Mark Grebler has interviewed more than a thousand people, and he now ends every interview the same way: he tells the candidate what he thought of their performance, out loud, before the debrief. The framing is not mainly charitable. Feedback makes the interview two-way, and it gives the candidate a chance to correct a misreading while it is still a misreading rather than a hiring decision.

How the move works:

  • The line is roughly: “I’d like to give you some feedback. First, so you can get some feedback on the interview, but also so you have the chance to correct anything that I have misinterpreted. Is that ok with you?” The “is that ok” is not seeking permission — it is a two-second reset from interview mode into feedback mode. Other interviewers give their feedback too.
  • The bad questions this replaces (“if you were an animal…”, “how many windows are in the city”) share one property: they are one-directional. The candidate performs, the interviewer judges, and the judgment never travels back.
  • Feedback must be about capability, not personality. “You seemed arrogant” gives a candidate nothing to act on but the feeling of being judged; “you dismissed my alternative answers as unnecessary” is the same concern made actionable.
  • Positive feedback is specific enough to carry: the architecture walkthrough explained trade-offs at each level; the team-building answer had real examples; you were nervous narrating your projects, which I don’t weight, but brush up on those before your next interview.
  • Constructive feedback is specific enough to sting: you described teams you have been on but not how you would get a team delivering, which makes me doubt the depth underneath; your background is more senior than this role, which raises a retention question — tell me what draws you here.
  • The response taxonomy:
    • Corrects the misunderstanding — roughly one interview in five. A testing-depth concern answered with ten years of specific TDD practice.
    • Tries to correct it and confirms it instead — the candidate goes deeper and gets muddier. He treats a confirmed concern as a flag against earlier stages, not a verdict.
    • Accepts it gratefully — the majority, even when the feedback was scathing. He doesn’t count the gratitude as validation, since he can’t tell real gratitude from a performance for the person holding the decision.
    • Responds negatively — very rare. Thank them and move on; arguing turns it into a bad experience for both sides.
  • He names two ways this could be self-flattering: voicing a doubt tells a shallow candidate exactly where to aim a clean answer (mitigated by probing for application, which rote answers survive poorly), and once you have said a concern out loud you are partly invested in it.
  • His own failure: he told a candidate his confidence might read as arrogance to the team, and the candidate turned hostile. His conclusion was that the delivery was wrong, not the decision — and he supplies the behavioural rewrite. Given “you can’t improvise experience you don’t have,” most corrections either land or expose the gap.
  • The cost is real: synthesising your impressions fast enough to deliver at the end of the interview takes practice, plus the nerve to say them.
  • The institutional objection, from the comments: some large companies forbid live feedback to standardise interviewer levels and avoid raising offer expectations. The alternative is “there was a lot of competition for the role, and no. kthnxbye.”

The reframe worth keeping is that live feedback is an information-gathering instrument before it is a kindness. Delivering the concern inside the interview means the candidate either corrects your model or confirms it, and you walk into the debrief knowing which.

It also separates the two things that get conflated in hiring feedback: what a candidate can work on, and what you privately concluded about them. Only the first one belongs in the room.