A work sample test asks a candidate to do a short piece of real job work, then scores the result against a written guide. To use one well, test only the skills the job needs, keep the task short, give clear instructions, use a made-up case, pay for long tasks, score blind to name where possible, and give every candidate the same task, time and support.
Key takeaways
- Choose a task that mirrors a real part of the job and tests a must-have you cannot judge from the CV.
- Keep it short: 30 to 90 minutes in most cases. Pay if it takes longer or looks like real work for your business.
- Write the brief, the time limit and the scoring guide before you send it to anyone.
- Use a made-up case. Never use a task to get free work.
- Offer adjustments and an alternative format. Score every candidate on the same guide.
What a work sample test is, and when to use it
A work sample puts the candidate in a small version of the job. A bookkeeper reconciles a short set of fictional accounts. A support agent answers three sample customer messages. A designer improves a layout brief. A manager plans a short schedule with conflicts. Because it shows what the person can do, it can give stronger evidence than a conversation about what they have done.
Use a work sample when: the must-have is a skill that can be shown in an hour or less; the CV and interview cannot confirm it; and you are down to a few finalists. Do not use one as a first filter for all applicants. That asks too much time from too many people. See our interview plan guide for where it fits in the stages.
| Role | Possible task | Skill tested | Suggested time |
|---|---|---|---|
| Customer support | Reply to three fictional customer messages with different problems | Written clarity, empathy, policy use | 30 to 45 minutes |
| Operations coordinator | Plan a week of deliveries from a short list with conflicts | Planning, priorities | 45 minutes |
| Marketing coordinator | Write a short email and a subject line for a given audience and goal | Writing, audience focus | 45 minutes |
| Finance assistant | Find errors in a small sample sheet | Accuracy, method | 45 to 60 minutes |
| Developer | Fix a small bug in a given code sample, or review a short pull request | Reading code, reasoning | 60 to 90 minutes |
| Manager | Prepare a one-page plan to handle a fictional staffing gap | Judgement, planning | 60 minutes |
Design the task in seven steps
Step 1: Choose the must-have
Pick one to three must-haves from your role brief. The task must tell you something new about each. If the task tests many things at once, it is hard to score fairly.
Step 2: Write a made-up case
Create a short scenario with fictional names and data. Make it realistic but not secret. Do not use real customer data. Do not ask for work that you will use in your business.
Step 3: Write the brief
State the goal, the materials provided, the time limit, the format of the answer, the tools allowed and how you will score. Use short sentences. State that the task is not a promise of a job.
Step 4: Write the scoring guide
| Criterion | 1: weak | 2: partial | 3: good | 4: strong |
|---|---|---|---|---|
| Accuracy | Major errors | Several errors | Minor errors | No errors; checks shown |
| Clarity | Hard to follow | Partly clear | Clear | Clear and well structured for the reader |
| Method | No method visible | Some steps visible | Sound method | Sound method and trade-offs explained |
| Use of the brief | Ignores the brief | Partly follows it | Follows it | Follows it and notes assumptions |
Step 5: Test the task
Ask a current employee to do the task. Time them. Check whether the instructions are clear, and whether a good answer exists within the time. Adjust the scope or the time limit.
Step 6: Plan fairness
Give every finalist the same brief, time and support. Offer extra time or a different format when a person needs an adjustment, such as a screen reader, extra time or a spoken alternative. Do not make the task depend on tools or access that only some candidates have.
Step 7: Score and give feedback
Score each answer against the guide, ideally with names hidden. Two scorers on a sample reduce errors. Record the scores in the scorecard. Send feedback to all finalists. See how to give fast interview feedback.
Pay, time and respect for candidates
| Task length | Suggested approach | Reason |
|---|---|---|
| Up to about 30 minutes | Do it in the interview, supervised or as a live exercise | Low burden; no payment needed |
| 30 to 90 minutes | Offer a time slot, or a take-home with a clear time limit; do not expect more time to be spent than stated | Moderate burden; many candidates work full time |
| More than 90 minutes | Pay a fair fee, or reduce the task | Long tasks can be unpaid work, and exclude people with limited free time |
| Anything that creates value for your business | Pay, or do not ask | Unpaid work may be unfair and, in some places, unlawful |
The rules on paid and unpaid trial work differ by country. In the US, the Department of Labor explains when unpaid work may break wage laws in its FLSA pages. In the UK, check the minimum wage rules on GOV.UK. In the EU, each country has its own rules. When in doubt, pay for the time. This is general information, not legal advice.
Templates
Brief to the candidate
Hello [Name],
As the next step, we would like you to do a short exercise. It shows us how you work on a task similar to the job. It is a made-up case with fictional data.
Time: about [60] minutes. Please do not spend more than [75] minutes.
What to do: [two or three sentences].
What we give you: [file or text].
What to send back: [format] by [date and time].
How we score: [clarity, accuracy, method and use of the brief]. Every candidate gets the same task and the same guide.
Adjustments: if you need extra time, a different format or any other support, please reply and we will arrange it.
Pay: [fee, or “this exercise is unpaid because it takes under 60 minutes”].
Thank you,
[Name]
Feedback after the task
Hello [Name],
Thank you for completing the exercise. We scored it against the guide we sent. Your strongest point was [specific point]. The area where we saw less evidence was [specific point]. We will contact you by [date] with a decision on the next step.
Best wishes,
[Name]
A worked example (fictional)
This example is invented. A team hires a customer support lead. The must-have is “handles a hard customer message with clarity and care”. The brief gives three made-up messages: a late parcel, a wrong item and an angry complaint. The candidate has 45 minutes. Two scorers use the guide above. Candidate A writes three polite but generic replies, and scores 2 on method. Candidate B writes clear replies, explains the steps she would take internally and notes assumptions, and scores 4 on method. The scores go into the scorecard next to the interview scores. In the debrief, the group sees that the interview scores were similar, and the task separates the two candidates on a must-have that matters. The evidence is clear, and the decision is easy to explain.
Risks and how to reduce them
| Risk | Why it matters | Control |
|---|---|---|
| Bias | Scorers react to style, names or handwriting | Blind scoring, a written guide, two scorers on a sample |
| Unfair access | Some people lack time, tools or a quiet space | Offer time options, tools and adjustments |
| Free work | The task looks like real work | Made-up case; pay for long tasks |
| Over-testing | Too many tasks drive away good people | One task, only for finalists |
| Cheating or outside help | Take-home tasks may be done with help | Ask follow-up questions about the answer in the interview |
| Weak link to the job | The task does not match real work | Ask current employees to check it |
Use the task as one piece of evidence, together with the interview and references. Do not let it decide alone. Review the task after each search: did the scores relate to later performance? If not, change the task.
Example tasks with scoring notes
| Task (fictional) | What a strong answer shows | Common weak answer |
|---|---|---|
| Plan deliveries for five days with three vehicles and two driver absences | A clear schedule, explicit assumptions, a back-up plan and a note on which orders may slip | A schedule that ignores the absences or has no assumptions |
| Write a reply to a customer who received a damaged item | Acknowledges the problem, states what will happen and when, is short and polite | Long apology with no clear action |
| Review a short budget sheet with three deliberate errors | Finds the errors, explains each, and suggests how to prevent them | Finds one error and rewrites the sheet without explanation |
| Plan a 30-minute team meeting to resolve a conflict between two colleagues | Agenda, ground rules, fair process, follow-up action | A plan that assigns blame in advance |
Notice that the strong answer also shows reasoning. If the task allows a short note on assumptions, you can see how the person thinks, not only the final output. Add a line to the brief: “Please add a short note with any assumptions you made.”
Using the task in the interview
Ask each finalist to talk through the answer for ten minutes. Ask: What was the hardest part? What would you do with more time? What did you assume? This helps you check that the candidate did the work, and shows how they respond to questions. It also gives people who write less well a chance to show their thinking in speech. Score the answer and the talk-through on separate lines so that you can see each clearly.
What to do with scores
Treat the work sample as one input. Add the scores to the scorecard, next to the interview scores and the reference notes. Look at the pattern across methods. If a candidate does well in the interview but poorly on the task, ask why. The task might have been unclear, or the interview might have covered a different part of the skill. Do not discard either piece of evidence without a reason. Record your reasoning in the decision record. See how to run a debrief for a method to resolve conflicting evidence.
Keep a library of tasks. After each search, save the brief, the scoring guide and notes about what worked. Remove any task that candidates found unclear. Replace any task whose answers can be found online. Over time you build a set of tested tasks for each role family, and the next search starts faster and fairer.
Frequently asked questions
Should work sample tests be paid?
If they take more than about an hour, or if they create value for your business, pay. For short tasks done in the interview, payment is not usual. Check local wage rules.
How long should a work sample be?
Usually 30 to 90 minutes. Longer tasks reduce the number of people who can apply and can feel unfair.
Can I use a real problem from my business?
Avoid it. Use a made-up case so that you are not taking free work and no confidential data is shared.
What if a candidate refuses to do the task?
Ask why, and offer an alternative if they have a good reason, such as a disability or a time limit. Do not assume lack of interest.
Is a work sample better than an interview?
Each tests different things. A task shows what the person can do now. An interview shows past behaviour and how the person thinks. Use both when needed.
Related guides
- Hiring guides for hiring managers
- How to write a good interview plan for a new role
- Interview scorecard template with a worked example
- Behavioural interview questions with a scoring guide
- All hiring guides for employers
- How to hire your first salesperson without a sales manager
- How to hire your first developer when you are not technical
- How to interview for potential and not just experience
This is general information, not legal advice.