What do you do at work when you do not know the answer to something?
answer
- say the words, no decoration
- mine to solve or someone else's
- explicit timebox before asking
- ask small: tried, believe, one question
- close the loop in writing
basics
~20 sTests intellectual honesty and judgement. Say you admit the gap immediately, timebox your own investigation, escalate with what you already tried and one narrow question, then write the answer down where others will find it.
how to answer
5 beats- the words themselvesOpen by saying you say it plainly and early, and give the phrasing you actually use. Avoid decorating it with excuses or hedges; the flatness of the admission is the signal being read here.
- how I decide whose problem it isExplain the split between a gap you should close yourself because it is core to your work, and one where someone already holds the answer. One sentence of criteria is enough — cost of the delay, how core the knowledge is to your role.
- the timebox I give myself firstState a real limit and what you do inside it: reading, one experiment, a look at the artefacts. A concrete limit is far more convincing than saying you try for a reasonable while.
- how I ask so it costs minutes, not hoursDescribe the shape of your ask: what you already tried, what you currently believe, one specific question. Say who you pick and why. This is where a twenty-second real example is worth more than another sentence of principle.
- closing the loopFinish with what happens after the answer arrives: where you write it down, who else you tell, and one outcome that shows it mattered. Keep this short, but do not skip it — it is what turns a habit into value for the team.
your answer
4 story prompts- Recall the last time you said you did not know in a meeting, and what you did next.
- Write down your real timebox before escalating, honestly, not the one that sounds disciplined.
- Pick a moment where your framing made the answer cost a colleague minutes.
- Name one thing you now know because you admitted the gap instead of guessing.
draft and rehearse your own answer in a learn session
go deeper
This probes intellectual honesty and judgement under uncertainty. Interviewers want to know whether you can say you do not know without anxiety, whether you can tell a gap you should close yourself from one worth escalating, and whether your asks are cheap for colleagues. A strong answer shows a repeatable habit backed by one short real instance, not a claim about your personality.
My default is to say it quickly and without dressing it up. In a design review during our warehouse migration I was asked why my nightly job kept queueing behind the analysts' ad-hoc work. I said I did not know how compute was allocated on the new platform, and asked to bring the answer back at the end of the week rather than guess in the room. Then I set myself a limit: two hours of reading and one experiment before I take up anyone else's time. That afternoon I went through the concurrency documentation and ran the same job in an isolated compute pool. I got half an answer — I could see what was queueing, but not why our pool sized itself the way it did. So I went to a platform engineer with just the missing piece, not the whole topic. I pasted the two run logs, said what I already believed was happening, and asked one question. It took her about ten minutes. Afterwards I wrote it up in the team channel, because two other people were about to hit the same thing. The job moved onto its own pool, and its cost per run went from a little over ninety dollars to around thirty-eight. What I try to avoid is the middle state — nodding in the meeting, then burning a day reverse-engineering something a colleague could have told me before lunch.
The timebox is explicit and the escalation arrives pre-loaded with logs, a hypothesis and one question, so the ask costs ten minutes. Admitting the gap in the room rather than afterwards is the honesty signal; writing it up publicly and naming a result keep it from being a philosophy answer with no evidence behind it.
People ask me things now, so my answer has two halves: what I do with a gap, and what I refuse to do with one. When it is mine to resolve, I timebox it and say the timebox out loud. Partway through a warehouse migration, an internal auditor asked me in a meeting how row-level access would behave on the new platform for regional finance tables. I said I did not know yet, that I would have an evidenced answer by Thursday, and I put the question on the board so it could not quietly evaporate. What I have trained out of myself is answering plausibly. A confident guess is more expensive than an admitted gap, because several people start building on it the same afternoon. So I ran a small proof with two real tables and one masked role, and it surfaced exactly what a guess would have missed: the view-based approach we had sketched re-scanned the largest fact table on every query, which would have quietly added something near four thousand seven hundred dollars a month for a control nobody would have looked at until the invoice arrived. I came back with three findings and one recommendation, and said which of the three I was still unsure about. The habit underneath all of it is separating I do not know from nobody knows, and saying which one it is, because those need very different next steps.
The distinguishing move is treating a confident guess as the expensive option and pricing the consequence of being wrong. Committing to a date in public and returning with evidence rather than opinion carries the ownership signal, and flagging which finding remains uncertain reads as calibration. Dropping that uncertainty flag would make it sound merely confident.
Show an explicit limit on how long you dig before asking, and that you arrive with what you already tried. Nobody expects you to know much at this stage; they expect you not to sit stuck and silent for two days.
Separate the questions you should own from the ones worth a colleague's time, and show that your asks are compact: context, attempt, one question. Writing the answer somewhere durable afterwards is baseline expectation, not bonus.
You are now someone others ask, so cover both directions: how you model saying you do not know in public, and how you keep your own uncertainty from stalling a team. Naming what a confident wrong answer would have cost is strong.
Talk about calibration at scale — signalling confidence levels in writing, keeping your guess from hardening into an organisational assumption, and the norms that make people surface gaps early instead of at review time.
saying these in an interview costs you the question
- Answering plausibly instead of admitting the gap
- Describing a philosophy with no lived example attached
- No limit on digging alone, so being stuck lasts for days
- Asking open-ended questions that hand the whole problem to someone else
- Treating not knowing as embarrassing rather than routine
- Never sharing the answer once found
- How do you decide when to stop digging and ask someone?Give a real rule rather than a feeling: a time limit, a sign you are going in circles, or a cost threshold on the decision waiting behind you. Then say how the rule bends — a blocked teammate shortens it, a learning-valuable problem lengthens it. Interviewers are checking that the decision is deliberate, not mood-driven.
- Have you ever asked for help later than you should have?Answer yes, briefly, with what it cost and the signal you now watch for. A no reads as either lucky or unreflective. Keep it short and unglamorous — one situation, the delay, the consequence, the rule you adopted — and do not turn it into a second full story unless invited.
- What do you do when the person who would know is unavailable?Show a fallback path rather than helplessness: what you can settle from artefacts and experiments, how you make a reversible decision and mark it, and how you leave a question queued without blocking. Mentioning that you flag the assumption in writing, so it is easy to correct later, lands the calibration signal.