TL;DR: Act decisively without being reckless: gather what you can fast, make the call on the best available evidence and explicit assumptions, prefer the reversible path, and instrument a way to learn whether you were right. Use STAR, say "I," and quantify the outcome. The signal is judgment under uncertainty, not waiting for perfect information.
The senior move is the two-way-door test: classify the decision as reversible or not, then spend your caution budget accordingly.
How to approach it. This screens for bias for action balanced with sound judgment (Amazon tests it explicitly; speed-driven cultures like NVIDIA and Tesla too). Pick a real decision where you genuinely lacked full information and time mattered. Structure it STAR with the weight on how you reasoned and de-risked, not just that you were fast.
A strong answer. Walk the arc with a real example:
- Situation: real time pressure and missing data. "A customer-facing model was misbehaving before a launch; I had a few hours to decide whether to ship, delay, or roll back, without a full root cause yet." Make both the deadline and the uncertainty concrete.
- Gather what you can, fast. "I pulled the signals available in the time I had, error rates by segment, a quick comparison to the previous model, and the blast radius, instead of waiting for a complete investigation."
- Decide on the best available evidence plus explicit assumptions. "The data suggested the regression hit a small segment, so I shipped to the unaffected majority behind a flag and held back the affected segment, stating the assumption that the issue was segment-specific."
- Prefer the reversible path. "I chose the option cheapest to undo, a flagged partial rollout with one-flip rollback, so if my assumption was wrong the cost was bounded." Separating reversible from irreversible (and moving fast on the former) is the senior move; Amazon frames many decisions as two-way doors.
- Set up to learn and validate. "I instrumented the rollout so we would know within an hour whether the assumption held, and pre-agreed the rollback trigger." You do not just decide; you build the feedback that confirms or corrects it.
- Outcome and reflection. Quantify the result and note what you would gather earlier next time. If the call turned out wrong, owning that and how you corrected is also a strong answer.
The throughline: under uncertainty and time pressure, you act on the best available evidence with explicit assumptions, favor reversible decisions, and build in a fast way to validate, instead of freezing for certainty or betting recklessly.
Key takeaways.
- Classify reversible vs irreversible first; spend speed on two-way doors and caution on one-way doors.
- State the assumption out loud and bound the downside (flag, partial rollout) so being wrong is cheap.
- A pre-agreed rollback trigger and instrumentation turn a guess into a testable bet.
What interviewers probe next.
- "How did you decide it was reversible?" Could I undo it cheaply (flag, rollback) versus a one-way door like data deletion or a public commitment? Reversible warrants speed; irreversible warrants caution.
- "What if your assumption was wrong?" I bounded the downside (partial, flagged) and set a trigger to detect and revert quickly, so being wrong was a fast correction.
- "How do you avoid analysis paralysis?" Time-box the data gathering, decide on the best available evidence, and treat the decision as updatable.
- (Amazon) maps to "Bias for Action" and "Are Right, A Lot"; show you move fast on two-way doors.
Common mistakes.
- A story where you actually had all the data, or where you stalled until certainty (no bias for action).
- Acting recklessly with no risk assessment or rollback (the opposite failure).
- Not distinguishing reversible from irreversible decisions.
- No mechanism to learn whether the call was right, and no reflection.