Most supplier audits end the same way. The auditor arrives at ten, is shown the quality manual, the ISO certificate and a conference room, walks the shop with the quality head for forty minutes, fills in a checklist that scores 87 percent, and leaves with three observations about housekeeping and a signed attendance sheet. Six weeks later the same supplier ships a batch with the wrong case depth, and nobody can explain how an 87 percent supplier did that.
The audit did not find it because the audit was not looking. It was checking whether a system existed. The question that matters is different and narrower: can this process, on this equipment, with these people, make my part to my specification on a normal Tuesday with nobody watching? Everything below is about answering that question in one day on site.
Before you go
The audit is decided before the car leaves. Three choices set it up.
Which supplier. Not the one whose turn it is on the annual schedule. The one whose process carries the most risk to your product: a special process you cannot inspect out at receipt (heat treatment, plating, welding), a single-source part, a supplier whose rejections or complaints have moved, or one about to take a new part. An annual rota that audits the stable packaging supplier and skips the heat treater because they were 'done last year' is a calendar, not a risk assessment.
Which process and which part. Pick one part number and one process route, and tell the supplier which. The aim is depth, not coverage. A full-day audit of hardening and tempering on one rocker arm will tell you more about that supplier than a tour of every bay. If they have five furnaces, you will audit the one your part runs in.
What your own data says. Pull the last twelve months of receiving results, rejections, complaints, deviations and the last audit's open points before you go. Read the supplier's control plan and process sheets for your part. Half the findings are visible in the data from your desk; the site visit is to find out why they happen.
Then write the day plan around the part: incoming material, loading, process, quench, temper, testing, release, despatch, in that order. Send it to the supplier. A supplier who knows you will follow one part through does not spend the morning rehearsing a presentation.
On the floor
Skip the conference room. Ask for the most recent batch of your part and follow it backwards from despatch to the raw material bin, at each step asking the same three things: what does the document say should happen, what did the record say happened, and what is happening right now at the machine.
At the furnace, that means reading the set-point on the controller and comparing it to the process sheet and to the chart recorder for the batch that went through last night. Asking the operator, not the supervisor, what the soak time is and what he does when the temperature alarm goes. Checking when the thermocouple was last calibrated and whether the furnace has a temperature uniformity survey from this year. Looking at the quench tank: is the agitation running, when was the quenchant last tested, is there a record.
At the hardness tester, pulling the calibration certificate and the daily verification block results, then asking the inspector to test a part from the batch in front of you. At the microscope, asking for the case-depth or microstructure report for your batch and whether the sample came from the batch or from a coupon that rode with it, and where the coupon was placed in the load.
At the record table, taking the batch card and tracing every signature to a person on shift and every number to an instrument that could have produced it. A hardness reading filled in before the batch finished tempering is a finding; so is a chart with a gap where the recorder pen ran dry.
Ignore the things that fill reports and change nothing: the paint on the floor, the motivational posters, the ISO certificate on the wall, the state of the canteen. Talk to operators more than managers. Spend at least half the day at the process, not in the office.
Findings that matter
A finding is worth writing when it describes a way your part can leave the supplier out of specification without anyone noticing. Everything else is an observation, and observations should be few.
The findings that matter in a special process are nearly always one of these: the process parameter is not what the control plan says, and nobody knew; the instrument that proves the result is not calibrated, not verified, or not capable; the record was written from memory or in advance; the person running the process was not the person qualified to run it; the sample tested does not represent the batch; nonconforming or reworked product can re-enter the flow without a documented release; or a change was made to the process, the equipment or a sub-supplier without telling you.
Write each one as evidence, not opinion. 'Process sheet PS-14 specifies 860 °C; controller set-point on furnace 2 at 11:20 was 845 °C; chart for batch 4471 shows 843 to 848 °C throughout; operator stated the set-point was reduced three weeks ago to lower distortion. No change record exists.' The supplier cannot argue with that, and the reader back home understands exactly what the risk is.
Do not score the audit. A percentage invites the supplier to negotiate points and invites your own management to read 87 as acceptable. State the findings, state the risk to the part, and state what must happen before the next shipment.
After the audit
Close the day with the supplier's management in the room, read every finding aloud with its evidence, and agree on the spot which ones need containment before the next despatch. Do not leave the site with a finding the supplier has not heard.
The corrective action that comes back two weeks later will usually be a retrained operator and a revised form. Reject that. Ask for the root cause in terms of why the system allowed it, the action that removes that cause, and the evidence that the action is in place: the new change-control record, the calibrated instrument's certificate, the survey report, the revised control plan with your sign-off. A corrective action closes when you have seen the evidence, not when the supplier says it is done.
Re-audit on risk, not on the calendar. A supplier with a finding about process parameters gets a short follow-up visit, focused only on that process, within the quarter. A supplier with no findings of substance goes back on the normal cycle. The schedule should follow the findings, and the findings should feed your supplier rating in a way the purchasing team can see.
A supplier audit is not a courtesy visit and not an inspection of the supplier's paperwork. It is a day spent finding out whether one process can be trusted with your part, by watching it and tracing it rather than by asking about it. Pick the supplier by risk, pick one part, follow it through, write what you saw, and close only on evidence. Do that, and the audit stops being a day that produces a score and starts being the day that would have caught the bad batch.
Done Your Supplier Audits With Best Experience Auditors at Advice Consultancy
