Measuring a Chinese Factory's Real Capacity in a 40-Minute Walk
Part of my job used to be visiting factories with a clipboard and pretending to be interested in the framed certificates. The most useful thing I learned is that capacity is arithmetic, not a number on a brochure, and that most of the inputs are countable during a normal visit if you know what to look at.
The claim you usually get is something like 50,000 units per month. Sometimes it is true. Often it is a theoretical figure derived from the number of machines times a cycle time times three shifts, with nothing deducted for changeovers, downtime, rework or the fact that only one shift actually runs.
Count the constraint, not the machines
Every plant has one step that limits everything else, and it is rarely the assembly line where the tour spends most of its time. It is the injection moulding machine with the longest cycle, the single reflow oven, the paint booth with the slowest bake, the one CNC that does the tight-tolerance op, or the testing station where every unit has to sit for 30 minutes.
Find it by asking the production manager, casually, "what do you wait for?". Then count what that station can do. If it is a paint line, count hangers per pass and ask the line speed. If it is a test bench, count the positions and the dwell time. That gives you a hard ceiling for the whole plant, and you can compare it with the claim.
A worked example from a small electronics assembly line: 24 test positions, 45 minute burn-in, running 20 hours a day, gives about 640 units per day theoretical, less a realistic 15 percent for handling and failures, so around 540. If the brochure says 30,000 a month, someone is counting a second shift that is not happening. That is not necessarily a lie, it is what happens when the factory has a good month and a marketing department.
Look for the second shift
The simplest capacity signal in China is whether the plant runs one shift or two. You can see it in the parking, the lighting, the canteen, the number of finished-goods pallets waiting, and the shift schedule posted on the notice board near the entrance. Ask directly: "how many shifts are you running now, and how many last October?" The answer tells you whether they are busy or hunting for orders, which is arguably more valuable than the capacity number, because it tells you where your order will sit in the queue.
Related: ask how many days of finished goods are sitting in the warehouse. A plant that ships daily has different cash flow and different lead-time honesty than one that runs campaign production and stores a month of stock.
The real lead time hides in changeover and materials
Quoted lead time is almost never about production. It is about the queue, the mould change, and the arrival of a component the factory does not make. Three questions get you most of the way:
- How long does a changeover take for my product, and how many changeovers do you run per week? A factory that changes over twice a month will slot you behind a full campaign; one that changes daily can absorb smaller orders.
- Which components are bought in, and what is the current lead time on the longest one? If the answer is a specific number of weeks rather than "normal", someone is actually tracking it.
- What is your on-time delivery rate for the last six months, and can I see it? Export-oriented factories increasingly track this because customers ask.
If the quoted lead time is 25 days and the imported component has a 45-day lead time, the 25 days assumes stock already exists. Ask who carries that stock and what happens if the next shipment is late.
Walk the floor with a counter, not a camera
Three counts that take ten minutes and reveal a lot:
WIP at the bottleneck. A pile of semi-finished units sitting before one station means that station is the constraint, and the size of the pile tells you how well they manage flow. Very large piles mean batch-and-queue production, which means longer throughput time and more rework discovered late.
Rework and scrap containers. Every plant has them. Look at what is inside and ask what the defect is and what the current first-pass yield is. A plant that quotes a yield in the high nineties and has visibly full rework bins has a reporting problem. A plant that says "about 92 percent, mostly cosmetic" and shows you the chart is telling the truth, and 92 percent is not a disaster.
People at the line versus people in the office. For a labour-intensive product, headcount on the floor times the takt time gives a rough capacity cross-check. If the brochure claims 300 staff and you count 60 on the floor in a single shift, ask where the other 240 are.
Certificates are a filing exercise; process discipline is not
ISO 9001 certificates are cheap and common, and I have stopped treating them as evidence. What is evidence: a documented change-control process (can they show you the record for the last engineering change on a product similar to yours), calibration stickers that are in date on the instruments you can see, a first-article inspection report with real measured values rather than ticks, and an incoming inspection area that is actually used.
Ask to see one nonconformance record, opened and closed, from the last quarter. Watching how a factory handles that request tells you more about quality culture than an hour of PowerPoint. Some will happily walk you through it. Some will say it is confidential. Some will look for someone who knows where it is, and that answer is the most informative of the three.
What to write down before you leave
Bottleneck station and its measured throughput, shift pattern, changeover time, longest bought-in component lead time, first-pass yield, headcount on the floor, and the name and role of the person who answered each question. Then, back at the hotel, multiply the bottleneck throughput by your realistic availability and see whether the number still looks like the quote.
If it does, you have found a factory worth a second visit. If it does not, you have found out for the cost of a taxi rather than for the cost of a missed season.