You pass the recruiter screen and the technical rounds, and instead of a final panel the company asks you to come work with the team for a day, or a few days. You build something real, sit in their standup, and at the end everyone involved votes on whether to make an offer. That is a trial day, also called a work trial or a working interview, and more companies are now publishing how they run it rather than treating it as an experiment.
The idea comes straight from the complaint every engineer has about whiteboard loops: a few hours of puzzles predict little about whether someone is good to work with over a quarter. A trial swaps the proxy for the thing itself. You do the job for a bit, the team sees your real output and how you collaborate, and you get a close look at the codebase and the people before committing.
Who runs them and what they look like
The clearest published example is Linear, the issue-tracking company. In a December 2023 post, Linear says every hire, from engineers to C-level, goes through a paid work trial of two to five days as the final step, with senior roles usually taking the full five. The trial is remote, you get access to their actual tools (Slack, Notion, GitHub, Figma, and Linear itself), and you work on “a real project that we plan to implement (or as close as possible to that).” An engineer might build a feature to triage incoming requests; a designer might design a way to resolve comments. Linear lists no manager or C-level prompt, but it confirms those roles go through a trial too, so for a leadership loop expect a planning or strategy exercise rather than shipped code. There is a kickoff, a mid-trial check-in you drive, and a final presentation to the trial panel. The same post says candidates are told the daily rate in advance. As of late 2023 it reported the format had helped Linear grow to over 50 people with a 96% retention rate, and it is blunt about the bar: “anything other than a strong yes is a no.”
The length and shape vary a lot. Some run a single paid day, some stretch to a full week, and a few fold the trial into a weekend to spare employed candidates the time off; at least one AI-first law firm, Crosby, has made Sunday interviews-as-trial-days its standard so candidates do not have to burn a weekday, as covered by Entrepreneur in 2026. The common thread is that you produce real work with the real team, not a contrived prompt.
These rounds cluster at early-stage startups, where a single bad hire is expensive and the team would rather spend money on a few paid trials than carry a mismatch for months. The search company Exa runs an onsite version, flying finalists out to build with the team; see this site’s writeup of Exa’s working trial. Larger companies mostly stick to the traditional onsite final round, so if you are deep in a startup process and the “final” turns into a work period, that is normal for the stage, not a sign something is off.
| Trial format | Typical pay expectation | Where it sits in the interview loop | What it is actually testing | Best suited to |
|---|---|---|---|---|
| Half-day scoped exercise (a few hours) | Often unpaid when it is really a take-home done live; some pay a flat fee | A technical round, not a true trial | Can you do a bounded task cleanly under light observation | Any candidate; low time cost |
| Single paid trial day | Paid: a day rate agreed in advance | Replaces or caps the final round | Collaboration, ramp speed, how you ask questions | Candidates who can take one day off |
| Multi-day work trial (2 to 5 days) | Paid: a daily rate agreed in advance for the whole period | The last step before an offer | Real output on a real project, plus team fit over time | Between-jobs candidates, or weekend-flexible ones |
| Full paid trial week | Paid: a week at a contractor rate | Stands in for the onsite entirely | Sustained performance and mutual fit | Candidates with a notice period or time between roles |
Paid versus unpaid, and the line that matters
At a company worth working for, real work on a trial is paid, for practical and legal reasons. Under the US Fair Labor Standards Act, an employer that “suffers or permits” someone to work generally owes at least the federal minimum wage of $7.25 an hour; a fact-check by WCNC applied that logic to unpaid trial shifts covering a real work period. State rules can be stricter: a 2026 law-firm note on restaurant “stages” points out that California’s labor enforcers have repeatedly treated unpaid trial shifts as a violation when the work replaces or supplements paid staff. None of this is legal advice, and the details turn on whether you are producing real value and who benefits, but the direction is clear enough that a serious tech employer pays.
On paper a paid trial is usually a short contractor agreement; you invoice for the days worked. A good company names the day rate up front, the way Linear does. If no number has been stated, ask the recruiter directly: “what’s the daily rate for the trial?” Treat the role’s annual base divided by about 260 working days as a floor, not a target: a contractor day rate should sit above it, since you get no benefits, no paid time off, and you cover the employer-side tax yourself. If a company wants three or five days of your real work and offers nothing, that is not a trial, it is free labor, and it tells you how they will treat you on payroll.
What actually gets evaluated
Less than you would fear, and different from what a coding screen measures. The panel is watching how you operate, not whether you ship a finished feature by Friday: whether you ask good questions early instead of silently guessing, scope a problem down to something shippable, communicate progress without being chased, take a code review comment well, and hold your own in a standup with people you met yesterday. The work artifact matters, but no one expects you to learn their whole codebase in a day, so trajectory and collaboration outweigh raw output. Because the round scores demonstrated competence over a tidy résumé, it can work in your favor if you are between jobs or returning after a gap: what you show in the room outweighs what the CV is missing.
This cuts both ways, which is the real advantage of the format. A trial is the best look you will get at a team before joining. Treat it like an extended reverse interview: watch how decisions get made, whether people are kind under deadline pressure, how messy the codebase really is, and whether the manager you would report to is someone you want to learn from.
How to prep for one
Prep is lighter than for a gauntlet of algorithm rounds, because you cannot cram a company’s codebase. Spend the time on setup instead: a working dev environment and your tools ready, so you are not fighting your laptop on day one. Read whatever they send in advance and skim their public docs and repos. Come in with a plan for the first hour: get oriented, get something running, and surface your questions early.
During the trial, bias toward communication. Post your plan before you build, give a short progress update partway through, and if you are blocked, say so within minutes, not hours, because how you handle being stuck is a large part of what they are reading. Scope aggressively toward something you can show by the end; a small, finished, well-explained slice beats an ambitious half-thing. If there is a final presentation, like Linear’s, be ready to walk through what you did, what you would do next, and what you learned about their system. Keep your own examples tight using a STAR structure for the inevitable conversations about past work, and split your remaining prep time with a study plan if the trial sits on top of earlier technical rounds.
Logistics when you already have a job
This is where trials get genuinely hard. A two-to-five-day trial is awkward to fit around a full-time job, and the company’s flexibility tells you a lot. Good companies offer evenings, weekends, or a compressed schedule; Linear says it sometimes uses weekends to minimize the time a candidate takes off. If a company refuses to flex and expects a week of unpaid leave on spec, weigh that heavily.
Two constraints people miss until they bite. First, read your current employment agreement before you write a line of code for anyone else: many contracts carry invention-assignment or moonlighting clauses that could let your employer claim work you produce for a prospective one while still on their payroll. Second, if you are on a work visa, a paid trial is a second source of employment income, and most visa categories do not let you accept pay from a second employer without separate authorization. Confirm with immigration counsel first; an unpaid shadow day may be the only safe option, and that is worth raising with the recruiter directly.
The red flags
Most trials are run in good faith, but a few patterns should end the conversation. Watch for these:
- Real multi-day work with no pay offered, especially framed as “proving you want it.”
- Unpaid work that ships. A real project is fair game when it is paid, but a brief that is both unpaid and headed straight to production, like building their live reporting dashboard instead of working a mock dataset, is free labor they need now.
- No defined scope, schedule, or end point, so the “trial” could stretch indefinitely.
- Pressure to start immediately with no written agreement about pay or deliverables.
- Vagueness about who is evaluating you and what a pass looks like.
A legitimate trial is specific: here is the project, here are the days, here is the rate, here is who you will work with and how we decide. If a company cannot give you those four things in writing, the problem is not your performance.
When the trial ends, send the recruiter a short note the same day: thank the people you worked with by name, restate what you built, and say you want the role. You just spent real days inside the team, so you can decide with more information than any onsite gives you. Before you counter an offer, run the package through a total comp calculator and anchor your salary negotiation on your level and comparable offers, not on relief at getting in.
Practice the behavioral round:
