Work sample
Also called: job sample, performance task
What separates a work sample from a generic test
A work sample uses the same kind of inputs and outputs as the real role:
- Sales role: a mock discovery call with a fake prospect persona.
- Designer role: a design exercise with the kind of brief and constraints they’d actually get on the job.
- Engineer role: a small feature in a representative codebase, not a leetcode-style algorithm.
- Customer support: a sample ticket with the company’s tone guide.
The test is whether someone reviewing the candidate’s submission could mistake it for output from a current team member.
Why work samples beat interviews
Forty years of research (Schmidt & Hunter and follow-ons) puts work samples at the top of predictive validity for job performance — substantially above unstructured interviews and tied with structured interviews.
The reason: a work sample is a behavior measurement, not a self-report. Candidates who say they can do the work and candidates who actually can do the work are two different populations.
How to use them well
Cap them at 3-4 hours. Pay for substantial samples. Pair with a discussion. Don’t substitute for an interview — work samples test skill, interviews test fit and motivation.
Where Join fits
Work samples are first-class artifacts on the candidate record in Join, with the scorecard for the sample reviewer attached. See the features page.
