The Take-Home Brief Standard, Graded to Accept, Rescope, or Decline

A recruiter has sent a take-home or case-presentation brief. Before it costs you a weekend, you have to decide whether to do it, rescope it, or decline. This standard grades the employer's request, not your output. It checks the brief against eight fixed pass or fail criteria, maps the result to one of three verdicts, and gives you a clause you can send for every criterion that fails. Two people grading the same brief should reach the same verdict.

The eight criteria a take-home brief must pass

A take-home brief passes this standard only when it meets all eight criteria:

• a stated time cap
• expected hours at or under the unpaid line
• placement after a human conversation
• a defined deliverable
• synthetic inputs
• clean work-product and retention terms
• a written review and feedback commitment
• a completion window that includes a weekend

A take-home is any assignment you complete on your own time, outside a live interview, so the employer can assess you. A case-presentation brief is a take-home that ends in a live presentation, and this standard grades both the same way.

Time and process criteria

Content and terms criteria

What each criterion proves

• Stated time cap shows the employer has considered your time. It does not show they considered it accurately. That is criterion 2's job.
• Realistic-hours check shows the real cost fits inside what unpaid assessment can fairly ask.
• Stage placement shows a person has already judged you worth the hours. Without it, the take-home is a filter that costs the employer nothing and costs you a weekend.
• Defined deliverable shows the grader will score what the cap allows and not what an unlimited budget would produce.
• Synthetic inputs shows the output is an assessment and not unpaid production work. This is the criterion with the most legal weight behind it.
• Work-product and retention shows you will still own what you made, and that it will not be kept and reused.
• Review and feedback shows someone is accountable for reading the work by a known date.
• Completion window shows you can do the work without giving up a working week.

How long should a take-home assignment take?

Published guidance puts the unpaid default at 2 hours and the ceiling at 3 to 4 hours, with anything past that paid. The stated number is roughly half of what the work really costs, so this standard grades the doubled figure against a fixed line of 4 expected hours.

None of the caps below comes from a survey. They are coach and employer-side rules of thumb, and they mostly agree.

The only published distribution I found says about 76% of assessments were scoped at between 50 minutes and 4 hours, with a median of 3 hours. That page does not state its sample size. No recruiter survey breaking caps down by role and seniority has been published.

The 2x conversion

The stated hours are written by someone who already knows the codebase and the answer. Alex Yang, founder of Tapioca, puts it this way: "candidates that do well on take-home assessments tend to take a lot longer, around 2x longer as a rule of thumb." In one documented case, a take-home listed at 3 to 5 hours took the engineer who was hired closer to 7 to 8. rankid's working rule is that a brief saying three to four hours, with no other detail, should be planned as six.

Expected hours = stated hours x 2, plus setup time for any tooling or environment you have not used before. I put the pass line at 4 expected hours because that is the highest unpaid ceiling any source in the cap table allows. A team adopting this standard can set a stricter line, but it has to fix the line before grading starts, or two graders will disagree.

Two consequences follow from that line:

• A brief stated at 2 hours sits exactly on the line and passes. A brief at the 3-hour median costs about 6 hours and fails unless it is paid.
• Seniority does not move the line. One guide suggests 4 to 8 hours for senior roles but does not source its figure. Under this standard, senior work over the line is not unacceptable, but it is paid work.

Is this take-home assignment free work?

A take-home is free work when its output is usable by the business for anything beyond assessing you. The legal tests I could find depend on what the output is used for, not on how long the task takes. That makes criterion 5, synthetic versus shippable inputs, the most legally grounded criterion in this standard, and the hours criterion the weakest.

What the sources say, by jurisdiction:

• California. Under section 46.7 of the DLSE Enforcement Policies and Interpretations Manual, "try-out time can only be unpaid if all of the following are met": no training, "no productivity derived from the work performed by the prospective employee", and a period that is reasonable under the circumstances. If the work is real correspondence, or a meal served to a paying customer, it must be paid.
• UK. "Current law does not define a 'trial work period' or state precisely when at least the NMW / NLW must be paid." The minimum wage guidance gained a section on unpaid work trials in December 2018. In a written answer (PQ122161), minister Kevin Hollinrake said "an unpaid trial lasting more than one day is highly likely to be illegal in all but very exceptional circumstances."
• US federal. I found no primary Department of Labor source on take-homes specifically. The six-factor training test appears only as cited in a Massachusetts opinion letter, so the federal position has not been confirmed from the DOL itself.

This is not legal advice. It is the reason criterion 5 is graded hard: shippable work on an unpaid basis is the case these tests were written for.

The test that works in practice is whether the output could ship on Monday. Real quarterly numbers, a live product's backlog, a pitch for an account they are actually working: each of those fails, however the brief describes it.

If the employer offers to pay for shippable work at a stated rate, you are no longer looking at a take-home. You are looking at a paid work trial, and a different standard should grade it.

Ownership terms live in more than one place

One secondary source sums up the default: "by default, you own what you create". That can change if you sign anything that grants rights, including the terms of an assessment platform account. On an unpaid task, this standard fails any clause saying the work "constitutes 'work made for hire' owned exclusively by Employer" or that you irrevocably assign patent and copyright. That wording comes from employment agreements, which is exactly why it has no place in an assessment.

No primary US or UK legal source specific to take-homes has been published, and I have not verified the UK statutory default. If any layer is ambiguous, ask the employer to confirm in writing that you keep the rights, and ask what happens to your submission after review.

How the verdict is mapped

The verdict comes from a fixed mapping. All eight criteria pass: Accept. Every fail is one the employer can fix and they agree in writing: Accept on Rescoped Terms. A hard fail that stays unresolved, or a rescope request that is refused or goes unanswered: Decline with a Substitute.

Criteria 5 and 6 are hard fails on unpaid work, because the fix changes what the task is, not just its size. The other six are soft fails. They are gaps in scope or process, and a written answer can close them.

Why the order of checks does not matter here

The sources disagree on what to check first. hirecomb treats stage placement and a written feedback date as gates that come before hours. Career Vision asks about hours first. rankid says to decide on the doubled figure before anything else. In this standard the order changes nothing, because the mapping reads all eight grades at once. Check in whatever order the brief makes easiest.

Decline the format, not the role. A substitute keeps you in the process when the brief does not.

Step-by-step: grading a take-home brief before you agree

The procedure takes about an hour of your time plus 1 to 3 days waiting for the employer's answers. It ends with a verdict and a sent message, not with a decision you are still turning over.

Before you spend that hour, check that the role justifies six hours of anyone's time. Refolk scores how well your own history fits the posting, so you know whether the role is worth the hours before you start negotiating them. The fit score does not change the grade. It tells you whether grading is worth doing at all.

Reporting the time you spent is the step most people skip, and it is the one that protects you. Reviewers spend 10 to 20 minutes on a submission, so they are comparing finished artifacts, not effort. Putting the hours in the submission itself is how you make the cap matter to the person grading.

Take-home assignment red flags: how this standard goes wrong

Every criterion can show a pass on paper while the real terms fail. Each failure mode below gives the criterion, what the false pass looks like, and the check that catches it. These are the cases where a careful grader and a hurried one disagree, so they are where this standard earns its keep.

1. Stated time cap. The brief says "2 hours" but the scope is a week of work, as in one Hacker News thread. Check: count the deliverables and set the count against the hours. Five distinct outputs in two hours is not a cap, it is a slogan.
2. Realistic-hours check. The brief says 1 to 2 hours, and environment setup alone took one candidate 5 hours, in a case reported on Blind involving a Snowflake environment. Check: list every tool, account, and dataset you need before you start, and add the setup to the doubled figure.
3. Stage placement. "After screening" turns out to mean an automated screen. Check: can you name a human you spoke to live, and the date? If not, criterion 3 fails.
4. Synthetic inputs. The data is labelled "anonymised" but is plainly their real Q3 numbers. Check: the Monday test from Career Vision. Could this output ship as it stands?
5. Work-product terms. The brief says nothing about rights, but the assessment platform's terms assign them. Check: read the platform terms before you create the account, not after.
6. Feedback commitment. The promise is "we'll be in touch", and the outcome is a generic rejection, a pattern reported on both Hacker News and Blind. Check: ask for a date in writing. No guide sets a minimum feedback standard, so a named reviewer and a date is the whole test.
7. Completion window. "A week" is offered, but it falls entirely on weekdays. Check: count the weekends inside the window, not the days.
8. Time cap vs grading. The cap is stated, but the grader penalises incomplete work. In one Datadog case, a submission done in 3 hours was rejected for "not enough tests". Check: ask whether capped work with notes on next steps scores fully. If the answer is no, criterion 4 fails, because the real deliverable is "finished", whatever the cap says.

Where the evidence behind this standard is thin

• The 2x multiplier is practitioner testimony, not a measured benchmark. No controlled study of it has been published.
• The unpaid caps are rules of thumb. The single median comes from a page that does not state its sample.
• The share of take-homes that are paid has not been published.
• The US federal position on take-homes has not been confirmed from a primary DOL source.

What to check locally: log your stated and actual hours on every take-home you do. After three, you have your own multiplier, and it beats mine.

How to push back on a take-home assignment

Push back with one written clause per failing criterion, each asking for a specific fix. If you decline, offer a substitute in the same message so the decline reads as a change of format and not a withdrawal. Sent before you start, a rescope request looks like professionalism. Sent halfway through, it looks like an excuse.

Hi [Name],

Thanks for sending the brief. I'd like to do it well, and I have a few questions so I can scope it properly.

[1 Time cap] Could you confirm how many hours you expect this to take?
[2 Hours] Doing this properly looks like closer to [expected hours] than [stated hours] once setup is included. Could we trim it to [named subset], or would you pay for the extra time at a stated rate?
[3 Stage] Before I start, could I have a short call with [hiring manager] to make sure I'm solving the right problem?
[4 Deliverable] If I stop at the time cap and include notes on what I'd do next, will that be scored as complete?
[5 Inputs] Could I use synthetic or sample data instead of [the live data or real problem]?
[6 Terms] Could you confirm in writing that I keep the rights to my submission, and let me know how long it's kept after review?
[7 Feedback] Who will review it, and by what date should I expect to hear back?
[8 Window] Could the deadline include a full weekend? [Proposed date] would work for me.

Happy to start as soon as these are settled.

[Your name]

Hi [Name],

Thank you for the brief. I'm still very interested in the role, but I can't take on [the specific term: unpaid work on live data / an open-ended scope / the rights assignment in the platform terms].

Could I offer an alternative that tests the same skills? I could:
- walk your team through a focused version of the problem live, in about [an hour], or
- present [specific past project] and take questions on the decisions I made, or
- do a shorter written piece plus a live session on the rest.

If none of these works, I'd welcome another way to show [the skill being assessed].

[Your name]

The substitutes with documented use

• A live walkthrough of a focused version. Career Vision's pushback script proposes this.
• A mixed asynchronous and live format. Bumble's iOS hiring, run through Tapioca, combines both.
• Existing portfolio work. Pointing the team at work you have already done is reported on Blind. The report is anecdotal, but the move is low-risk to offer.

The portfolio substitute only works if the comparable work is easy to find. Refolk writes your resume from your own history and tailors it to each posting, so the project you would offer to present is already named on the version the hiring team has.

Payment as the counter-ask

Asking to be paid for over-line work is cheap for the employer, which makes it a reasonable request. The published figures:

• Lemonly pays a flat fee scaled to seniority: $50 for an entry role, up to $200 for senior or principal-level work. Six finalists at those rates cost $300 to $1,200.
• WordPress (Automattic) paid $25 an hour for every role, according to an older post. Check whether that still holds before citing it.
• Peoplism pays final-round candidates for 10 hours of work.

Adjusting the standard by role and market

The eight criteria do not change by role or country. What changes is the hours line for high-volume roles, and possibly how your pushback reads in a different hiring market. Sales development is the clearest case. RecruitingDaily puts a fair SDR take-home at 30 minutes to 1 hour, so an SDR brief stated at 2 hours is already double the role-specific guidance before you apply the multiplier.

The two pools below come from Refolk's index of professional profiles. They are the SDR candidate pool and the specialist recruiting role that most often sends technical take-homes.

How I read this, with the caveat stated plainly:

• The legal footing differs. The UK line is about whether recruitment is genuine and whether a trial runs past a day. The California line is about productivity, training and duration. Criterion 5 holds in both. The phrasing of a wage argument does not carry over.
• The person reading your message may differ. In Refolk's index the dedicated "Technical Recruiter" title is far rarer in the UK than in the US, even allowing for the undercount. Pushback scripts are mostly written for US hiring teams, and how well they land in the UK has not been established. Ask who reviews the work (criterion 7) and write to that person.
• Many of the published caps assume software roles. Where a source gives a role-specific cap, as RecruitingDaily does for SDRs, use the lower number in place of the general line.

Before you reply: the grading checklist and how to keep it current

Run the checklist below before you send anything. If any item is unchecked, the grade is not finished. If you are grading as a team, two people should run it on the same brief and compare verdicts. A mismatch means one of you skipped the missing-equals-fail rule.

Keeping the standard current

Two inputs to this standard change over time. Re-check them at their sources rather than trusting a figure copied from somewhere else:

• The legal tests. Re-read section 46.7 of the DLSE Manual for California, and the unpaid work trials section of the UK government's minimum wage guidance. If the employer pays for over-line hours in the UK, check the current National Living Wage on GOV.UK rather than relying on a figure quoted in a blog post.
• Your own multiplier. Every take-home you accept gives you a stated figure and an actual figure. Once you have a few, replace the 2x with your own number and update the expected-hours line on your grading sheet.

The window figure is a check worth keeping in mind. One secondary source, citing recruiting-operations benchmarks, puts the median interview loop at 14.4 days for business roles and 17.9 for technical roles. A two-weekend window would take up most of that loop, which is why employers push for shorter ones. A single full weekend is the minimum this standard asks for, and it fits comfortably inside either median.