Candidates prepare for interviews as examinations. Firms run them as simulations. The distance between those two understandings decides most outcomes before the last round.
Thinking without a script
The purpose of a hard question is rarely the answer. It is to watch what you do when the prepared material runs out: whether you reason aloud, state assumptions, and move — or stall, guess, and defend. Firms are staffing for the moments when there is no script, because those are the moments that cost money.
Risk asymmetry
Interviewers listen for how you weigh being wrong. A candidate who treats every claim as equally confident is signaling that they have never been expensively wrong. The strongest answers carry their own error bars: what you believe, how strongly, and what would change it.
Simplifying a mess
Somewhere in the process you will be handed a mess — a vague case, a broken dataset, a contradictory brief. The test is whether you can find the two or three things that matter and say them plainly. Complexity is easy to perform; simplification has to be earned, and it is the scarcest skill in the building.
Intellectual honesty
"I don't know" is not a failure state; unfounded fluency is. Firms have long memories of the damage done by people who could not say those three words. Saying them early, then reasoning toward an answer, is often the single strongest signal a candidate sends.
Who are you at 2 a.m.?
Ultimately the panel is asking one question: when it is late, the position is against you, and nobody is watching — what does this person do? Every filter above is a proxy for that. The candidates who pass are the ones whose answers, under pressure, sound the same as their answers in comfort.