Splitting a dozen part numbers across a dozen vendors looks like risk management. Measured over a decade, it usually is not. A view from the factory floor.
Procurement orthodoxy on molded plastic parts has been stable for years: split the business, keep everyone honest, re-quote annually. On a bill of materials where the molded items are worth a dollar or two each, that discipline feels almost free. You are managing a rounding error, and you are managing it competitively.
Then something on the line stops. A handwheel slips on a valve stem. A seat inside a check valve goes slightly out of round and the assembly leaks on test. Now the one-dollar part owns your week — and the vendor who made it has been on your books for eleven months and has never seen that failure before.
I have watched this from the other side of the transaction. I run a second-generation injection molding factory in Ningbo, China. We mold handwheels, check valve seats, valve stem guides and complete flow meters for a European medical-gas equipment OEM, and dental consumables for a North American brand. Those two relationships are seventeen and twenty-nine years old respectively — old enough to have outlived several generations of the products they started on. What follows is what that longevity actually produces for the buyer, and what it does not.
1. The asset is the tooling record, not the machine
Machines are not a moat. Any competent factory can buy tonnage, and a buyer comparing two quotes is usually comparing two companies with broadly similar equipment lists.
What cannot be bought quickly is the accumulated record of how one specific tool behaves. The oldest mold on our floor was cut in 1996 and is still in production today. The program it belongs to now runs across seventeen dedicated tools for a single customer, at roughly one million parts a year, a 17-second cycle, 99% yield and tool life in the region of three million shots.
None of those numbers were true in year one. They are the residue of thirty years of small corrections — gate position, cooling, ejection sequence, the order in which the tool gets cleaned during a long run. Move the program to a cheaper quote and none of that transfers. You restart it, at your own cost, in the currency of first-article rejections and stopped lines.
2. Buy part families, not part numbers
Consolidation done well is not “fewer suppliers.” It is one supplier per part family, chosen because the family has internal logic that somebody has to hold in their head.
The handwheel family we build for the medical-gas OEM runs to about forty variants, 26 to 66 mm across. They are not one part in different sizes. The material is selected by the torque the wheel has to survive: small flow-setting wheels in unfilled PA6, higher-load wheels in PA66, the largest cylinder-valve wheels in 30% glass-filled nylon, which roughly doubles stiffness.
That choice carries a penalty you only price correctly if you know the whole family. The 33% glass-filled PA66 grade on our shelf reaches 186 MPa in tension but breaks at around 3% elongation — strong, and brittle. A sharp internal corner on a part in that grade is a crack initiation site you designed in yourself. One supplier holding the family makes that trade once, deliberately. Twelve suppliers holding one part each make it twelve times, by accident.
3. Failure modes stay solved
The numbers in this article come from our own production records at Weilin Plastic Co., Ltd rather than from a market survey, which is also why I can give you the failures.
Early in the handwheel program, parts left the machine dimensionally correct and went out of tolerance after cooling — shrinking more than the tool had been cut to allow. The cause was not the tool. It was drying: nylon that enters the barrel wet does not behave like nylon that does not. The fix was a longer drying time, and it has stayed fixed for a decade because it lives in that part’s process sheet rather than in somebody’s memory. Moisture uptake in these materials is measured to ASTM D570, and for unfilled PA6 it runs on the order of 1.5–1.8% in twenty-four hours of immersion — easily enough to move a dimension that matters.
Second example. The check valve seats we mold in acetal homopolymer keep shrinking after ejection: roughly 2.8–2.9% along the direction of flow and 2.1–2.4% across it, with the part still settling about twenty-four hours after it comes off the machine. So we hold parts for a day before dimensional inspection. A supplier who gauges at the machine and ships is gauging a part that no longer exists by the time it reaches your incoming inspection. Mold shrinkage is determined to ASTM D955; it is worth asking which value the tool was actually cut to.
Neither problem is exotic. Both cost a batch to find. You pay that tuition once per supplier, per material, per part — which is the real reason supplier churn is expensive on parts that look cheap.
4. One audit, one standard, one definition of yield
The administrative saving is what buyers usually cite to justify consolidation. It is real but smaller than the slide deck suggests: one supplier audit instead of six, one quality manual, one set of ISO 9001 certified production records to review.
The larger saving is comparability. When one factory holds a part family, yield becomes a signal instead of noise — 99% on the dental program, 98% on the handwheels, measured the same way, by the same people, over years. Six suppliers reporting six private definitions of yield give you six numbers you cannot add together.
Tolerance behaves the same way. We hold ±0.05 mm as standard and ±0.02 mm where the drawing genuinely demands it, and whether a feature lands in the second category is partly a tooling decision taken long before the first shot. Our rule of thumb, built up over three decades of family-factory production rather than from a textbook, is that each additional cavity in a mold costs roughly 5% of dimensional precision, and that where precision really matters you should not go past four cavities. That is a conversation worth having once, with one supplier, about an entire family — not twelve times, in twelve separate quotes.
5. What it costs you, honestly
Consolidation is not free, and any supplier telling you otherwise is selling.
Single-source exposure is the obvious cost, and it is not theoretical: a fire, a power curtailment, a shipping disruption or a simple falling-out now takes out a whole family rather than one part number. Price tension also drops. A supplier who knows the tooling is not going anywhere quotes differently from one who is defending the business every March.
The version that works, from what we see, is narrower than “consolidate.” It is one supplier per part family, with the families deliberately spread across more than one factory so that no single event stops a whole assembly, and with a second source kept qualified — actually qualified, with sampled parts on the shelf — on the two or three items that would stop your line tomorrow.
The question worth asking at re-quote
The useful question is not “who is cheapest this year.” It is: what does this supplier know about my part that a new one would have to re-learn, and what is that re-learning worth?
On a genuinely commodity part, the honest answer is often “nothing much” — and you should split that business and enjoy the savings. On the one-dollar part that decides whether your instrument reads correctly, seals, grips, or holds pressure, that knowledge is most of what you are buying. It just never appears on the quote.
Frank Lai is the second-generation owner of Weilin Plastic, an injection molding factory in Ningbo, China, running sixteen injection molding machines from 100 to 1,000 tons. The factory has supplied a European medical-gas equipment OEM for seventeen years and a North American dental consumables brand for twenty-nine, and holds ISO 9001 certified production.