A relocated line is not moved until its night shift can run it

Since 2022 contract manufacturers across Southeast Asia have spent much of their capital, and most of their management attention, on lines arriving from China. Customers want parts for tools and devices sold in the United States made somewhere else, and Penang, Johor, northern Vietnam and Thailand's eastern seaboard have absorbed a remarkable amount of work in a short time. At Kelvorne we moved 212 part numbers out of our Suzhou plant in 2023 and 2024, sending 138 gas-line weldments to Penang and the frames and cable assemblies to Bac Ninh. Many of our peers have made larger moves.

Our industry has a well-developed method for these transfers, and I respect it. The customer approves the change of site. The process is frozen and copied. First articles are measured against the drawing, a validation run is inspected, and the customer signs. The old line is switched off, the visiting technicians go home and the project is reported to the board as complete. In most companies I know, the bar on the programme chart ends at that signature.

I think the signature marks the midpoint of a move, not its end. What remains is the part that decides whether the move was worth making, and it is rarely measured at all.

What a first article cannot show

A first article proves that the receiving plant can make a good part when the right people are present. It cannot show whether the plant can make the same part when they are absent. During qualification, the line is staffed by the best people from both sites. The transferring plant's technicians stand beside the new operators, engineers are on the floor for every shift that matters, and nobody runs a qualification lot at three in the morning.

The knowledge that keeps a line stable seldom lives in its process sheets. In gas-line welding it lives in how a welder prepares a tungsten electrode, how long a joint is purged before the arc is struck, what a welder hears when the argon flow is wrong and which tubing lot an experienced hand distrusts. In 2001, as a young engineer at a disk-drive component maker, I watched a line moved from Singapore to Thailand fail for eleven weeks because six operators had been adjusting a fixture by feel. The drawings had travelled. The feel had not.

We met the same problem in Penang twenty-three years later, with better methods and the same result. The day shift qualified on schedule. In August 2024, after our Suzhou technicians had returned home, night-shift first-pass yield on weldments fell from 97.6 per cent to 91.2 per cent within three weeks. Nothing in the process had changed. What had changed was who stood beside the welders when something went slightly wrong.

Count the people who can stop the line

When we examined the night shift, the gap was easy to state. On days, eleven welders could diagnose a purge or electrode problem and stop the line with confidence. On nights there were three, and on two nights a week there was one. The others were qualified to weld, which is a different thing. They could make a good joint, but they could not yet tell why a joint had gone wrong, so they did what sensible people do in a plant with a delivery commitment. They kept going and let inspection find it.

Since then I have asked every transfer team for a number that appears on no customer checklist: for each shift and each critical process, how many people can recognise a developing fault, stop the line and lead the diagnosis without help. We call it stopping depth. A transfer plan now has to show that number rising on every shift before the visitors leave, alongside the count of certified operators. In Penang we brought four Suzhou technicians back for ten weeks, paired each with two night-shift welders, and kept them until the night shift had seven people with that depth. By the first quarter of 2025, first-pass yield was 98.4 per cent on both shifts.

The cost of this approach is plain. Secondments run longer; extending ours cost about S$1.9 million, part of a total transfer bill of S$19.4 million against a budget of S$13 million. The customer, which wanted the Suzhou line closed and its programme reported as finished, was patient but not pleased. Seconded technicians spend more months away from their families, which deserves to be counted and managed rather than assumed. And a programme that ends in month fourteen rather than month nine looks worse in every quarterly review, however sound it is.

There is a fair counter-argument, which I hear most often from commercial colleagues. Some moves cannot wait. A tariff or a customer's deadline may make it better to move quickly and accept a period of weaker yield, sorting good parts from bad at inspection while the new team learns. I accept that this is sometimes the right choice. My objection is to making it silently. If a company decides to hand over a line before the receiving shift can run it, the board should know that it is buying speed with inspection cost and escape risk, and for how long.

An exit criterion boards can ask for

At Kelvorne we now define the end of a transfer by three conditions rather than one. Each shift at the receiving plant must hold its yield and delivery for eight consecutive weeks with no visiting staff on site. Each shift must have a stated minimum of people with stopping depth for every critical process. And the receiving team must have carried at least one engineering change through the customer's approval process on its own, because a line that cannot change without its original owners has not really been handed over.

The Bac Ninh lines that took frames and cable assemblies from Suzhou met all three conditions within ten months. They had the advantage of a plant whose local supervisors were trained before it opened. The Penang weldment lines took fourteen. Both figures are less flattering than the dates on which the customer signed, and both are closer to the truth.

Directors overseeing a relocation do not need to understand orbital welding to use this test. They can ask when the visiting staff will leave, what yield the night shift has held since they left, and how many people on each shift can stop the line. A management team that answers those questions with numbers has either finished its move or knows how far it still has to go. One that answers with the customer's approval date has finished the paperwork.