"Walk me through how you approached this."
Hiring guide
How to build a work-sample test for a remote hire
A CV tells you where someone has worked. A short, real task tells you how they work. Here's how to design one that's fair to candidates and useful to you.
Interviews are good at showing how someone talks about their work. When you hire remotely, though, most of the job happens in writing, in shared docs and in a task queue, not on calls. A work-sample test closes that gap: you ask finalists to do a small piece of the actual job and judge the result.
This guide covers what a good test looks like, how to build one in an afternoon, and the mistakes that make tests unfair or useless.
What a work-sample test is
The US Office of Personnel Management (OPM), which publishes assessment guidance for federal hiring, describes work-sample tests as tasks that mirror what people actually do in the job. Its example is an administrative assistant asked to type up a memo or file a stack of paperwork.
OPM's summary of the research is useful for anyone hiring:
- 1
Tasks drawn from the real job are representative of the work.
- 2
Performance on them relates closely to performance on the job.
- 3
Candidates tend to see them as fair.
OPM also flags two limits:
- Good tests take time to build.
- Tests should only measure skills the person needs from day one. If you plan to train someone on a task after they start, don't test them on it.
In other words, a work sample isn't a puzzle or a trick question. It's a preview of the job.
Start with the brief
The test should come straight from your role brief. If you've used our one-page hiring framework, you already know the role's outcomes for the first 90 days and the few skills that matter most. Pick the task that best represents those skills.
Ask yourself: "What will this person do in their first month that tells me, by week four, whether this is working?" That task, cut down to size, is your test.
Some examples:
Swipe to see more →
| Role | A test drawn from the real job |
|---|---|
| Executive assistant | Sort a sample inbox of 20 emails into reply, delegate, schedule and archive, then draft three replies. |
| Bookkeeper | Reconcile a short sample bank statement against a ledger and flag anything that doesn't match. |
| Developer | Fix a small, known bug in a sample repo and write the pull-request description. |
| Customer service agent | Answer five sample tickets in your brand's voice, including one where the right answer is "no". |
| Marketing or content | Turn a rough set of notes into a short post and three subject lines. |
Notice what each one includes: some writing. For remote roles, how someone explains their work is part of the work.
Keep it small
Long tests filter out the wrong people. Strong candidates often have the least spare time, and a weekend-long project tells you mostly who had a free weekend.
A few rules of thumb:
- Aim for a task that takes one to two hours. Tell candidates how long you expect it to take and ask them to stop there.
- If it has to be longer, offer to pay for it. It's fair to the candidate and it shows them you value their time.
- Test only finalists. A work sample is for the last two to four candidates, not the first 50 applicants.
- Give a sensible deadline. Two or three working days lets people fit it around their current job, and it suits candidates in other time zones.
Never use real data or real client work
Build the test on sample material: a made-up inbox, an anonymized spreadsheet, a toy repo. That's important for three reasons:
- 1
Security
A candidate isn't on your team yet and shouldn't see live customer data, credentials or financial records.
- 2
Fairness
Using a test task as unpaid production work is a fast way to lose good candidates' trust.
- 3
Comparability
Every finalist should get exactly the same task, so you're comparing like with like.
Write the instructions like a real brief
How you set the task is part of the test. Write it the way you'd brief a new team member:
- the goal and who the output is for
- what's provided and what format the result should be in
- the time limit and the deadline
- any tools they can use, including your rules on AI tools (allowed, allowed with disclosure, or not allowed)
- who to contact with questions
Then watch what happens. Candidates who ask one or two sharp clarifying questions are showing you exactly what you'd want from them on a remote team. Candidates who guess at something important without asking are showing you something too.
Decide how you'll score it before anyone submits
This step is easy to skip, and it's the one that makes a test fair. Before you send the task, write down three to five things you're looking for and what "good" looks like for each. For an executive assistant inbox test, for example:
- Judgment: Are the right emails flagged as urgent?
- Accuracy: Are dates, names and details right?
- Writing: Are the drafted replies clear, polite and ready to send?
- Communication: Is there a short note explaining the choices made?
- Instructions: Was the time limit and format respected?
Score each one on a simple scale (for example 1 to 3) and add a line of notes. If more than one person reviews the work, score separately first and compare afterward. That keeps the loudest opinion in the room from deciding.
Talk it through afterward
A test result on its own only shows you the output. A 20-minute follow-up call shows you the thinking behind it. Ask:
"What would you have done with another hour?"
"Where did you have to make an assumption?"
"If I pushed back on this choice, what would you say?"
These questions are hard to fake. They also give the candidate a real taste of working with you, which helps them decide too.
Close the loop with everyone
Every finalist who completes a test should hear back, with a short reason if they're not moving forward. It costs you ten minutes, and it's how good candidates remember you when the next role opens.
Where this fits in a vetting process
A work sample is one stage, not the whole process. It works best alongside a clear brief, a look at past experience, structured interviews and references. We cover how those fit together in what "vetted" should mean, and the common traps in five mistakes to avoid when hiring remotely.
When JiveTalent runs a search, the test is built from your brief, so the shortlist you see has already shown how they work. You still make the final call.
Want help designing the test for your role?
Tell us about the role. We'll talk through the brief, what a fair test would look like, and how a search would run. Book a call, or see how it works.
Sources
- US Office of Personnel Management, "Work Samples and Simulations": https://www.opm.gov/policy-data-oversight/assessment-and-selection/other-assessment-methods/work-samples-and-simulations/