You can assess some of a candidate's skills before an interview through work they share, test results and a task related to the role. This does not prove future performance or make every conversation unnecessary. The aim is to reach a decision with relevant evidence and explicit questions, not replace a personal impression with a score lacking context.
This guide proposes a workflow for recruiters and technical managers. The exercise and matrix are editorial examples, not actual Nort tests, customer research or a validated predictive method.
Start with the work to be done
Describe a concrete task: investigating an API failure, reviewing a change or explaining a technical decision. Separate what someone must know on entry from what you will teach after hiring. Not every tool your team uses needs to be a knockout requirement.
Work samples reproduce job tasks. Reference: OPM.
For our proposed exercise, document the problem, expected deliverable and what you will not assess. If a portfolio or previous work already provides relevant evidence, consider that before requesting another task. Avoid requiring a public portfolio: an inability to share confidential work is not evidence of inexperience.
Agree on conditions before starting
Give participants the same essential instructions and criteria. Our suggested brief includes:
- Scope: what to deliver and what will not be requested.
- Time: expected effort, deadline and adjustment options; no duration suits every role.
- Environment: permitted tools, documentation and AI use.
- Access: an accessible format, a question channel and how to request accommodations.
- Feedback: who will review the work and when a response is expected.
Use synthetic data and a separate environment. Do not request access to former employers' private repositories or use selection to obtain production work. Agree with the responsible team on data access, retention and handling; this guide does not certify legal compliance.
Example: an API returning duplicate items
Imagine a fictional API returning repeated items across pages. Supply a small code sample, synthetic data and examples of expected behavior. Ask for a possible cause, a proposed fix and a way to verify the result.
Do not hide an essential business rule and expect the candidate to guess it. If multiple solutions are plausible, document the constraints that allow comparison. The task reveals part of the work, not mastery of the entire role.
| Criterion | Evidence to record | Follow-up question |
|---|---|---|
| Diagnosis | A cause consistent with the supplied code and data | What observation would contradict your hypothesis? |
| Correction | A change connected to the problem and constraints | What alternative did you consider? |
| Verification | Test cases linked to expected behavior | What failure might still escape detection? |
| Communication | Clearly stated assumptions and limits | What information was missing? |
Use states such as observed, partial and not observed, with a reason and a reference to the submission. “Not observed” does not automatically mean “cannot do it”: it may reflect ambiguous instructions, access problems or a need for another question.
The matrix is neither an automatic ranking nor a validated scale. Before using it, ask reviewers to discuss a fictional response and identify differences in interpretation. If they subsequently change a criterion, revisit affected assessments.
What the conversation still needs to clarify
Structured interviews use common questions and criteria. Reference: OPM.
In this workflow, prepare questions about the choices, alternatives and limits observed in the task. Leave room for work expectations, responsibilities and the candidate's questions. An early conversation can help avoid requesting an exercise for a role that does not match those expectations.
Record evidence, interpretation and decision separately. When reviewers disagree, a human owner should revisit the submission and context, not simply select the highest score or defer to the most senior voice.
Personality and AI: support is not a decision
The relevance of personality traits depends on the job. Reference: OPM.
Do not treat a single trait or profile description as proof of technical competence. Before adopting an instrument, request documentation of its purpose, limitations and suitability for the intended use. This article does not establish the psychometric validity of Nort tests.
For AI, explain in advance whether it is allowed and how to describe its contribution. If permitted, ask candidates to explain what they accepted, corrected and verified. Polished code is not proof of cheating, and a generated answer alone is not sufficient evidence of understanding.
Where Nort fits
Nort is a reverse recruiting platform: candidates build profiles with information and test results, and companies search for professionals. Those records can support initial assessment, but do not guarantee hiring, freedom from bias or the elimination of interviews.
Explore Nort for employers and the reverse recruiting guide. When comparing solutions, check what each result measures, the conditions under which it was produced and which decisions still require your team's assessment.
Frequently asked questions
Can you assess candidates before interviewing them?
Yes, you can gather evidence about specific skills before a conversation. It may not cover every job requirement and does not guarantee future performance.
Does a test eliminate all bias?
No. Instructions, criteria and interpretation can introduce distortions. Review participation conditions and document the reasons for decisions.
Can personality be the only criterion?
That is not the approach proposed here. A profile does not replace job-related evidence or consideration of an instrument's limitations.
How should AI use be handled in the exercise?
Set the rules in advance. If use is allowed, ask candidates to explain their choices and verification, applying the same essential criteria to participants.
Does Nort guarantee hiring without interviews?
No. The platform supports search and access to candidate information; the decision and necessary stages depend on each hiring process.
Editorial note
Revised September 9, 2026. We removed unsupported promises about hours saved, universal test superiority and product capabilities from the previous version. The sources explain assessment methods; they do not certify Nort. Corrections: [email protected].