair to candidates who took the instruction literally, which most employers care about. It produces work that is not representative of what you would deliver under normal conditions, which is what they are trying to see. And it tells them nothing about how you prioritise, which is frequently the main thing being assessed.
If you genuinely could not complete it in the time, say so and say what you would have done next. "I ran out of time before the sensitivity analysis. My approach would have been X." That is a good answer and it is honest. Employers see straight through work that clearly took four times the stated time.
On using AI
Assume some candidates will, and increasingly employers assume it too.
The practical position: if the employer has not prohibited it, using AI as you would in the job is defensible, because it is how the work is now done in many roles. If they have prohibited it, do not.
What matters either way is that you can explain and defend every part of your submission. Most employers now follow a take-home task with a conversation about it, precisely for this reason. Someone who cannot explain why they approached something a particular way, or who cannot answer what they would do differently, has told the interviewer everything they needed to know.
So if you use a model, understand what it produced, check it, and be ready to justify it. Submitting something you cannot discuss is the one genuinely fatal approach.
Present it well, briefly
Say what you did and why, in a few lines at the top. This is where you demonstrate the thinking that the output alone does not show.
State your assumptions. Real work involves incomplete information and naming what you assumed is a professional habit.
Say what you would do with more time. Almost always asked, and answering it pre-emptively shows you know the limits of your own work.
Check it before you send. Errors in a work sample are weighted heavily because this is you at your most careful.
When a task is unreasonable
Some are. It is worth knowing where the line is.
A task taking many hours, unpaid. Anything beyond about ninety minutes should be paid, and this is increasingly the expectation. It is reasonable to ask.
A task that solves a live problem. If the brief is suspiciously specific to a real current problem and they would clearly use the output, that is free work. Asking what happens to the output is a fair question.
A task before any human contact. Being asked for two hours of work before anyone has spoken to you is a poor signal about how the employer sees candidates.
You are allowed to ask about any of these, and how the employer responds is itself useful information. A reasonable employer explains, offers payment or adjusts the scope. One that reacts badly to a polite question has told you something worth knowing.
Common questions
How long should I spend on a take-home task?
The stated time. Spending several times longer is unfair to other candidates, unrepresentative of your normal work, and obvious to the person reviewing it.
Can I use AI to complete a work sample?
Unless prohibited, using it as you would in the job is generally defensible. What matters is that you can explain and defend every part of what you submit, because most employers now discuss the task afterwards.
Should I be paid for a work sample?
Anything substantial, certainly beyond about ninety minutes, should be paid, and it is reasonable to ask.
What if I cannot finish in the time?
Submit what you have and say clearly what you would have done next and how. That is a stronger answer than something that obviously took four times as long.
The short version
Answer the brief they actually set, stick to the stated time and say what you would do next if you ran out, state your assumptions, be ready to explain every part of what you submitted, and ask politely about anything unreasonable.