Failure analysis questions.
83 failure analysisquestions from the bank — open to read. Pick one and practice it out loud; a coach note comes back in seconds.
Learn the ideas first
All 83 questions
A 30-year-old highway bridge developed unexpected cracks in a girder. Walk me through your initial hypotheses.Walk me through how you'd preserve and prioritize evidence at the site of a structural failure.Walk me through how you'd reconstruct the sequence of a failure when no one was present at the time.Walk me through how you'd distinguish primary cause from contributing factors in a complex failure.Walk me through how you'd structure prevention recommendations after identifying a failure cause.A weld in a pressure vessel cracked under nominal operating conditions. Walk me through your hypotheses.Walk me through how you'd handle a failure where the most informative evidence has been destroyed by responders.Walk me through how you'd handle conflicting witness accounts of a failure.Walk me through how you'd handle a failure where multiple parties have potential liability.Walk me through how you'd communicate prevention recommendations when implementation would require an industry-wide change.Walk me through how you'd approach a failure that doesn't match any known pattern.Walk me through how you'd handle evidence collection when there's pressure to restore operations quickly.Walk me through how you'd use simulation to reconstruct a failure event.Walk me through how you'd identify latent factors that contributed to a failure but weren't direct causes.Walk me through how you'd write recommendations that are specific enough to act on but general enough to apply broadly.Walk me through how you'd handle a failure where multiple causes look equally likely.Walk me through how you'd respond when the manufacturer or operator wants to limit what you can examine.Walk me through how you'd handle a failure where data logs and physical evidence don't match.Walk me through how you'd handle a failure that's primarily organizational rather than technical.Walk me through how you'd avoid recommendations that just transfer the problem to another part of the system.Walk me through how you'd handle a failure where the suspected cause has been ruled safe by the standards body.Walk me through how you'd handle suspected counterfeit components found post-failure.Walk me through how you'd handle a failure where the timeline is contested.Walk me through how you'd respond when stakeholders push to limit the investigation to a narrow set of causes.Walk me through how you'd handle prevention recommendations that the client doesn't want to implement.Walk me through how you'd structure an investigation team for a complex multi-disciplinary failure.Walk me through how you'd handle a failure where the relevant standards have changed since the original design.Walk me through how you'd test a hypothesis about a failure using a physical sample.Walk me through how you'd describe contributing factors in a report when the primary cause is contested.Walk me through how you'd think about whether a failure means redesigning, refurbishing, or replacing.A gear in a conveyor system broke after six months of service, well before its rated lifespan. What are your initial hypotheses for why this might have happened?You arrive at a site where a pipe has ruptured and fluid has leaked everywhere. What physical evidence would you try to collect first, and why?An electronic component failed, but the failure wasn't noticed until hours later. How would you determine when the failure actually occurred?A motor overheated and failed. You find it was undersized for the application and also poorly ventilated. How would you explain which factor was more important?After finding that a bearing failed due to contaminated lubricant, what recommendations would you make to prevent this in the future?A software-controlled valve closed unexpectedly, causing a process shutdown. What are three possible root causes you would investigate?You need to document a failed component that will be shipped to a lab for analysis. What information and photos should you capture before it leaves the site?A tank overflow occurred, but the level sensor logs show the tank was only half full. How would you figure out what actually happened?A bolt broke during installation. You find it was both overtorqued and had a pre-existing crack. How do you determine which was the primary cause?You've determined that a pump failed because the operator didn't follow the startup procedure. What recommendations would you make beyond just 'follow the procedure'?A battery pack in a device swelled and stopped working. What hypotheses would you consider about what caused this?During a failure investigation, you notice corrosion on multiple components, not just the failed one. How would you decide what role corrosion played in the failure?A root-cause review just concluded 'human error' and everyone seems satisfied. Why is that answer suspect, and what would you ask next?Your team's ML model degraded quietly for three months before anyone noticed, and the post-incident review is yours to run. Where do you start, and what makes this different from reviewing an outage?An automated decision system produced a harmful outcome while operating exactly as designed — nothing technically broke. How do you investigate a failure with no defect in it?Midway through an incident review you realize the telemetry needed to answer the central question was never built. What do you do with the investigation — and what do you do about the gap?Your team shipped a fix after what felt like a thorough root-cause review, and the same failure recurred a month later. What now?Your organization runs blameless postmortems, but this time an engineer skipped an explicit runbook step. How do you keep the review honest about that without turning it into a trial?Tell me about a near miss — a failure that almost happened on your watch but didn't. How did you treat it differently from one that did, and should you have?A critical vendor's failure took your operation down, and the vendor's own postmortem is two thin paragraphs. How do you run your side of the investigation when half the evidence belongs to someone else?A model failure traces back to bad training data that passed every validation check you had. How would you reconstruct how it got in — and decide where the real defect lives?In a root-cause review, how do you decide when to stop asking 'why'? Give me a real example of where you drew the line and what you left unexamined.Six months after a serious incident, how would you audit whether the postmortem actually changed anything?Customers are reporting real harm, but every dashboard for that period stayed green. How would you investigate a failure your own metrics insist never happened?You've read your organization's last ten incident reports and the same contributing factor keeps appearing under different names. How do you make the case that this is one systemic failure, not ten one-offs?A junior engineer's postmortem draft names individuals and assigns blame throughout. As the reviewer, what do you change in the document — and what do you teach them?A data center cooling system failed, causing server overheating in three racks before automatic shutdown. Walk me through your initial hypotheses about root cause.Walk me through how you'd prioritize evidence collection when a failure involves both physical components and digital logs that are overwriting in real-time.A pharmaceutical batch failed sterility testing three days after production. Walk me through how you'd reconstruct what happened during those 72 hours.Walk me through how you'd separate design flaws from maintenance issues when both appear to contribute to a mechanical failure.After identifying operator error as the proximate cause, walk me through how you'd frame prevention recommendations to avoid a blame-focused response.Walk me through how you'd facilitate a postmortem when the team is defensive because this is their third failure in six months.A surgical robot experienced a momentary loss of position tracking that corrected itself before any patient harm. Walk me through your near-miss analysis approach.Walk me through how you'd build a failure hypothesis when the component that failed was a black-box third-party module with no internal documentation available.A wind turbine blade separated during operation. Walk me through how you'd secure the failure site when debris is scattered across private farmland.Walk me through how you'd distinguish between a latent design defect and recent environmental changes when analyzing a failure in a system that's operated successfully for five years.Walk me through how you'd structure prevention recommendations when your analysis reveals the failure was predictable from ignored warning signs.Walk me through how you'd run a postmortem when the failure caused significant customer impact and executives are attending the session.An oil pipeline pressure sensor briefly exceeded its alarm threshold but returned to normal within 30 seconds. Walk me through your near-miss investigation approach.Walk me through how you'd gather evidence when a failure occurred in a clean room environment and your investigation could contaminate ongoing production.Walk me through how you'd reconstruct the failure sequence when the only available evidence is contradictory telemetry from redundant sensors.Walk me through how you'd assess whether inadequate training was a contributing factor without access to training records or the ability to interview the operator.Walk me through how you'd develop prevention recommendations when the root cause is a rare combination of conditions unlikely to recur naturally.Walk me through how you'd facilitate a cross-functional postmortem when hardware, software, and operations teams are each blaming the other groups.A construction crane's load limiter activated unnecessarily, stopping work but preventing no actual danger. Walk me through your near-miss analysis approach.Walk me through how you'd develop failure hypotheses when the failed component is so damaged that forensic testing cannot determine its pre-failure state.Walk me through how you'd reconstruct the timeline of a chemical reactor upset when the control system's clock was found to be 14 minutes fast.Walk me through how you'd evaluate whether organizational cost-cutting was a contributing factor when you have limited visibility into budget decisions.A finished tablet lot returns an out-of-specification assay result on one of six samples. Walk me through your initial hypotheses before you decide whether it's real or a lab artifact.A blend uniformity failure shows up intermittently across several batches of the same product. Walk me through how you'd separate the true root cause from the contributing factors.A settle plate in your Grade A filling zone grows an organism above the action limit. Walk me through what evidence you'd secure, and in what order, before it's disturbed.A commercial product fails a stability time point at 18 months when earlier points passed. Walk me through how you'd reconstruct what changed to explain the drop.You've traced a cross-contamination event to a shared piece of equipment and an incomplete cleaning verification. Walk me through how you'd structure prevention so it addresses the system, not just that one line.