What is a work sample test? How to design one fairly, and how it differs from a take-home
On this page
- What counts as a work sample test?
- What does the federal government's guidance say about work samples?
- How does a work sample differ from a take-home, a skills test or a case interview?
- How do you design a work sample test fairly?
- What do the anti-discrimination laws require of a work sample?
- A copy-ready work sample specification
- How long should a work sample take, and should you pay for it?
- Who runs the work sample when a staffing agency is involved?
- Which work sample mistakes are most common?
- Questions people ask
A work sample test is a selection step in which candidates perform a task taken from the job itself, such as drafting a letter, reconciling a ledger, answering customer emails or fixing a bug, under the same conditions for everyone, and trained assessors score the result or the way it was done against criteria set in advance. The U.S. Office of Personnel Management describes work samples as tasks that "mirror the tasks employees perform on the job" and says they should be used only for skills applicants are expected to have on entry (as of November 2026). A take-home assignment is one way to deliver a work sample: on the candidate's own time and unsupervised, which adds questions of unpaid hours and authorship that a live or proctored work sample avoids.
This page defines the term and its variants, shows how to design one fairly and what the federal rules require, and gives a copy-ready specification. How to score a take-home with a rubric and two reviewers is covered in how to evaluate a take-home assignment, and how to check who did the work in how to verify take-home assignment authorship.
What counts as a work sample test?
Anything where the candidate does part of the job, rather than talking about it or answering questions about it. Common forms:
| Form | What the candidate does | Example (invented) |
|---|---|---|
| Live task | Performs the task during the interview, observed | A recruiter coordinator schedules a four-person panel from three calendars |
| Workstation or hands-on test | Uses the job's tools or equipment on site | A machinist sets up and runs a part to a drawing |
| In-basket exercise | Works through a stack of emails, requests and documents in a set time | An office manager prioritises and answers twelve messages |
| Role-play or simulation | Handles a scripted situation with a trained actor or interviewer | A support agent calms an upset customer and finds the right policy |
| Written work sample | Writes something the job produces | A grant writer drafts a one-page needs statement from supplied data |
| Take-home assignment | Does a task at home by a deadline, often presented afterwards | An analyst builds a short dashboard from a sample dataset |
| Paid trial shift or trial project | Works a short paid period on realistic work | A line cook works a paid prep shift |
OPM's page on work samples and simulations uses the same two families: work samples that re-create specific tasks (its example is an administrative assistant transcribing a memo at a workstation like the one on the job) and broader performance tests, such as a role-play with a trained actor, that measure general competencies like problem solving and communication.
What does the federal government's guidance say about work samples?
OPM's assessment guidance, checked November 2026, describes work samples this way:
- Validity: the tasks are very representative of the job (high content validity) and performance on them relates highly to performance on the job.
- Candidate reactions: applicants often see work samples as very fair.
- Group differences: generally little or no difference between men and women or between races, though OPM notes this depends on the competencies assessed.
- Cost: costly to develop and to run, often one candidate at a time, with trained people observing and rating; tasks need updating when the job's tools change.
- Best use: when the competencies are critical on day one and the applicant pool is small enough to test.
- Limit: if the work will be taught after hiring, a work sample may not be appropriate.
That last point matches the federal Uniform Guidelines, which say content validity is not an appropriate strategy when a selection procedure involves knowledge or skills an employee will be expected to learn on the job (29 CFR 1607.14(C)(1)).
How does a work sample differ from a take-home, a skills test or a case interview?
| Term | Defined by | How it relates to a work sample |
|---|---|---|
| Work sample test | Content: a task from the job | The category itself |
| Take-home assignment | Delivery: unsupervised, on the candidate's time | A work sample done at home; see take-home assignment emails |
| Skills or knowledge test | Questions about the work (multiple choice, short answer) | Asks what the candidate knows, not what they can do |
| Case interview | A business problem solved out loud | A work sample only when the case is built from the job; see what is a case interview |
| Assessment center | Several exercises rated by several trained assessors | Usually contains work samples; see what is an assessment center |
| Realistic job preview | Showing candidates the job's hard parts | Not scored; a work sample can double as one |
| Portfolio review | Past work the candidate already did | An alternative when a new task would be unfair or redundant |
How do you design a work sample test fairly?
- Analyse the job first. List the important tasks and the skills each needs on day one. The Uniform Guidelines expect a job analysis behind any content-valid procedure (29 CFR 1607.14(C)(2)).
- Pick the one or two tasks that matter most and that strong and weak performers do differently. A representative slice of the job beats a long, everything-included exercise.
- Remove what the job does not need. Unfamiliar software the employer would train on, insider vocabulary, or a time pressure the job never has all measure something else.
- Standardise it. Same instructions, materials, time and help for everyone, written down. Decide in advance what the assessor may answer if asked.
- Write the scoring criteria before anyone takes it. What a strong, adequate and weak result looks like, with examples. OPM notes scores come from trained assessors observing behavior or measuring outcomes, such as the number of errors.
- Build in accommodations. Under the ADA, a test must be given so that the result reflects the skill being measured rather than an impaired sensory, manual or speaking skill, unless that skill is what the job requires (29 CFR 1630.11). Say in the invitation how to ask for one; see ADA accommodations in interviews.
- Never use the output. Use a past, disguised or invented problem, and say so in the brief.
- Pilot it. Have two current employees in the role take it. If good performers struggle, the test is measuring the wrong thing or is too long.
- Check results over time. Compare pass rates across groups where you have voluntary self-identification data, and revisit the task when the job's tools change.
What do the anti-discrimination laws require of a work sample?
The EEOC's technical assistance on employment tests and selection procedures lists "sample job tasks (e.g., performance tests, simulations, work samples, and realistic job previews)" among the procedures the federal laws cover, and says employers should give them without regard to protected characteristics, make sure they are validated for the job and purpose, and keep them current as the job changes. In outline, as of November 2026:
| Law | What it means for a work sample | Source |
|---|---|---|
| Title VII | Same test and scoring for everyone; a test that screens out a protected group more than others must be job-related and consistent with business necessity, and scores may not be adjusted by group | 42 U.S.C. 2000e-2(h), (k), (l) |
| Uniform Guidelines | Performance tests are "selection procedures"; validation methods, including content validity, are set out | 29 CFR part 1607 |
| ADA | Reasonable accommodation in how the test is given; no test that screens out people with disabilities unless job-related | 42 U.S.C. 12112(b)(5)-(7) |
| ADEA | No test given only to older applicants; neutral tests with age impact need a reasonable factor other than age | EEOC technical assistance |
| Records | Keep the materials, scores and decisions for at least one year | 29 CFR 1602.14 |
Executive Order 14281 (April 2025) told federal agencies to deprioritize disparate-impact enforcement (90 FR 17537); the statute still allows candidates to bring those claims, and state laws can add their own. Honesty and personality tests raise different questions, covered in pre-employment testing laws. This is not legal advice.
A copy-ready work sample specification
WORK SAMPLE SPECIFICATION (internal)
Role and level: [title, level] Requisition: [number]
Skill(s) measured (needed on day one): [1-2 skills, from the job analysis]
Task: [what the candidate produces or does, in two sentences]
Why this task: [the job duty it represents; how often it happens]
Delivery: Live in interview / On-site workstation / Take-home / Paid trial
Time allowed: [minutes] Total candidate time in the process: [hours]
Materials given: [files, data, tools; all invented or disguised]
Allowed: [calculator, internet, AI tools: yes/no, and how to disclose]
Assessor may answer: [list] Assessor may not: [list]
Scoring criteria (written before first use):
Criterion 1: strong = [...] adequate = [...] weak = [...]
Criterion 2: strong = [...] adequate = [...] weak = [...]
Assessors: [two names, trained on this spec]; score independently
Accommodations: offered on request via [contact]; examples [extra time,
screen reader compatible files, printed materials]
Alternative for candidates who cannot do the task format: [portfolio
walkthrough / live session], scored on the same criteria
Use of output: none; submissions deleted after decision on [date rule]
Review date: [when the task is checked against the job again]
How long should a work sample take, and should you pay for it?
As short as it can be while still showing the skill. The design steps above usually cut the task to the one piece of the job that predicts the rest. Add up everything your process asks: interviews, the work sample and any presentation. For longer or unsupervised tasks, many employers pay candidates and offer a live alternative; the options and a published example of a paid trial are in how to evaluate a take-home assignment. Whatever you choose, apply it to every candidate for the role.
Who runs the work sample when a staffing agency is involved?
Agree it in writing with the client before the first submittal. If the agency runs the test, it needs the client's task, criteria and pass level, and it should give every candidate for the order the same test, not only the ones it is unsure of. If the client runs it, the agency's job is to tell candidates what to expect, how long it takes, whether it is paid, and how to ask for an accommodation. Either way, the scores and materials are hiring records that someone has to keep for the retention period.
Which work sample mistakes are most common?
| Mistake | What it causes | Fix |
|---|---|---|
| Testing skills you will train | Screens out people who would learn quickly | Test only day-one skills |
| A different task for each candidate | Results cannot be compared | One spec per role, versioned |
| Scoring by overall impression | Assessor bias decides | Criteria written first, two assessors |
| A live client problem as the task | Free work, and candidates notice | Invented or disguised materials |
| No accommodation route | ADA risk and lost candidates | Name a contact in the invitation |
| An eight-hour task for a first-round screen | Strong candidates with jobs drop out | Short task early, longer only at the final stage |
Questions people ask
What is an example of a work sample test?
Asking an administrative assistant applicant to format and file a set of documents, a bookkeeper to reconcile a short ledger, a support agent to answer three real-looking customer emails, or a developer to fix a bug in a small codebase. The task is taken from the job, every candidate gets the same one, and trained assessors score the result or the way it was done against criteria set in advance.
Is a take-home assignment a work sample test?
It is one way to deliver one. A work sample test is defined by its content, a task from the job, while a take-home is defined by where and when it happens: on the candidate's own time, unsupervised, usually with a deadline. A work sample can also be done live in the interview, at a workstation on site, or in a paid trial shift. Take-homes add two problems of their own: unpaid hours and the question of who did the work.
Are work sample tests legal?
Yes, when they are job-related and given the same way to everyone. The federal Uniform Guidelines treat performance tests as selection procedures, so a work sample that screens out one race, sex or ethnic group more than others must be shown to be job-related, for example by showing that it is a representative sample of the job's content. The ADA also requires reasonable accommodation in how a test is given. This is general information, not legal advice.
How long should a work sample test be?
As short as it can be while still showing the skill that matters. Pick the one or two tasks that best separate strong from weak performers, cut everything else, and add up the total time your whole process asks of a candidate. Longer tasks, especially unsupervised ones, are where employers increasingly pay candidates and offer a live alternative.
Can I use a work sample test for a job you train people to do?
Usually not for the parts you will train. The Office of Personnel Management says work samples should be used only where applicants are expected to have the skill on entry, and the Uniform Guidelines say content validity is not an appropriate support for testing skills an employee will be expected to learn on the job. Test the entry-level skills and leave the rest to training.