A role with four hundred applicants is not a harder version of a role with twenty. It is a different problem, and the failure mode is specific: the bar does not get lowered deliberately, it drifts downward through fatigue somewhere around candidate ninety.
Filter before the interview, not during it
Anything that is a genuine yes-or-no requirement — the right to work, a licence, a minimum of nights available — should be settled on the application form. Using interview time to discover that someone was never eligible is the most expensive way to find out, and it is unfair to them as well.
Be careful what you put on that list. A hard filter is a hard filter, and "five years of experience" excludes people who would have been excellent. Keep the list to things that are actually binary.
Keep the round genuinely short
At volume the arithmetic is brutal. Four questions at two minutes is eight minutes per candidate; four hundred candidates is fifty-three hours of recording. You will not watch all of it, and you should design the round knowing that. Put the most discriminating question second and let reviewers stop when the answer is clear.
Review by question, in batches
Watching question two across forty candidates in one sitting keeps your scale steady in a way that candidate-by-candidate review cannot. It is also faster, because you stop re-reading the question and re-establishing what good looks like.
Rotate the reviewers
Nobody should review a hundred answers in a day. Split the batch across the panel, overlap a handful of candidates between reviewers, and compare the overlap — that comparison is the cheapest calibration check there is, and it tells you when someone's bar has drifted.
The overlap is not wasted duplicate work. It is the only evidence you have that two reviewers are applying the same standard.
And keep the rejections quick
Volume makes slow rejections worse, not more forgivable. Four hundred people waiting three weeks is four hundred bad impressions in a market where you will be hiring again next quarter.
