How we vet a specialist before they reach the bench

Nobody reaches our bench on a CV. Every specialist does a paid task in the role they applied for, and a reviewer who does that job for a living scores it against a written rubric.

The paid task

The task is real work, scoped to roughly three hours, and we pay for it whether or not the person passes. A Shopify developer gets a theme task on a seeded store and we review it as a pull request. A bookkeeper gets a trial balance with two planted errors. A support specialist answers twenty real tickets cold.

We score process as well as output. Someone who gets the wrong answer but asks the right clarifying question scores better than someone who guesses and happens to be right.

English, out loud

Written English is tested on a real thread, not a grammar quiz, because the job is replying to people. Spoken English is tested on a call where the interviewer pushes back and changes their mind halfway through. We report the level we scored rather than a label, so you can judge it against the job instead of trusting a badge.

What makes us say no

Three things, in order of how often they come up. Work that looks copied, which we check. An unwillingness to say "I do not know" during the call. And a task submitted late without telling us it would be late, because that is the same habit that costs you a deadline later.

Skills we can usually teach. The last two we cannot.