Why should a candidate prepare a per-round question bank to ask at the end, not one generic list?
answer
- Who in this room can answer it?
- Different rounds, different authority
- One card per interviewer, not one list
- Rank by what changes your decision
- Four prepared, two asked
basics
~20 sA per-round bank matches each question to whoever can actually answer it: a recruiter knows process and timeline, an engineer knows daily practice, a manager knows scope. One generic list burns a five-minute window on the wrong person.
solid answer
~40 sEvery round ends with a short window for the candidate's own questions, and different people at the table can answer different things. A recruiter can tell you how the loop is sequenced and how the role is levelled; the engineer you would sit beside can tell you how work actually gets reviewed and shipped; the manager can tell you what the team owns. So I keep one card per round with about four questions on it, ordered by how much the answer would change my decision, and I expect to ask two — five minutes does not hold four real answers. A generic list forces me to improvise the routing live, and the questions I most want answered end up aimed at someone who has to say `I'm not the right person for that`.
go deeper
Walk in with something written down for every round. Know that a recruiter and an engineer answer different things, and that having four prepared questions per round is normal and reads as prepared, not stiff.
Be ready to explain the sizing: four prepared, two asked, because a five-minute window only holds two real answers plus the follow-up. Show that you order the card by decision value rather than by topic.
Show that you route deliberately — that you can say which question belongs to which interviewer and why, and that you built the card from what you still do not know rather than from a template you found.
Own the tradeoff between coverage and depth across a whole loop: a broad card gathers a little from everyone, a narrow one drills the two people whose answers decide it. Say which you chose and what you accepted not learning.
## What a per-round question bank is A question bank, in this context, is the candidate's own prepared list of questions to ask at the end of an interview round — not the interviewer's questions to the candidate. "Per-round" means it is sliced by *who is in the room*, not kept as one long list carried through the day. The concrete artefact is small on purpose: **one card per round, about four questions on it, ordered top to bottom by how much the answer would change your decision, with a line struck through each one as the day answers it.** It fits on an index card or one note on a phone. It is not a script; it is a routing table plus a priority order. ## Why routing matters more than content The window is short — commonly around five minutes at the end of a round — and it is the only part of the interview the candidate controls. Two things spend it badly: asking something the person cannot answer, and asking something you no longer need answered. Different interviewers hold different authority: | Round | Can genuinely answer | Cannot really answer | |---|---|---| | Recruiter screen | How the loop is sequenced, who you will meet, how the role is levelled, timeline | What the release process feels like day to day | | Peer engineer | How changes get reviewed, what the on-call reality is, what the codebase feels like to work in | How headcount was approved | | Hiring manager | What the team owns, how work is prioritised, how performance is assessed | Line-level detail of another team's roadmap | | Skip-level or director | How the group is organised and where it is heading | Which ticket you would pick up first | A generic list ignores that table. The predictable result is a recruiter politely deflecting a question about code review depth, and the candidate leaving with nothing they could not have read publicly. ## Sizing the card Four prepared, two asked, per round. Four is enough that two being covered mid-conversation still leaves you two live ones; it is few enough that you can rank them honestly. Two asked is what five minutes actually holds if you want the *answer* to have room — a good answer to a good question runs a couple of minutes on its own, and the follow-up you ask off the back of it is often where the real information is. Across a four-round day that is roughly eight asked questions from sixteen prepared. That asymmetry is the design, not waste: the unasked ones are the reserve that keeps you from improvising. ## A worked example: a mobile role Suppose you are interviewing for a mobile app team, and your first round is a call with a recruiter. Your card for that round might read: 1. How is the loop sequenced after this, and who would I meet in each round? 2. How was this role levelled, and what did the team want that they do not have today? 3. Is the team building new surfaces in the app, or mostly maintaining what ships now? 4. How long has the role been open? All four are things a recruiter can genuinely answer. None of them asks the recruiter to characterise the release train or the crash triage rotation — those go on the peer engineer's card, where they will get a real answer. ## The failure this prevents The common failure is not asking a bad question; it is arriving at the last five minutes with nothing live. The prepared question got answered in passing at minute eleven, there was no second tier, and the candidate improvises something vague to fill the silence. A ranked card with a reserve makes that structurally hard: you strike the answered one and read the next line. ## What the card is not It is not a memorised script read aloud in order — you ask two of four, and which two depends on what the round already surfaced. It is not hidden either: taking out a small card and saying you wrote a few down reads as prepared, not as stiff. And it is not a place to store questions you do not care about; a question you would not act on the answer to is taking a slot from one you would.
- How many questions belong on the card for a single round?About four, with the expectation of asking two. Four gives you a reserve when the conversation answers one or two in passing, and it is still few enough to rank honestly. More than that and nothing is prioritised; fewer and a single covered question leaves you improvising.
- What belongs on a recruiter's card that would be wasted on a peer engineer?Anything about the shape of the process rather than the work: how the remaining rounds are sequenced, who is on the panel, how the role was levelled, how long it has been open, what the timeline looks like. A peer engineer usually cannot answer those and will say so, which costs you a slot.
- What if there is only time to prepare one card before the loop?Prepare the round whose answers would most change your decision — usually the hiring manager or the peer engineer, because they describe the work you would actually do. For the other rounds, carry two general-purpose questions and rely on what earlier rounds surface to sharpen them.
saying these in an interview costs you the question
- Carrying one generic list and asking a recruiter about code review depth
- Preparing a dozen questions with no priority order
- Improvising in the last five minutes because the prepared question was already answered
- Storing questions the candidate would not act on the answer to
- Treating the question window as a formality rather than a decision tool