How to Conduct a Technical Interview as an Interviewer
Advanced4 minInnovialabs editorial team
In short
A good technical interview is structured: a clear role profile, the same questions and tasks for every candidate, a written scoring rubric, independent feedback based on evidence and a respectful candidate experience.
What a good interview looks like
A technical interview answers one question: can this person do this job in this team? Not "are they as smart as me" and not "do they know my favourite trick".
Good interviews share five traits:
- a clear profile of the role, written before the first candidate;
- structured questions and tasks, the same for everyone at the same level;
- a scoring rubric that describes what weak, solid and strong answers look like;
- evidence-based feedback, written independently before discussing with other interviewers;
- a respectful experience that leaves candidates with a fair impression, even when rejected.
Step 1. Define the profile
Agree with the hiring manager on what the person will actually do in the first months. Turn it into 3-6 competencies to evaluate, for example:
- coding and code quality;
- problem solving and debugging;
- system design (for mid-level and above);
- knowledge of the main stack;
- collaboration and communication.
For each competency, define the expected level. A junior is not expected to design a distributed system; a senior is expected to discuss trade-offs without prompting.
Assign each competency to a specific interview stage so nothing is checked twice and nothing is missed.
Step 2. Design questions and tasks
Prefer tasks close to real work. Debugging a small piece of code, extending an existing function, designing an API for a familiar feature or reviewing a pull request tell you more than puzzles.
Good task properties:
- Solvable in the time slot by a candidate at the target level, with room to go deeper.
- Layered: a basic version, then follow-ups that raise complexity.
- Clear input and expected output, so the candidate does not waste time guessing the requirements.
- Tested on colleagues before using it with candidates.
Prepare follow-up questions in advance: "How would you test this?", "What changes if the data grows significantly?", "Where could this fail in production?".
Step 3. Write a scoring rubric
Without a rubric, scores reflect mood and personal taste. Describe each level with observable behaviour:
| Score | Problem solving (example) |
|---|---|
| 1 - Below bar | Could not move forward even with hints; did not clarify requirements |
| 2 - Mixed | Reached a working solution with significant help; missed obvious edge cases |
| 3 - Meets bar | Clarified requirements, reached a working solution, handled main edge cases |
| 4 - Strong | Considered alternatives, explained trade-offs, tested the solution unprompted |
Score each competency separately, then give an overall recommendation.
Step 4. Reduce bias
- Ask everyone the same core questions for the same level.
- Take notes during the interview: what the candidate said and did, not your impressions.
- Write feedback before talking to other interviewers so opinions do not anchor each other.
- Separate signal from style: accent, confidence, the university name or similarity to you are not competencies.
- Avoid "culture fit" as a vague reason; if collaboration matters, define what you evaluate.
- Calibrate regularly: compare scores with other interviewers and review past decisions.
Step 5. Candidate experience
The candidate is evaluating your company too. Simple rules:
- Introduce yourself and the format at the start: how long, what stages, whether they may use documentation.
- Let them ask clarifying questions and treat questions as a positive signal.
- Give hints when stuck and note the hint in your feedback, instead of letting silence last for minutes.
- Leave time for their questions at the end and answer honestly.
- Do not hint at the outcome during the interview; the decision is made after all feedback is collected.
- Make sure they hear back within the time you promised.
Step 6. Write useful feedback
Feedback is for the hiring decision, so it must be specific:
Recommendation: Hire (mid-level backend)
Problem solving - 3: Clarified input format, built a working solution,
handled empty input after a prompt about edge cases.
Code quality - 4: Clear names, extracted helper functions, wrote two tests.
System design - 2: Proposed a single database without discussing load;
improved after the question about read-heavy traffic.
Concerns: Limited experience with caching; can be addressed in onboarding.
Avoid "smart", "not a fit" or "weak" without evidence.
Common mistakes
- Improvising questions for each candidate.
- Testing trivia that a developer would look up in seconds.
- Talking more than the candidate.
- Judging by the final answer only, ignoring the reasoning.
- Rushing to fill a vacancy and lowering the bar without saying so.