Interviewing an experienced hire is largely about verifying a track record. Interviewing an intern is not: there is no track record, so an unstructured conversation ends up measuring confidence, polish and similarity to the interviewer — none of which predict much, and all of which correlate with background rather than capability.
This guide covers how to design rounds that actually discriminate, how to score without the panel anchoring on one voice, and how to reject people in a way that does not cost you next year's applications.
Design the rounds first
| Round | Length | What it decides | Common mistake |
|---|---|---|---|
| Screening call | 20 min | Interest, availability, basic fit. | Turning it into a mini technical interview. |
| Work-sample task | 2–3 hrs, timeboxed | Can they actually do the work. | Making it multi-day, or never reviewing it. |
| Craft interview | 45 min | How they think about real problems in this role. | Quizzing definitions instead of reasoning. |
| Ways of working | 40 min | How they will behave in week six. | Asking "where do you see yourself in five years". |
| Closing conversation | 20 min | Selling honestly; surfacing what would make them decline. | Skipping it, then wondering why offers are declined. |
What actually predicts performance at this level
- Work-sample task
- The strongest single predictor available for early-career hiring, because it measures the work rather than the candidate's access to opportunity. Keep it short, realistic and timeboxed — and review every submission.
- Structured interview
- Same questions, same order, anchored scoring. Substantially more predictive than an unstructured conversation, and the gap is widest exactly where candidates have no history to discuss.
- Aptitude testing
- Moderately predictive and scales well. Useful as a defensible first cut at campus volume; weak as a sole basis for a decision.
- Unstructured interview
- Measures confidence and similarity to the interviewer. Feels highly informative, which is precisely the problem.
- College tier
- Largely a proxy for circumstances at seventeen. It shrinks a pool without improving it.
- Brain teasers
- Predict nothing job-relevant and disproportionately lose strong candidates who are nervous.
Questions that work when there is no work history
- "Walk me through one project — what was it, and which part was actually yours?" Probe: "What would you do differently now?" You are testing whether they can separate their contribution from the team's without inflating it.
- "Tell me about a time you got stuck. How long did you stay stuck, and what got you unstuck?" You are testing escalation instinct, which is the single most useful trait in an intern.
- "Take the hardest problem you have solved and walk me through it." Probe: "Which part would you struggle to do again without help?" You are testing honesty about their own boundary of competence.
- "Here is something you have not seen — read it and tell me what it does." You are testing whether they read before judging, which is most of junior work.
- "What would change your mind about that?" Asked after any confident claim. You are testing coachability directly rather than inferring it.
Score before you talk
Every interviewer submits scores and a recommendation before the debrief begins. If the panel discusses first, the most senior or most confident voice sets the anchor and the rest converge — you end up with one opinion held by four people. Read the independent scores aloud, then spend the debrief on the disagreements. Where two interviewers differ by two points or more, ask each for the evidence rather than the conclusion; most divergence turns out to be two people assessing different things.
What not to ask
- Marital status, family plans, pregnancy or childcare arrangements
- Caste, religion, region or mother tongue
- Age or date of birth beyond confirming legal eligibility to work
- Health conditions or disability, except a genuine and specific job requirement
- Current or expected stipend as a filter — anchor on your own published band instead
- "Why should we hire you?" — it rewards rehearsal, not capability
- "Where do you see yourself in five years?" — no intern can answer this honestly and you would not act on the answer
Rejecting well is part of the process
Most of the people you interview will be rejected, and campus word of mouth travels through student groups faster than any employer brand campaign. Reply to everyone who reached a screening call within a week. Give one line of specific feedback to anyone who reached a final round — it takes a minute and it is remembered for years. Never say you will keep a CV on file unless you actually will. The cost of doing this well is about an hour per cycle; the cost of doing it badly turns up as a smaller applicant pool next year that nobody connects to the cause.
Frequently asked
How many interview rounds should an internship have?+
Two to three, plus a short closing conversation. A screening call, one round on the craft and a ways-of-working round is enough to decide a three-month internship. More rounds lose candidates with faster offers without improving the decision.
How do you interview someone with no work experience?+
Ask about the work they have done — projects, coursework, anything real — and probe for the part that was actually theirs. Pair it with a short work-sample task, which is the strongest predictor available at this level because it measures the work rather than their access to opportunity.
Should we give a take-home assignment to interns?+
A short, timeboxed one — two to three hours on a realistic problem — and only if someone will actually read it and give feedback. Multi-day unpaid assignments select for free time rather than ability, and the strongest candidates decline them.
How do we stop the panel agreeing with the most senior person?+
Independent written scores before the debrief, every time. Group anchoring is strong and fast; the only reliable fix is to capture each reading before anyone speaks, then use the debrief on the disagreements.
Should we give feedback to rejected candidates?+
One specific line to anyone who reached a final round, and a prompt reply to everyone who reached a screening call. It costs about an hour a cycle. Silence is what candidates discuss publicly, and it shows up as a smaller applicant pool next year.