Skip to content

Interviews

Are Unpaid Take-Home Assignments Worth Doing?

A take-home assignment can win you the job or waste a weekend. Use this framework to decide fast, negotiate scope, and spot free-work traps.

ResumeFlex guidesIdeas for your next move
PublishedUpdated 13 min read12 sections
Share article
Share on XShare on LinkedIn

On this page12 sections

Key takeaways

  • Work sample tests still rank among the top five predictors of job performance (.33 validity in Sackett et al.'s 2022 re-analysis, summarized by SIOP); that's why employers keep asking.
  • The cost is real: before Dropbox dropped take-homes, 20% of candidates never finished them and only about 10% passed (Holloway, 2022).
  • The fair version is narrow: 2–4 hours, hypothetical material, stated criteria, late stage (Wonsulting, June 2026; Ask a Manager).
  • Paid trials exist but are rare, Linear pays every trial candidate a daily rate disclosed in advance (April 2026) and Automattic pays $25/hour (November 2026).

Sometimes; a take-home assignment is worth doing when it’s short, sits late in the hiring process, uses hypothetical material, and comes with clear evaluation criteria. It’s worth declining when it eats a weekend, mirrors real client work, or lands in your inbox before a single human at the company has spoken to you. The answer lives in a narrow band, and most of the frustration comes from employers who push past it.

A take-home assignment is unpaid, self-directed work a company asks a candidate to complete on their own time to evaluate job-relevant skills. The sections below walk through the research that keeps this format alive, the real cost, the free-work traps, and a decision framework with a script you can send before you commit a single evening.

Why do employers still use take-home assignments?

Employers keep asking because work samples predict job performance better than almost any other hiring signal. In a 2022 re-analysis of roughly a century of selection research, psychologist Paul Sackett and colleagues re-estimated what actually forecasts performance: structured interviews came out highest (.42), with job knowledge tests (.40), biodata (.38), and work sample tests (.33) all in the top five, ahead of general cognitive ability (.31). SIOP, the Society for Industrial and Organizational Psychology, summarizes those revised validity estimates, and they explain why the format survives decades of candidate complaints.

Before Dropbox moved away from take-homes, one in five candidates never completed the assignment, and the pass-through rate was roughly 10%. That’s a lot of unpaid hours for a one-in-ten shot. (Holloway Guide to Technical Recruiting, 2022)

There’s a fair argument for the format, too. The Holloway Guide notes a take-home has the lowest false-negative rate of any interview format, because you can’t hide weak work over six hours the way you can survive an hour of polished talking. It also favors strong candidates with unconventional backgrounds who interview badly but work brilliantly.

The ground is shifting, though. Fabric, an interview-integrity vendor, reported in January 2026 that 59% of hiring managers suspect candidates of using AI on assessments, and that AI tools can finish most take-home tasks in under five minutes. Read vendor numbers cautiously, but the direction matches what candidates report across engineering and design communities: a take-home now partly measures whether you have a subscription, which weakens the old validity argument.

The best employers have responded by pairing the task with a live walkthrough of your decisions, which, as you’ll see, is also your best negotiating lever.

What does an unpaid take-home assignment actually cost you?

Two to twenty hours of unpaid, non-transferable work, with no guarantee anyone reads it. The Wonsulting guide updated in June 2026 puts a fair scope at roughly 2–4 hours, and calls 12 or more hours “poor scope management” rather than commitment.

But real cases run long: a Hacker News thread from August 2022 starts with a candidate who spent about 6 hours on a 2,000-line assignment, got ghosted, and concluded they’d never again do more than an hour unpaid. Ask a Manager readers have reported 15-plus-hour marketing campaigns demanded just to reach a second-round interview, and ~20-hour cases besides.

The math is worse than the hours. Because employers rarely set an upper bound, you’re compared against rivals who sank 40 hours into the same prompt, while your own expected value is hours spent times a low pass rate.

Holloway’s Dropbox data adds a sharper edge: less-competitive candidates were the most likely to finish, because the strong ones already had other offers. Take-homes quietly filter for people with free evenings; Holloway flags that this discriminates against candidates with families, second jobs, or financial strain.

Companies have skin in the game on the other side. CareerPlug’s candidate-experience report from February 2025 found 36% of job seekers had declined an offer after negative interview interactions, and 66% said a good experience influenced their decision to accept. Greenhouse’s 2025 Workforce Survey Report, built on feedback from more than 2,000 job seekers and published in early 2026, documents application drop-off and ghosting as defining features of the current process.

If a take-home is how you’re being screened, you’re also screening, one-way interviews and async tasks tell you how a company treats people’s time before you’re inside it.

When is a take-home assignment just free work?

When the company could use what you produce. A legitimate work sample tests a skill on fictional material; free work is deliverable-quality output the company could hand to a client or ship. The line isn’t fuzzy in practice, and candidates keep documenting cases on the wrong side of it.

The examples from Ask a Manager’s archive are instructive: a commenter who spotted four of her interview story ideas published on the company’s site within a year, a bakery that allegedly ran about 50 unpaid “working interviewees” through four weeks of real shifts, and a social media plan request that felt wrong precisely because, as the candidate put it, it could be applied directly. Business Insider’s July 2024 report quotes a candidate who describes building social media plans for real clients as “free consultancy”, and notes a company that paid her a $90 gift card for a two-hour test when she pushed.

Here are the near-dealbreaker red flags:

  • The brief uses the company’s real client, real data, or live production code.
  • Nobody will tell you the expected hours, the evaluation criteria, or how the output will be used.
  • The request arrives before any conversation with a human.
  • Scope keeps expanding after you’ve agreed to the original ask.
  • Submission is followed by silence, or worse, an offer that feels pre-scripted.

And the green flags of a legitimate one:

  • A stated time ceiling of roughly 2–4 hours that matches reality.
  • Fictional or sanitized scenarios, named evaluation criteria, and a promise of feedback.
  • A slot late in the funnel, after you’ve met the hiring manager.
  • A debrief conversation where you defend your choices; Wonsulting calls that debrief half the evaluation.

In design circles this is the old spec-work fight, and designers have been refusing unpaid “challenges” for years; engineers reached the same place once AI made long take-homes cheap to fake. Both communities learned the same lesson: the more the output resembles the job, the more you should treat the request as negotiation, not homework.

Should you do it? The hours × stage × fit × signals framework

Use these four checks before agreeing. If the assignment passes all four, set a time limit and complete it. If two or more raise concerns, ask for changes or decline.

  1. Stage check. Has a human interviewed you already? A take-home before any conversation is a screening shortcut they should be paying for, and it’s fair to push it later.
  2. Hours check. Ask how long the assignment should take. If the estimate is over four hours, or they won’t give one, ask to reduce the scope. The 2–4 hour norm gives you a starting point for that conversation.
  3. Fit check. Does the task test the core skill of the job, or is it busywork? A pricing analysis for a pricing role is a work sample. A three-month content strategy for a role that would mostly manage vendors is free labor.
  4. Signals check. Fictional data, named criteria, promised feedback, and a respectful deadline are green flags; real client work and vague answers are disqualifying ones. Then weigh stakes: a top-choice role late in the funnel earns more of your patience than a maybe.

One caveat as this format ages: seniority changes the calculus. The more senior the role, the more your portfolio already answers the question a take-home tries to ask, which is why senior candidates increasingly decline on principle, and companies increasingly accept it.

The four kinds of take-home assignment, ranked by what they’re worth

Most requests fall into one of four buckets, and the right response differs for each.

Scenario What it looks like Worth it? Your move
Small and scoped 1–3 hours, fictional data, criteria named, after a real conversation Yes Do it well, cap your hours, document assumptions
Large but legit 4–8 hours, senior role, stated scope, feedback promised Only for roles you’d take Negotiate down to 2–4 hours or a narrower slice first
Real-work-shaped Their actual client, real data, “treat it like the real thing” No Decline, or offer a parallel version on a fictional brief
Compensated trial Paid at a disclosed rate, late stage, real project Yes Treat it as a multi-day audition and track your hours

How to negotiate scope before you start

Ask three questions in writing before you open the brief, hours, criteria, and intended use. Asking isn’t risky; Wonsulting’s June 2026 guide notes that a company which penalizes you for scoping questions is showing you its culture early. Here’s a script to adapt:

Happy to do this, I want to give you something genuinely evaluative. Before I start, three quick questions: roughly how many hours is this scoped for, what will you be assessing it on, and will the output be used only for my evaluation? If it’s likely to run past four hours, would you be open to a shorter version, or to me walking your team through comparable work I’ve already done?

That last sentence is the pivot. If they hold firm on a 10-hour deck, escalate in this order:

  1. Restate your cap: “I can give this 4 focused hours; here’s what I’d cut to stay inside that.”
  2. Offer a paid rate for anything longer, the way Datapeople compensates every candidate who completes its work demonstration.
  3. Propose an alternative format entirely: a live working session or a past-work walkthrough.
  4. If they refuse all three, decline warmly. “I’ve enjoyed the conversations, but I can’t commit to 10 unpaid hours. I’d still welcome a shorter option if the role stays open.”

If you accept: timebox it and scope it yourself

Set a time limit before starting the assignment. State the scope and assumptions in your submission so the reviewer can see what you completed within that limit.

  • Ask the three questions from the script above and save the answers.
  • State your assumptions in one paragraph at the top: scope, data gaps, and what you’d do with more time.
  • Name your tools honestly, if you used AI, say so and be ready to defend every decision, because the debrief is where you’re really evaluated.
  • Stop at your cap even if it’s imperfect. A B+ submission at 3 hours beats an A submission that cost you 12, and interviewers know the difference between polish and judgment.
  • Structure the writeup around the story of your choices; the STAR format works better than most candidates expect for framing situation, task, action, and result in a README.

What fair looks like: the paid-trail handful

A small group of companies has already solved this, which makes the rest harder to excuse. Linear hires every role through paid work trials (2–5 days of real project work, at a daily rate shared in advance) and argues that interviews “favor people who present well under pressure, not necessarily people who are capable of great work.”

Automattic pays every trial participant $25 per hour, typically for 25–40 hours of work on real tools (November 2026). Business Insider profiled Datapeople’s practice of compensating every candidate from entry level to VP.

None of these companies report regretting it, and the formats keep them honest: when a company pays for your work sample, it stops asking for work and starts paying for signal.

No federal rule addresses interview take-homes directly, so the answer depends on whether the work is real. The Department of Labor’s position, stated in Fact Sheet #71 (updated January 2018), is that for-profit employers must pay for work performed; its eLaws advisor allows unpaid status only where the trainee benefits, the employer gets “no immediate advantage,” and both sides understand no wages are owed. Apply that lens and most short hypothetical tests are fine, while a deliverable the company could bill for edges toward compensable work.

This isn’t legal advice; treat any real-work ask as both an ethics problem and a legal gray zone.

Ownership is cleaner. Absent an employment relationship and a signed IP assignment, copyright in what you create stays with you; a point Ask a Manager commenters raise about candidates whose work was later used. Read anything the company asks you to sign before submitting, and keep your own copy of everything you send.

If the assignment terms claim your work product, that’s another data point about the employer, worth carrying into your pre-offer checks.

The verdict

A well-built take-home assignment is the single fairest hiring format we have, and most companies implement it badly. The safe trade: do the 2–4 hour, hypothetical, late-stage version with clear criteria, because it’s your best chance to be judged on real work. Decline the 15-hour, real-client, pre-screening version without guilt; the research on validity doesn’t require 15 hours, and the employer that demands them is telling you how they’ll treat your time after you’re hired.

The format isn’t the problem. Scope is.

FAQ: take-home assignments, answered

How long should a take-home assignment be? About 2–4 hours is the fair ceiling for most roles. Wonsulting’s June 2026 guide recommends that range and treats 12-plus hours as a scope failure; anything longer deserves pushback or compensation.

Do employers have to pay for take-home assignments? There’s no federal rule that addresses them by name. DOL guidance requires pay for productive work that benefits a for-profit employer, so short hypothetical tests are generally fine while real deliverables occupy a gray zone.

Who owns the work I submit in a take-home assignment? You generally do, because copyright starts with the creator and you’re not an employee. Read any agreement they ask you to sign before submitting, and keep copies of everything you send.

Can I use AI on a take-home assignment? Only if their policy allows it, and then transparently. Fabric reported in January 2026 that 59% of hiring managers already suspect AI use, so the safe move is to declare your tools and be able to defend every line.

Should I ask to be paid for a take-home assignment? You can, and it works at companies that already pay. Linear, Automattic, and Datapeople all compensate candidates, but they’re exceptions, so ask politely, tie the request to length, and expect refusal at most employers.

Do take-home assignments actually predict job performance? Work samples do. In Sackett and colleagues’ 2022 re-analysis of selection research, work sample tests scored .33 (among the top five predictors, ahead of general cognitive ability at .31) though AI use has since muddied the signal for long tasks.

What’s the difference between a take-home assignment and a paid work trial? A take-home is unpaid homework, usually a few hours on a simulated task. A paid work trial is a disclosed-rate, multi-day stretch of real work at a late stage, closer to a contract audition than a test.

Can I decline a take-home assignment without losing the job? Sometimes, especially with seniority or a strong portfolio. Offer alternatives (a live working session, a past-work walkthrough, a shorter version) and take the refusal of all three as useful information about the employer.

Sources

Written by ResumeFlex. Questions or corrections? Contact ResumeFlex. Read our source and correction standards.

Filed under

Back to guidesBack to top

Keep reading

More in Interviews