How do you turn an interviewer's vague answer to your question into a concrete one?
answer
- Trade the policy for an episode
- Ask when it last happened
- Ask who actually made the call
- One follow-up per hedge, then move
- Curious register, never cross-examination
basics
~20 sAsk for the most recent instance instead of the general policy. A follow-up that requests one episode — what happened, when, and who decided — converts an abstraction into something checkable, and one follow-up per hedge is the right budget.
solid answer
~40 sSwap the policy question for an episode question. Instead of asking again how something works, ask when it last happened: `What is the most recent example of that?`, `Who made that call?`, `How did that one go?`. An abstraction can be produced from nothing; an episode has to be remembered, so it either arrives with detail or visibly does not exist. Keep the register curious rather than prosecutorial — you are interested in the work, not building a case — and spend one follow-up per hedge, then move on. Two things count as a real result: a concrete story, or a clean admission that the person does not know. Both go in your notes. What you must not do is accept the abstraction and move to the next question on your list.
go deeper
Learn one reusable sentence you can say out loud without rehearsing: ask for the most recent example. Practise it until asking a follow-up feels ordinary rather than confrontational.
Explain why an episode is harder to fabricate than a policy, and be able to name the three probes that produce one: when it last happened, who decided, how it went. Show that you budget one follow-up rather than an interrogation.
Show you can read the room while probing — choosing which single hedge in a round is worth the follow-up, keeping the register collaborative, and stopping cleanly when the honest answer is that the person is not close enough to the work.
Own the tension between gathering decision-grade information and being evaluated at the same time. Decide beforehand which one or two subjects would actually change your answer, and accept that everything else stays at policy level.
## The move: policy out, episode in A vague answer is almost always a *policy-level* answer — how things are meant to work in general. The single most reliable repair is to ask for an *episode* instead: one real occurrence, recently, with an actor and an outcome. This works because the two require different things of the speaker. A policy answer can be assembled on the spot from what sounds right. An episode has to be recalled, and recall leaks detail — who was in the room, what was traded away, how long it dragged on. Three phrasings do most of the work: - **When did that last happen?** — forces a date-shaped answer and rules out aspiration. - **Who made that call?** — forces a role, which tells you where decisions actually sit. - **How did that one go?** — invites the cost, which is the part recruiting language removes. ## Worked example A candidate interviewing for a data platform role asks the skip-level leader how the team handles it when an ingestion job fails overnight. > *We take reliability really seriously. There's a strong ownership culture — people don't throw things over the wall here.* Policy, values, no episode. The follow-up: > *When was the last overnight failure, and what did the person on call actually do that morning?* A substantive answer now has to contain a shape: the job that failed, whether anyone was paged or it was found at standup, who fixed it, whether the fix was reverted, whether anything was written down afterwards. A leader who owns the team and cannot produce that shape has told you something real either about the team's memory or about their own distance from it. ## Register matters as much as wording The same follow-up can land as interest or as cross-examination. Three things keep it on the right side. First, **attribute the gap to yourself, not to them** — you are trying to picture it, so a concrete example would help. Second, **ask one, not three**; a rapid second and third probe on the same soft answer changes the room. Third, **accept the answer you get**. If the reply is an honest 'I'm not close enough to that to give you a good example', that is a real answer, and pressing further buys nothing except a worse impression. Budget one follow-up per hedge. It is enough to distinguish a thin answer from a thin subject, it is affordable inside the few minutes a candidate usually has at the end of a round, and it keeps the exchange feeling like a conversation between engineers rather than a deposition. ## Follow-ups that do not work - **Repeating the question in different words.** The same altitude produces the same abstraction. - **Asking a leading question** — a probe that supplies the answer you fear invites a denial and tells you nothing, because it is easier to say no than to describe reality. - **Asking for numbers you cannot interpret.** A count with no context is another abstraction wearing a different costume; an episode you can picture beats a figure you cannot check. - **Stacking a compound probe.** Two questions in one sentence lets the person answer the easier half, which is what the hedge was already doing. ## Recording what came back Write the follow-up and its answer into the notes column for that interviewer while it is fresh, and record the raw material rather than your verdict: the words you got, and whether an instance appeared. That distinction matters when you later compare the same question across the loop, because 'gave no example even when asked directly' is comparable across people, while 'felt cagey' is not. ## The failure mode this defends against Candidates rarely fail here because their follow-up was badly phrased. They fail because they never asked it: the answer felt incomplete, the silence felt awkward, the prepared list was sitting there, and the moment passed. Treat an unfinished answer as the live thing in the room and the next prepared question as expendable. The prepared question you skipped costs you nothing; the hedge you accepted is the one you will be guessing about after you sign.
- Your follow-up gets a second abstraction. Do you press a third time?No. Two abstractions on the same subject is already the finding — record it and move on. A third probe buys little new information and changes how the room reads you. If the subject genuinely matters to your decision, carry the same question to a different interviewer later in the loop rather than escalating with this one.
- How would you phrase a follow-up so it reads as interest rather than suspicion?Put the gap on yourself and ask for one instance: something like 'I'm trying to picture what that looks like day to day — what's the most recent example?'. Ask a single question, not a stack, and let the person land it. The wording that reads as suspicion is usually leading or compound, not merely direct.
- Is asking for a number a good way to make a vague answer concrete?Usually less good than asking for an episode. A bare count arrives without context and you have no baseline to judge it against, so it feels concrete while telling you little. An episode you can picture — what broke, who decided, how it ended — is harder to produce from nothing and far easier to interpret.
saying these in an interview costs you the question
- Hearing the non-answer, nodding, and moving to the next prepared question
- Re-asking the same question at the same altitude in new words
- Stacking two or three probes onto one soft answer
- Phrasing the probe as a leading question that invites a denial
- Refusing to accept an honest 'I don't know' as a real answer