Quality Engineer Interview Questions

"I studied mechanical engineering and my first job was as a production engineer on a machining line. The part I kept getting pulled into was scrap. Every time we had a spike, I'd end up with the quality team going through the data and walking the line. One month we traced a run of oversize bores to a tool offset that one shift kept adjusting by hand, and fixing that felt better than anything else I'd done. After that I moved into quality properly. What keeps me here is that it's detective work with a real payoff: when you find the true cause, the problem stops and the operators stop fighting it every shift. Now I want a role where I own customer and supplier quality, not just the internal side."

"From your website and the trade show material, you make die-cast housings and machined brackets, mostly for vehicle and industrial customers. In that kind of work, the risks I'd expect are porosity in the castings, leak failures, and dimensional drift on the machined features as tools wear. Because some of your parts go into braking or steering systems, I'd also assume the customer treats certain characteristics as safety-critical, so traceability and a tight control plan matter a lot. I've worked with castings before, on a line where porosity was our top scrap cause, and I spent a year on the leak test and X-ray side. So I'd want to start by learning your top scrap and complaint causes, because that's where I think I could help quickly."

"I'd say checking parts is the smallest part of it. Inspection tells you a part is bad after it's already made. A quality engineer's real job is to set things up so the process makes good parts in the first place, and when something goes wrong, to find the true cause and make sure it can't come back. So I spend time on risk reviews before a launch, control plans, measurement systems, and working with suppliers so their parts arrive right. When a customer complains, I lead the investigation and the fix. I'd sum it up as: inspection is a safety net, and my job is to need it less every month."

"Quality assurance is about the process: the systems and checks that make it likely you'll make good parts, so it's mostly prevention. Quality control is about the product: inspecting and testing what you made to catch anything that's wrong, so it's detection. On a stamping line, an example of assurance would be the first-piece approval procedure, operator training on the work instruction, and a preventive maintenance plan for the dies. An example of control would be the final inspector checking hole positions on a sample from each batch with a gauge, or an in-line vision system rejecting parts with a missing feature. You need both, but a good system leans more on assurance, because control only tells you after the fact."

"ISO 9001 doesn't tell you how to make your product. It asks you to run the business as a set of defined processes, control them, and prove it with evidence. In practice that means understanding what customers need, deciding which processes matter and how they connect, thinking about risks and opportunities, and keeping the documents and records that show things were done as planned. It's built around plan, do, check, act. The check part is internal audits, monitoring and a management review where leadership looks at the data and makes decisions. When something goes wrong, you're expected to correct it, find the cause and stop it recurring. And the whole thing is meant to improve over time. Certification just means an outside body has audited that you really work this way."

"In an internal audit, the auditor found a torque wrench on our assembly line that was past its calibration date. The quick fix would have been to calibrate that one wrench, and we did that the same day. But I also checked every other gauge in the area and found two more overdue. The cause was that our calibration list lived in a spreadsheet one person maintained, and when she was on leave nobody checked it. So we moved the schedule into the maintenance system, which sends reminders to the area lead, and added a coloured due-date label on each tool so operators could see it. I also checked whether any parts had been assembled with the overdue wrench and confirmed its readings were still fine when calibrated. The follow-up audit closed it with no repeat."

"I'd stay calm and first make sure I understand it. I'd ask which requirement they're citing and what they saw that doesn't meet it. Sometimes a finding comes from the auditor not seeing a record that exists, so I'd offer to show the evidence right there, like the training record or the procedure section that covers it. If it's a difference in interpretation, I'd explain how we read the requirement and why our approach meets its intent, politely and briefly. Auditors usually change a finding when you show objective evidence. If they still stand by it, I wouldn't argue on and on in the closing meeting. I'd accept it for the record and, if I really believe it's wrong, use the certification body's formal appeal process afterwards. And honestly, I'd still ask myself whether they've spotted a weak point, even if it isn't technically a nonconformity."

"I use the fishbone to go wide and the 5 whys to go deep. Say we're getting burrs on a drilled hole. I'd get the operator, the setter and maintenance together and fill in the fishbone under the usual headings: people, machine, method, material, measurement and environment. That gives us maybe fifteen possible causes. Then I check them against data, like whether the burrs follow one spindle, one shift or one material lot, and cross off what doesn't fit. On the cause that survives, I ask why repeatedly. The drill is dull. Why? It ran past its tool life. Why? The counter was reset during a changeover. Why? The changeover sheet doesn't say not to. That's a cause I can fix and test. Five is just a guide; I stop when I reach something actionable."

"Day one, I'd make sure the customer's protected: is the scrap being caught, or could some be escaping? If there's any doubt, I'd tighten checks at the end of the line. Then I'd get the scrap data and break it down by defect type, machine, shift, material lot and date, and build a Pareto, because doubled scrap is rarely one problem, it's usually one or two defect types that jumped. Once I know which defect grew and when it started, I'd go to the line and talk to the operators and setters, and pull together a list of everything that changed around that date: a new material supplier or lot, a tool change, a program edit, a maintenance job, new staff, even a new gauge. By the end of the week I want a focused problem statement and one or two hypotheses I can test, not a fix I picked on day two."

"At my last company we had a leak failure on an aluminium housing that came and went for months. The team kept blaming casting porosity. When I split the failures by machine, they all came from one of three machining centres, which porosity wouldn't explain. We sectioned failed parts and found the leak path was a tiny burr on a sealing face, left by a worn deburring tool that nobody had on a tool-life counter. We put the tool on a counter, added a probe check for the sealing face, and changed the tool path so the burr couldn't form. To prove it, I ran a worn tool on purpose for a short controlled trial, with every part quarantined, and the leaks came back, then went away with a fresh one. After that, leak failures from that machine dropped to zero over the next three months, and we copied the fix to the other two."

"We had cracked plastic clips at a customer, and early on we found a batch of resin with high moisture, so I blamed drying and we tightened the dryer checks. Cracks dropped for two weeks and then came back, with properly dried material. That told me I'd found a contributor, not the cause. I went back and looked at where on the clip the cracks started, and they were always at the same corner. That corner had a sharp internal radius, and the customer had recently switched to a faster assembly machine that flexed the clip further. So the real cause was a design margin problem made visible by a process change at the customer. We added a radius to the tool. What I do differently now is test the cause before I close anything: if I can't turn the defect on and off with it, I don't call it the root cause."

"8D starts with forming a cross-functional team, some versions add a planning step before that. D2 is describing the problem precisely: what, where, when, how many, what's different from good parts. D3 is containment to protect the customer right away, like sorting stock. D4 is root cause, and I always look for two: why the defect was made, and why our controls didn't catch it. D5 is choosing corrective actions and proving they work, D6 is implementing them and validating with data. D7 is preventing recurrence more widely, updating the FMEA, control plan and similar processes. D8 is recognising the team. Teams usually cut corners at D2 with a vague description, at D4 by stopping at the first plausible cause, and at D5 by writing 'operator retrained', which is almost never a real fix."

"A correction deals with the bad product itself. If we find twenty parts with a missing weld, reworking or scrapping them and sorting the stock is the correction. A corrective action goes after the cause, so if the missing weld came from a robot program that was edited without control, locking down program changes is the corrective action. It's aimed at stopping that problem coming back. A preventive action is about a problem that hasn't happened yet. If we realise the robot on the next line has the same unlocked setup, fixing that before it ever fails is preventive. Since the 2015 edition of ISO 9001, preventive action isn't a separate clause any more; it's covered by risk-based thinking. The idea is the same, and a lot of CAPA systems still track it."

"I wouldn't try to close forty items in a week, because that just produces paper closures. I'd go through the list and sort it by risk: anything tied to a customer complaint, safety or an audit finding goes to the top. Then I'd clean it up. Some will be duplicates, some will actually be done but never closed, and those need evidence before I close them. Some will be corrections pretending to be corrective actions. For the ones that matter, I'd sit with each owner, agree what's really needed and set dates they can actually meet. Then I'd look at why it built up in the first place: often actions are assigned to people with no time, or nobody reviews the list. So I'd set a short weekly review with the managers who own the resources, and track the count and age so everyone can see it shrinking."

"A control chart plots a measurement over time so you can see whether the process is stable. The centre line and the control limits are calculated from the process's own data, typically three standard deviations of whatever you're plotting on each side. As long as points stay inside and don't form patterns, you're seeing common cause variation, the normal noise of the process. A point outside the limits, or a pattern like a long run on one side of the centre line, is a signal of a special cause worth investigating. Specification limits are completely different: they come from the drawing or the customer and say what the part must be. So a process can be in control and still make bad parts, or be out of control while every part is in spec. That's why you never draw spec limits on an averages chart."

"Cp compares how wide the tolerance is with how wide the process spread is, six sigma. It ignores where the process sits. Cpk takes the distance from the process mean to whichever spec limit is closer and divides by three sigma, so it penalises being off-centre. If Cp is good but Cpk is poor, the process is capable of fitting inside the tolerance but it's aimed wrong, sitting close to one limit. That's usually good news, because re-centring through an offset or a setting is easier than cutting variation. Before I trust either number, the process has to be in statistical control, the gauge has to be good enough, and the data should be roughly normal. Many customers ask for at least 1.33, and often higher for new processes. I'd also say whether I'm quoting Cpk from within-subgroup variation or Ppk from overall variation, because they can differ a lot."

"I'd explain to the supervisor that the chart is doing its job. In spec tells us today's parts are fine. Out of control tells us something has changed in the process, and if we ignore it, the next parts might not be fine. The whole point of SPC is to react before we make scrap. So I'd go and look straight away: check the measurement wasn't a one-off error, then look for what changed, like a tool change, new material or a temperature shift. I'd follow the reaction plan in the control plan, which might mean checking the last few parts more closely. If the cause is obvious and fixable, we fix it and keep running. If we can't find it and the trend is heading towards a limit, I'd agree with the supervisor on increased checking while we investigate. I'd note the cause on the chart either way."

"I start from the process flow diagram with a team that includes the operators, the process engineer and maintenance, not just quality. For each step we ask how it could fail, what the effect would be on the next operation and on the customer, and what could cause it. Then we list the current prevention and detection controls and rate severity, occurrence and detection. The trouble with ranking purely by RPN, which multiplies those three, is that very different risks can get the same score. A safety-related failure with low occurrence can score lower than a cosmetic defect that happens often, and teams chase the wrong one. So I always treat high severity items on their own, whatever the product is. The joint AIAG and VDA FMEA handbook replaced RPN with an action priority table for exactly this reason, and severity weighs heaviest in it."

"APQP is the planning framework for launching a new part. It runs in phases: plan and define the programme, product design and development, process design and development, product and process validation, and then feedback and corrective action once you're running. PPAP is the evidence package you give the customer to show that the production process, on real tooling at real rate, makes parts that meet every requirement. So PPAP is really the output of the validation phase. The package includes things like the design records, process flow, PFMEA, control plan, measurement system studies, full dimensional results, material and performance test results, initial capability studies, and the part submission warrant, which is the summary document you sign. The customer sets the submission level, which decides how much of that you send versus keep on file."

"We were launching a new assembly with two nearly identical O-rings, one for fuel and one for coolant, different materials, same size. In the process FMEA, an operator on the team asked what would stop someone swapping them. Nothing did: same bins, same colour, same size. The effect would have been a seal failing in the field, so severity was at the top of the scale. We had the supplier change the fuel seal to a different colour, gave the two separate feeding points, and added a vision check at the station that confirms the colour in each groove. During the pilot run the vision system rejected two assemblies where someone had grabbed from the wrong bin. Without that control, those would have shipped. I always credit that operator when I tell the story, because it came from the shop floor, not from me."

"For attribute inspection I'd normally use a standard sampling table. You take the lot size and the inspection level, usually the general level two, and that gives a code letter. The code letter and the agreed AQL give you the sample size and the accept and reject numbers. There are also switching rules to tighten or relax inspection based on recent results. Where it falls short is that the AQL is about the long-run quality level of the process, not a guarantee about the lot in front of you. You can accept a lot that has defects in it, just by chance. So for anything safety-related or critical to the customer, I wouldn't rely on sampling at all. I'd push for error-proofing or full automated inspection. Where sampling is still used on those features, many customers expect a zero-acceptance plan, where one defect rejects the lot."

"A gauge R&R tells me how much of the variation I'm seeing comes from the measurement system rather than the parts. Repeatability is the gauge itself: the same person measuring the same part several times. Reproducibility is the difference between people. The usual study is about ten parts that cover the real process range, two or three appraisers, and two or three trials each, measured blind and in random order. Then I compare the measurement variation with the tolerance, and with the process variation if the gauge feeds a control chart. As a rough guide, under a tenth of the tolerance is good, up to about three tenths can be acceptable depending on the feature, and beyond that I wouldn't trust it. If I skip this, a capability study blames the process for gauge noise. For go/no-go gauges I'd run an attribute agreement study instead."

"First I'd make sure we don't ship anything bad. I'd find out how long the check has been skipped and hold or re-check the parts made in that time. Then I'd talk to the operator calmly, one to one. I'd want to know why, and usually the answer is that the rate isn't achievable with the check included, so they're choosing between two things they've been told to do. That's a system problem, not a character problem. I'd raise it with the shift supervisor and the production manager: either the rate or the staffing changes, or we find a way to make the check faster, like a better gauge or a fixture. I'd also check whether the other shifts were doing the same thing. If the operator had been skipping it knowingly with no pressure, that's a different conversation, but I'd start by assuming there's a reason."

"A customer found mixed parts in a box: two variants of a bracket that look almost the same but have a different hole position. I called their quality contact the same day, agreed a sort at their site and in our warehouse, and sent our own people to do it. We found a handful more in stock. For root cause, the two variants ran on the same press back to back, and leftover parts from the first run stayed in a bin that got emptied into the next box. It escaped because final inspection sampled by appearance, not hole position. The fix was a proper line clearance check at changeover and a simple fixture at packing that only accepts the correct variant. I sent the 8D within the time they asked, updated them weekly, and they closed it after three clean months."

"We had a plastic moulding supplier whose parts kept failing our incoming check for short shots and flash, and they were on our worst-supplier list for months. Their 8Ds all said 'operator retrained'. I asked to visit instead of emailing another rejection. On the floor, I found the moulds were running without any process monitoring, and setters adjusted pressure by feel after each start-up. We agreed a joint plan: they set a validated process window, added monitoring on cavity pressure, and did a proper first-off check after every start-up. I also realised our drawing had an unclear callout on flash, so we fixed that on our side. I visited monthly for a quarter. Rejects dropped to near zero, and we moved them from full incoming inspection back to normal sampling."

"I wouldn't let it ship quietly, however small the miss. Out of spec means nonconforming, and it's the customer's decision, not ours, whether they can use it. But I also don't want to just say no and walk away, because the delivery problem is real. So first I'd get the facts: which characteristic, how far out, how many parts, and whether it's a safety or fit feature. If it's safety-related, it doesn't ship, full stop. If it's minor, I'd call the customer's quality contact with the data and ask whether they'll grant a written deviation for that quantity. At the same time I'd look at sorting out the good parts, reworking what we can, or running a fresh batch fast. And whatever happens, I'd open a corrective action, because the real question is why the process made them."

"First I'd get the details from the customer: part number, lot or date code, photos, how many, and where on their line they found it. I'd thank them, not argue. Then containment, because that protects them while we investigate. I'd put a hold on all suspect stock in our warehouse, in work in progress and anything in transit, and arrange a sort at their site, either with our people or a sorting service, using a clear check for cracks. I'd ask for the cracked part back urgently so we can look at it. Before the day ends, I'd send the customer a short note of what we've contained, how the sort works and when they'll get the next update. Then I'd open the 8D, get the team together and start looking at the lot history and the process for that date."

"I'd take their claim seriously, because sometimes they're right. First I'd get us measuring the same thing: pick a handful of parts, measure them in their lab and ours, and compare. If the results don't match, we need to agree on the method, the datum setup and the gauge before anything else. I'd also sit down with our design engineer and read the drawing callout they're disputing. If it's ambiguous, that's on us, and we clarify it formally with a drawing change. If the measurement correlates and the drawing is clear, then the parts really are bad and I'd expect a proper corrective action from them, with evidence, not an argument. Either way, until it's settled, I'd keep containment in place so bad parts don't reach our line while we talk."

"On a welding line, the supervisor was proud that his team could set parameters by eye and hated the idea of fixed settings. I needed them locked because we had lack-of-fusion and burn-through defects that tracked with different welders. Instead of issuing a new instruction, I sat with his best welder for a shift and asked what he adjusted and why. It turned out most adjustments were compensating for poor fit-up from the previous step. So I proposed a trial: fix the fit-up issue first, then lock the parameters on one cell for two weeks and compare. The trial cell had far fewer defects and the welders said it was easier, because they weren't fighting the parts. The supervisor presented the results to the plant manager himself, and we rolled it out to the other cells."

"People hide bad parts when they think they'll be blamed, or when the pressure on output means stopping feels like the wrong choice. So the first thing is how we react. If an operator flags a problem, the first words from me or the supervisor should be thanks, not who did this. I'd make it easy too: a clear way to raise it, like a signal light or a simple card, and a promise that someone from quality or the team leader arrives quickly. Then I close the loop, telling the operator and the shift what we found and what we changed. When people see their report actually fixed something, they report more. I'd also track the number of issues raised as a good thing, not a bad thing. It takes months, and one supervisor shouting can undo it in a day, so I work hard to get the supervisors on board."

"I try to be on the floor a lot, so production sees me when things are going well, not only when I'm there to hold a batch. When I do have to say no, I try to pair it with a way forward: this can't ship, but here's how we sort it quickly, or here's what the customer might accept with a deviation. I also try to make their job easier, like fixing a gauge that's slow to use or simplifying a work instruction nobody could follow. That builds trust, so when I dig my heels in on something safety-related, they know it's not just me being difficult. In the end, we both want the same thing: good parts, on time, without firefighting. If production sees me helping them get there, the us-and-them feeling goes away."