Work samples, job simulations, take-home assignments and real-work hiring
Employers use several formats to see applicants do something job-like before deciding. This page compares four of them on the same dimensions, says where each fits, and states what is and is not known. None is always best. The right choice depends on the hiring question and on how much evidence, effort and preparation you can support.
Summary
- The four formats overlap. A take-home can be a work sample, and a simulation can be a work sample delivered in a recreated setting.
- What differs most is whether the situation changes in response to what the applicant does, how much the employer must prepare, and how the work is reviewed.
- “Real-work hiring” is Calavion’s proposed category language. It is not an established scientific classification, and it has no validation evidence of its own.
- A simpler method is often enough. Choose the least burdensome format that answers the hiring question.
Method and scope
This is an editorial comparison, not a study. It draws on the peer-reviewed and official sources listed below, which are summarized in What research says about seeing applicants perform real work, and on Calavion’s Product Truth Reference and application code as of 25 September 2026. Cells describe typical tendencies, and any format can be designed well or badly. Calavion offers one of the four formats, so read its column with that in mind. This page was first published on 10 September 2026 as a three-way comparison and was revised on 25 September 2026 to add job simulations and real-work hiring, and to remove two citations that could not be re-verified.
Working definitions
- Work-sample test. The US Office of Personnel Management (OPM) describes work-sample tests as requiring applicants to perform tasks or work activities that mirror the tasks employees perform on the job.
- Job simulation. A recreated work setting or scenario. Lower-fidelity forms include situational judgment tests; higher-fidelity forms include assessment-center exercises (Lievens & Patterson, 2011).
- Take-home assignment. An exercise completed independently over a stated period from a prompt, usually ending in a submitted output. This is a working editorial definition.
- Real-work hiring. Calavion’s proposed term for an approach in which a company brings work it has already handled into hiring, an environment is prepared from that work and approved by the company, applicants work through the approved version, and the employer reviews a record of what happened. It is a product category description, not a scientific classification.
Comparison at a glance
Four ways to see applicants perform job-like work
Scroll sideways to see all columns.
| Dimension | Work-sample test | Job simulation | Take-home assignment | Real-work hiring (Calavion) |
|---|---|---|---|---|
| Source of the work | Tasks chosen by the employer to mirror the job | Scenarios designed by the employer, a vendor or assessment specialists | A prompt the employer writes, often a stand-in for real work | Work the employer has already handled, prepared by Calavion with the company |
| Relationship to actual responsibilities | Direct by design | Ranges from generic to job-specific | Varies; often a proxy | Intended to be direct because it starts from the company’s own work. No real company case has yet been prepared and run, so this is untested |
| Changes in response to actions | Not by default | Can, for example role-players or branching | No, once the prompt is issued | Yes: state changes, delayed messages and consequences are part of the design |
| Company rules and records | As designed by the employer | As designed | Whatever is attached to the prompt | Records and an operating procedure are included. Rules are not enforced: a rule-breaking act is recorded, not blocked |
| Communication | Depends on the task | Ranges from choosing a response to live role-play | A written deliverable, sometimes a presentation | Free-text messages to simulated parties. Replies from those parties are scripted events, not open conversation |
| Consequences | Usually only the task result | Can include them | None until someone reviews the output | Immediate and delayed consequences inside the environment |
| Time burden on the applicant | A bounded task | Varies widely | Set by the employer and often unclear; no reliable figures were found in the sources reviewed | A hard session limit exists. Calavion has no validated figure for how long applicants take and claims none |
| Review method | Raters or a scoring key | Scoring keys or trained assessors | People review the deliverable | A person reads Replay. The employer can also type its own observations and interpretations against approved requirements |
| Automated scoring | Varies | Varies | Usually none | Calavion does not currently produce automatic scores, applicant rankings or hiring decisions |
| Applicant experience | Reviewed favorably in meta-analyses of applicant reactions | Varies by design | No peer-reviewed evidence located | No evidence yet. No applicant has completed a real hiring process with it |
| Setup effort for the employer | Define the task, conditions and review | Substantial for high-fidelity designs | Low to moderate | High today: Calavion prepares each environment by hand with the company, and the company reviews and approves an exact version |
| Appropriate use | A discrete capability can be shown directly | The capability depends on situational context | A self-contained deliverable is what matters | Sequence, checking records, communication, consequences and handover matter, and the work fits the one environment shape currently supported |
| Main limitations | Moderate validity in meta-analyses (.33 in Sackett et al., 2022), mostly concurrent evidence, group score differences reported | Cost and, for high-fidelity forms, effort; some evidence comes from single studies | Effort can expand, outside help is invisible, and no peer-reviewed evidence specific to take-homes was located | No validation, one environment shape, manual preparation, and recorded actions are not a record of reasoning |
Cells describe tendencies, not rules. Evidence figures are explained in the evidence review. “Not by default” means possible but not required.
When each format fits
A work-sample test
Use it when a capability can be demonstrated through a bounded activity under clear conditions, such as editing a passage or drafting a customer update from supplied facts. Its strength is directness. Its limit is scope: a short task may show one capability and leave investigation, collaboration and follow-through outside the frame. OPM notes such tests may be costly to develop and suit cases where the measured competencies are highly critical and the number of applicants is limited.
A job simulation
Use it when the capability depends on a situation, for example weighing a scenario or handling a role-play. Lievens and Patterson (2011) found in one study of 196 applicants that a higher-fidelity assessment center added predictive value beyond a lower-fidelity situational judgment test, but that is a single setting and cost was not the question.
A take-home assignment
Use it when independent planning and a considered deliverable are genuinely relevant and the prompt is bounded enough to keep applicant effort proportionate. Clarify the expected effort, whether outside tools or help are allowed, whether the company will use the work commercially, how it will be reviewed, and how an applicant can ask a question or request an adjustment. A take-home can hide the process: which alternatives were considered, where the applicant struggled, and what help was used.
Real-work hiring
Consider it when the order of decisions matters, information must be checked across records, communication and unfinished work matter, and the responsibility resembles the one environment shape Calavion can currently run, an order-cancellation and refund handover in customer support. If a single output answers your hiring question, a simpler format is the better choice. See Suitable work and Availability.
Questions to ask before choosing
- What job-relevant question must this exercise answer?
- Do we need a final output, the sequence of work, or both?
- Do changing circumstances matter to the responsibility?
- What is a proportionate amount of applicant effort, and have we stated it?
- Can the exercise run without production access or unnecessary sensitive information?
- Can every applicant work under comparable conditions?
- Do reviewers have the time and guidance to inspect the evidence?
- How will we check whether the exercise disadvantages any group of applicants?
- What can this method not tell us?
- Would a simpler method answer the question?
Multiple-choice tests
What they are good at: they are cheap, quick to administer, and the answers are easy to count. If a role needs a fixed body of knowledge (a regulation, a procedure), a well-built knowledge test measures that directly.
Where they stop: a selected answer shows recognition, not conduct. It cannot show whether someone investigates before responding, writes to a frustrated person clearly, notices that two records disagree, or leaves a handover. The reviewer sees a result, not a route.
How Calavion differs: the applicant is not limited to selecting a prewritten answer to a question; they inspect information, communicate and act. The cost is that a reviewer must read evidence rather than count answers.
AI résumé screening
What it is for: reducing a large pile of documents to a manageable one. Calavion does not claim that such tools do not work for that purpose.
The structural limit: it processes an account of work, not work. Whatever the tool is good at, the input is still what the applicant wrote about themselves. Ask any vendor whether a rejected applicant can see why.
How Calavion differs: Calavion does not give applicants an overall score, generate an ordered shortlist, or recommend whom to hire. People decide, with a record they can open. What Calavion produces is a record a person can open and read. These are different jobs; they can be used at different stages.
Limitations of this comparison
- It compares tendencies across formats that vary a great deal internally.
- The strongest evidence concerns work-sample tests and interviews. Little peer-reviewed evidence was located for take-homes, and none for interactive environments with delayed consequences or replay records.
- Calavion’s column describes one implemented environment and manual setup. It is not evidence that the approach works for any employer.
- Calavion has not demonstrated predictive validity, fairness, reduced bad hires or improved hiring outcomes.
Sources
- US Office of Personnel Management. Work samples and simulations. opm.gov (page shows no publication date; accessed 25 September 2026).
- Sackett, P. R., Zhang, C., Berry, C. M., & Lievens, F. (2022). Revisiting meta-analytic estimates of validity in personnel selection. Journal of Applied Psychology, 107(11), 2040–2068. doi:10.1037/apl0000994
- Roth, P. L., Bobko, P., & McFarland, L. A. (2005). A meta-analysis of work sample test validity. Personnel Psychology, 58(4), 1009–1037. doi:10.1111/j.1744-6570.2005.00714.x
- Lievens, F., & Patterson, F. (2011). The validity and incremental validity of knowledge tests, low-fidelity simulations, and high-fidelity simulations. Journal of Applied Psychology, 96(5), 927–940. doi:10.1037/a0023496
- Hausknecht, J. P., Day, D. V., & Thomas, S. C. (2004). Applicant reactions to selection procedures. Personnel Psychology, 57(3), 639–683. doi:10.1111/j.1744-6570.2004.00003.x
- Anderson, N., Salgado, J. F., & Hülsheger, U. R. (2010). Applicant reactions in selection. International Journal of Selection and Assessment, 18(3), 291–304. doi:10.1111/j.1468-2389.2010.00512.x