When the Take-Home Is a Free Consulting Project, Walk

A take-home assignment in an interview is a fair test when it's short, the same for everyone at your stage, and scored against a rubric. It stops being a test the moment it asks for a deliverable the company could ship, scoped to "as much effort as you want." That's not screening. That's unpaid consulting with an interview wrapper, and you should price it accordingly.

So here's the part nobody tells you: the assignment that "shouldn't take more than a few hours" is almost always a lie the company is telling itself, not you. Scope creep in take-homes isn't a mistake. It's the default physics of an open-ended request. The real question is never "am I good enough to do this?" It's "has this company done the work to make this test actually predict anything?" Most haven't. And when they haven't, your free Sunday is buying them nothing but your free Sunday.

Are take-home tests just a scam to get free work?

No. And this is where most takedowns get lazy. Work sample tests are among the better predictors of job performance we have. A 2005 meta-analysis by Roth, Bobko, and McFarland in Personnel Psychology, across 54 studies and over 10,000 people, put their validity at around r = .33 after correcting for measurement error, and that was a downward revision. The headline numbers everyone quoted for years came from Schmidt and Hunter's 1998 meta-analysis, and a later reanalysis found those figures ran high once the methods were re-examined.

A 2022 reanalysis by Sackett and colleagues pushed the corrections further: combining the best hiring tools tops out around .47, with every individual coefficient lower than the 1998 figures, and structured interviews moving to the top of the pile. So the method is real. A well-built work sample beats a gut-feel interview. The problem is almost never the concept. It's the execution and the incentives behind it.

What separates a real work sample from spec work?

A real work sample and a free consulting project look identical from the outside. They feel completely different once you start. The difference is design discipline, and it shows up in five tells.

Tell	Real work sample	Spec work in disguise
Time	Boxed under ~3 hours, stated up front	"No hard deadline, take the time you need"
Specificity	Role-representative, fictional or generic data	Maps to a current, named business problem
Consistency	Identical prompt for everyone at this stage	Tailored to you, or "surprise us"
Scoring	Rubric named ("we evaluate reasoning, not polish")	No criteria shared
Placement	After at least one real conversation	First or second step, before anyone's talked to you

The rule of thumb: if the output is something the company could use without modification, or it scales in scope the harder you try, you're not being tested. You're working. Count the tells. Two or more, and treat it as spec work.

What does a weak take-home actually look like versus a strong one?

Weak. "Build a three-screen prototype for our onboarding flow and propose a content strategy to go with it. There's no hard deadline, take the time you need." Read it again. It's a company-specific deliverable. The scope is unbounded. There's no rubric. Two tells in the first sentence, and the "take the time you need" line guarantees the strongest applicants will pour in 14 hours to stand out. That onboarding flow is a thing they need built. You'd be building it.

Strong. "Here's a dataset with churn signals from a fictional SaaS company. In 90 minutes, find the top three intervention points and draft a retention email for the highest-risk segment. We're looking at clarity of reasoning, not polish." Time-boxed. Not their actual product. The criteria are stated. Nothing you produce ships anywhere. That's an audition. You can do it on a weeknight and learn something about how the team thinks while you're at it.

Same skill being measured. One respects your time as a finite thing. The other treats it as free inventory.

Who actually pays the price for long unpaid assignments?

The people with the least slack in their lives. A take-home that eats a weekend is a quiet filter for who has a free weekend. No second job. No kids. No parent to care for. The format doesn't mean to discriminate. It does anyway, because time is not evenly distributed, and like "culture fit," it's a place where bias hides in plain sight inside a process that looks neutral on paper.

The research is starting to catch this. One 2025 experimental study found that family caregivers were 4.4 percentage points less likely to be hired than equally qualified non-caregivers, an early-funnel barrier that's barely been studied. Long assignments sit right on top of that fault line.

And here's the twist that should worry the companies more than the candidates: the strongest people walk first. At Wolt Market Israel, the CEO said publicly that 25% of candidates for management and junior roles drop out at the assignment stage. Dropbox saw the same shape before they fixed their process: 20% of candidates never completed the take-home, with only about a 10% pass-through, on a roughly 15-hour ask per candidate. The people who can afford to vanish are the ones with competing offers. A bloated take-home doesn't filter for skill. It filters for desperation, and leaves you hiring from the bottom of your own funnel.

What does a company that's serious actually do?

It pays, or it tightens. Automattic, the company behind WordPress.com, pays trial candidates $25 an hour for 25 to 40 hours of standardized work that's never put into production. That's the gold standard, and it's honest about what it is: a paid trial, not a screen. The trade-off is obvious. Paying $25 an hour for 30-hour trials across 50 candidates for one seat is real money a seed-stage startup may not have. Fair. But there's a cheaper way to be serious that costs nothing: box the test to two hours, give everyone the same prompt, and tell candidates the rubric.

The tell is whether the company has bothered to make the test predictive at all. Standardized, timed, rubric-scored, and placed after a human conversation. If it skips all four, the assignment has roughly zero incremental validity over just reading your portfolio. They're not learning anything they couldn't get for free. You're the only one spending.

So when do you actually walk?

Not on reflex. The "always walk" advice mostly helps people who already have another offer in their pocket. If you're a career changer with no public portfolio, or you're in a thin market where this is the only door, a long take-home might genuinely be your best shot at proving you can do the work. Agency means reading your own situation honestly, not following a rule.

Here's the decision, in order:

Count the tells. Zero or one: probably fine, proceed. Two or more: it's spec work. Keep going.
Check your bargaining power. Competing offers or a strong portfolio? You can afford to push or walk. Thin market, no artifacts to show? The math changes, and that's allowed.
Ask one question before you start. It costs you nothing and tells you everything.

"Happy to do this. Two quick things: roughly how many hours do you expect this to take, and what are you evaluating it against? Want to make sure I'm putting effort in the right places."

A serious company answers in a sentence, because they already know. A spec-work shop gets vague, says "however much you want to put in," or goes quiet. That's your data. If they dodge and you have a strong hand, you can ask to be paid for a scoped version, or you can walk and lose nothing of value.

Name the trade-off

Walking costs you a shot at a specific job, and sometimes that job was real and the messy take-home was just one disorganized team's bad process, not malice. You'll occasionally walk away from something that would've worked out. That's the genuine cost, and pretending it's zero would be dishonest.

But the other side is realer. Your time and your work are career capital. Giving 15 hours of finished thinking to a company that hasn't decided whether it's serious isn't humility. It's a subsidy you're paying to a process that isn't paying you back, on the bet that effort alone gets noticed. It usually doesn't. The companies worth working for tend to be the ones who respected your hours before you'd given them a single one.

What to do now

Next assignment that lands in your inbox, do this before you open your laptop. Read it against the five tells and count. Send the two-question reply. Read the answer for whether they actually know what they're testing. Then decide with your real situation in front of you, not a rule and not your nerves. If you've got a strong hand and it's clearly spec work, price your time: ask to be paid, or price in the walk. If the market's thin and this is the door, go in with eyes open and time-box yourself even when they won't.

Want a second read on whether a take-home is a fair test or a free project before you sink your weekend into it? Send me the brief on WhatsApp and I'll help you count the tells and draft the reply.