Entry 0200·November 20, 2026·Sourcing·Specifications & Supplier Markets

The Downgauge That Passed Its Test, Then Failed

A packaging supplier tells you it is reducing the plastic in a molded container.
Truth · observed pattern

The change you approved, and the one you did not

A packaging supplier tells you it is reducing the plastic in a molded container. The notice arrives months before the change takes effect, which is exactly how a good supplier behaves. So you test it. You build pallets with the new material, move them around, look at them, and find nothing. You approve the change. The unit price goes down and the variance shows up in the next purchasing report.

Two quarters later the leak rate steps up. Not on one line, on the lines that share a handling path. Product is escaping the package. Some of it is caught in your own warehouse before it ships. Some of it is found at a customer's dock weeks after arrival, and when one container ruptures on a pallet the stain condemns the cases around it, so a single defect becomes a multi-case complaint. The team builds a cause list and it runs to a dozen entries: a manual liner fold that varies by operator, palletizer alignment, case strength variation between suppliers, nitrogen levels and the overfill they create, which way the fold faces when a forklift takes the pallet up a slope. Every one of those is plausible. One of them changed on a date somebody can name.

There is also a second change nobody announced. The corrugated supplier pulled flute out of the board and shifted the paper weight. The plant found that one on its own, because the shrink wrapper's tension was reading low against its normal setting and the cases were crushing under the film.

Why a passed test is not a qualified change

The test was not wrong. It was short.

A material qualification measures a failure path, and that path has a length. Its length is the number of handling legs, the dwell time at each one, the temperature cycles, and the stack height chosen by whoever is holding the pallet at the time. Building pallets and moving them around reproduces one leg under your own rules. The real path runs the plant to an outside warehouse, the warehouse to a distribution center or a third party, the third party to a customer, and then sits in the customer's own warehouse. Somewhere in that chain a facility double stacks pallets because its racking says it can, even though you never double stack yours. A package that survives one controlled leg and fails on the fourth uncontrolled one is not a package that passed; it is a package that was asked an easier question.

That explains the delay. It does not explain why the delay was expensive, and the second half of the mechanism is the one that costs real money.

The loss never enters scrap. Scrap is the category for what you throw away before it leaves the building. This leaves the building. A case with a stained panel caught in the warehouse becomes rework, and rework reports as labor minutes per thousand units in an operations report. A container that reaches the customer becomes a credit, and a credit reports in the commercial P&L. Neither one lands as yield loss on a packaging line, so the packaging cost per unit, which went down when the material got thinner, stays down on paper. The cost structure reads improved while the margin gets worse, and both statements are supported by their own numbers.

Then your reporting censors the evidence. If claims older than about 6 weeks are denied as a matter of policy, and the failure incubates for longer than 6 weeks, the complaint rate you review is not the failure rate. It is the failure rate truncated by your own claim window. On a format whose production-to-complaint lag can run past a full quarter, you are reading a number that structurally cannot show you the oldest and worst cohort.

What to require before you accept a material change

The decision in front of you is not whether to take the saving. It is what the qualification has to cover before the saving is yours to book. Four requirements make that concrete.

Exposure parity. The test has to carry product through every irreversible step and every handling leg the real shipment carries, and hold it at least as long as the longest field dwell. If a third party in your network stacks two high, the test stacks two high.

Detection before the irreversible step. This is the requirement most qualification plans miss, and it is the one that pays for itself immediately. In a process with a thermal step, hold dedicated racks for 24 to 48 hours before that step and count the defects there. At that point product can still be recovered and the line can still be adjusted. Past the irreversible step the same defect is direct product loss, and the only instrument still reading your material change is the customer.

A counted rate, not a verdict. Dedicate volume, one rack the first day and several the next, and report a defect rate per unit of product run. A qualification that returns "looked fine" cannot be compared to anything later, which is why the original approval survived contact with the field for two quarters.

An aligned record. Keep every supplier change notice with its effective date on file before first receipt, and read complaint data against production date rather than claim date. Then make one person responsible for reading the packaging price variance and the rework hours for the same format on the same page, because the saving and the cost currently live in two reports that never meet.

My rule is narrow and I apply it without exception: a material change is not qualified until a test has carried product through every irreversible step and every handling leg the real shipment carries, and has produced a counted defect rate at a point where the product can still be recovered. A supplier notice is information. It is not qualification, and accepting it as qualification is how a cost reduction becomes a working capital problem and a customer relationship problem at the same time.

Write that protocol so your own team can run it after the outside help leaves. A qualification standard that only works while a consultant is in the building is not a standard; it is a project.

What a well-run material change looks like

Every inbound material change has a notice with an effective date in a file, and the first receipt under the new specification is identifiable by lot. Qualification holds dedicated volume past the longest known field dwell, through every irreversible step, and closes with a counted defect rate rather than an opinion. Defect data is plotted against production date, and the claim window is longer than the longest known incubation period, or the gap is written on the face of the report. Packaging price variance and rework hours for the same format are reviewed together, by one owner, monthly. When a supplier changes a specification without telling you, an in-plant process reading finds it before a customer does.

The next test

Take the material change you are closest to approving right now and ask one question about its qualification plan: at what point in the process does this test count a defect, and can the product still be recovered there? If the answer is that the count happens after the product ships, you do not have a qualification plan. You have a notice from your supplier and a price you like.

Published November 20, 2026
Related reading in Specifications & Supplier Markets