Everyone quotes p95 in latency SLOs, but few can compute it without a library and fewer can explain why the index is ceil(p*n)-1, not p*n. The warm-up that screens whether you actually understand the metric you live by.
Updated Aug 2026 · Grounded in real Forward Deployed Engineer interview loops and written to a senior-engineer editorial bar.
Everyone quotes p95 in latency SLOs, but few can compute it without a library and fewer can explain why the index is ceil(p*n)-1, not p*n. The warm-up that screens whether you actually understand the metric you live by.
The trap is averaging. A candidate who reports mean latency when asked for p95 has missed the entire point of the question, which is that tail latency is what users feel. The second trap is off-by-one in the index: interpolation method matters, and a strong candidate names which one they picked (nearest-rank) rather than pretending there is one true p95.
No comments yet — be the first to share your approach.
