Customers should audit our records, not our tour route

In 2024 Kelvorne hosted 41 customer audits across our five plants. They covered quality systems, process controls, information security, export control, labour practice, environment, and health and safety. On top of them came the certification audits for our quality-management and medical-device standards, and the follow-up visits that close the findings of earlier ones. In an average month, at least one of our plants was preparing for an audit, hosting one or answering one.

I do not resent this. Our customers put our parts inside tools that cost millions of dollars and into fabs that cost billions, and they are right to check what we do. For four years I was one of the people doing the checking, as supplier-quality director for a wafer-equipment maker across Asia Pacific. I believe in audits. I have also come to believe that the way our industry conducts them tells customers less than they think and costs suppliers more than they see.

The route every auditor walks

An audit is announced weeks ahead. The plant prepares a room, a presentation and a route. The route is not dishonest; it usually goes where the auditor asks to go. But it goes there in daylight, on a weekday, with the plant's most articulate engineers present, after a fortnight in which every record the auditor might sample has been checked twice.

As a customer's auditor I once examined a supplier that had passed every visit for three years. Its weld-qualification records were complete and its parts were good. On an unscheduled return, we found that the night shift's weld-parameter checks had been filled in by the day supervisor the next morning, from memory. Nobody had meant to deceive us. The night shift had never been staffed to complete the record, and the day supervisor had made the paperwork whole. The audits had examined the record. None had examined the shift.

That experience shapes how I look at my own plants. The question a customer most needs answered is not whether our procedures are good, but whether a new operator in their second month follows them at three in the morning. A prepared visit is poorly designed to answer it.

What forty audits cost a plant

The cost of audits is rarely added up, so we did it. In 2024, hosting and answering customer audits took about 9,600 hours of staff time across Kelvorne, most of it from quality engineers, plant directors and the managers of our dormitories and environmental systems. That is roughly the annual working time of five experienced engineers. Much of it went on answering the same questions in different formats: each customer has its own checklist, its own scoring and its own preferred evidence, even where the underlying standard is common to the industry.

The less visible cost is attention. A plant preparing for a visit cleans, rehearses and rechecks its files, and the engineers doing that are not working on yield. A plant that learns that audits are passed by preparation also learns, gradually, to manage the record rather than the process. That is the habit I saw at the supplier with the tidy weld logs, and it is the one I most want to keep out of our own plants.

Customers pay as well, though they seldom count it as a cost of their own. Each of the 41 visits brought two or three auditors, usually for two days and often after a flight. That is well over a hundred auditor-days a year spent walking routes through our plants, a good share of them confirming what another customer's auditor had confirmed a month earlier. Those are scarce people, and I would rather they spent their days where a supplier's risk is genuinely unknown.

Open records instead of more visits

The alternative I have argued for, and which three of our largest customers have now tried with us, has two parts.

The first is a shared audit for practices that do not differ by customer. Labour practice, dormitory conditions, environment, and health and safety are judged by all three customers against much the same industry code. Since early 2025 an independent firm has audited those areas once a year, with interviews on every shift and visits at night and at weekends, and has reported the results to all three customers at once. Each customer can add questions, and any of them can still arrive unannounced. In its first year the arrangement replaced eleven separate visits.

The second part matters more. Between audits, the three customers have read-only access to the records a good auditor would sample: first-pass yield by process and by shift, open deviations and corrective actions with their ages, welder qualifications and their expiry dates, weekly working hours by worker group, and dormitory occupancy against capacity. These records are prepared for nobody. They are the same records our plant directors read each morning, and if the night shift's yield or hours drift, the customer sees it when we do.

That openness has a cost, and I should state it. Customers see our bad weeks without our explanations attached. In the first months two of them asked about yield dips that we would once have resolved before any auditor arrived, and the questions took time to answer. Some colleagues worried that a customer would use the data in price negotiations, and one commercial manager is still not persuaded otherwise. There are limits, too. Records that touch another customer's designs, or export-controlled material, cannot be shared, and we had to separate them with care.

The benefit is that our customers now judge us on what happens every day rather than on what happens when they are expected. The findings of this year's shared audit were fewer and more specific than those of earlier years: a Johor dormitory running above its planned occupancy for six weeks during a ramp-up, and overtime on one Penang line that approached the sixty-hour limit before we added a shift. Both findings were real. Neither would have been visible on a prepared route.

For customers' supplier-quality leaders, my suggestion is to ask suppliers for their records rather than their presentations, and to visit when the plant does not expect you. For directors of contract manufacturers, it is to ask how many hours the company spends on audits, and whether those hours make the plants better or only better prepared.