How to design a paid take-home for builder hires

Take-home, live build, or paid trial day

You hire a builder to build, so most published hiring advice for these roles includes a build task. Where it splits is the format: a take-home, a live build, or a paid trial day. Here’s how six guides and hiring processes size it.

Our GTM engineer interview questions compare the full Clay, Apollo, and Databar loops. Here’s how to pick the format:

• A take-home suits most builder finalists. It shows how they work alone, on their own schedule.
• A live build suits a candidate who can’t give up an evening, or when you want to hear how they think out loud.
• A paid trial day suits the last one or two candidates, when you also want to see how they work with your team. Expect to pay a full day’s rate.

Candidates notice when a take-home goes wrong. One Clay candidate reported more than 40 hours on an unpaid assignment, with scoring criteria they called ambiguous (Taro, May 2025). The same report warns that anyone who didn’t already use Clay would have to learn it first. A task like that tells a candidate how you’ll treat their time once they work for you.

Eight rules for a take-home strong builders will finish

• Write the rubric before the brief. If you can’t say what a top score looks like, the task is too vague. Ambiguous scoring was one of the complaints in the Clay report above.
• Pick one task close to the real job. Choose a problem they’d own in their first month. Skip puzzles and anything they’d never do at work.
• Use clean or made-up data. Strip names and emails from any export, or build a sample file. Don’t send live customer data to someone who doesn’t work for you yet.
• Set a hard cap of 2 to 3 hours. Tell them to stop at the cap and list what they’d do with more time. Yardstick tells hiring teams to keep take-homes to 2 to 3 hours at most, and GTME Pulse sizes its Clay build the same way.
• Let them pick the tools. Unless the job lives in one tool, accept Clay, n8n, Make, a script, or a spreadsheet. You’re testing how they think, and nobody should learn your stack for free.
• Pay a flat fee. The math is in the next section.
• Write down the AI rule. Companies differ here. Anthropic’s candidate guidance tells applicants to complete take-homes without Claude unless told otherwise. For builder roles, we’d allow AI and ask for the prompts, the way Databar asks candidates to document their prompts and tool choices. Building with AI is the job, so test it.
• Say who owns the work. Put it in the brief: you won’t use their build in production, and you’ll delete their files after you decide. PostHog tells candidates that work from its paid SuperDay isn’t published.

Send it to two or three finalists, never the whole pipeline. Same brief, same rubric, same deadline for each.

What to pay for a 3-hour task

Set one flat fee per role and pay it to every finalist who submits, whether or not they advance. A fair way to size it: the role’s U.S. base salary divided by 2,080 working hours, times the hours you’re asking for.

Databar suggests $200 to $300 for its GTM engineer take-home. When in doubt, round up. The fee is small next to a wrong hire. A U.S. GTM engineer at the $159,000 median costs about $18,900 a month with benefits: $159,000 ÷ 0.70 ÷ 12, because benefits were 30.0% of total compensation for private-industry workers in June 2026, per the Bureau of Labor Statistics. A $230 task is about 1% of one month.

Three briefs you can adapt

Each brief fits in 2 to 3 hours if you prepare the inputs first.

Want a full rubric to go with a brief? Our GTM engineer interview questions score an outbound build out of 25, and our marketing engineer interview questions weight a scorecard build by category. For the automation role, start with how to hire an AI automation specialist.

How to score it fairly

The rubric matters less than how you use it. Two finalists who send the same quality of work should get the same score, whoever reads it first.

• Two reviewers score alone. Each scores every submission before they compare notes. Talk only where the scores differ by more than a point.
• Score the work before the call. Then score the review call as its own line. A strong talker shouldn’t lift a weak build.
• Set the pass bar in advance. Write it next to the rubric when you send the brief.
• Score submissions in the same sitting. The first one you read shouldn’t become the benchmark for the rest.
• Hold the time cap. If someone went well past it, ask in the call what they built after the cap, and give that part less weight.

Each role also needs one line the others don’t:

Red flags in the submission, and in your own process

In the submission:

• A slide deck about a build, with no build to open
• They can’t explain a step when you ask why it’s there
• AI-written output went in unchecked, with wrong names or made-up facts
• A public template with your logo on it, and nothing changed underneath
• No list of what they’d do with more time, so you can’t tell what they chose to skip
• They solved a different problem than the brief asked

In your process:

• An unpaid task, or one that runs past 3 hours
• No rubric, or one you wrote after reading the submissions
• Live customer data, or a task that looks like work you’d ship
• A required tool the candidate has to learn on their own time
• Sending it to every applicant instead of two or three finalists
• No feedback after they submit

When to skip the take-home: if a candidate has shipped work you can open and run, a 60-minute live walkthrough of it tells you as much. And if you need one project built, hire a freelancer for that project at their rate. A take-home is for choosing a full-time hire who’ll build for you every month.

Where GrowthPair fits

We recruit, vet, and pay Marketing Engineers, who build GTM systems, marketing pages and tools, and AI automations inside your team. Every candidate goes through five steps: a resume screen, a phone interview, a proctored skills assessment, a live interview with a marketer, and your interview. 1.5% of applicants are accepted, and every one passes a C1 English test. You run the last step and choose who you hire. A full-time hire costs about $4,000 a month, all-in, covering the marketer’s pay, payroll, benefits, and HR.