Layer one: role and logistics
Start with the facts that determine whether the job can work: relevant responsibility, schedule overlap, communication requirements, compensation alignment, and the candidate’s ability to work reliably in the required environment.
This layer should be quick. Its purpose is to remove obvious mismatches without pretending it has proven job performance.
Layer two: written evidence
Ask candidates to explain a decision, diagnose a realistic problem, or show how they would organize the first steps of the work. Strong answers make assumptions visible and connect actions to an outcome.
Generic enthusiasm is not evidence. Neither is a polished résumé. The written step should resemble the thinking the role actually requires.
Layer three: a role-specific work sample
Give the candidate a bounded scenario with enough context to make a decision. Evaluate the reasoning, tradeoffs, communication, and finished output—not whether they guessed an internal trick.
Talented’s own aggregate data reinforces why this matters: among 729 candidates with both evaluations, résumé score and work-sample score were essentially unrelated. The correlation was 0.03.
Layer four: a decision interview
Use the final conversation to probe the evidence already collected. Ask why the candidate made a choice, what they would change with more context, and how they have handled similar constraints in real work.
By this point, the interview should decide among demonstrated options. It should not be the first serious attempt to determine whether the person can do the job.
What stays consistent at scale
- Every stage should test something the previous stage could not.
- The work sample should resemble the real role without becoming unpaid client work.
- Reviewers need a shared definition of strong evidence.
- Candidate identities and protected characteristics should never determine the score.
- The final shortlist should be small enough for a real decision.