Your Supplier Booked the Savings, You Booked the Leakers

A defect that arrived four months after the decision that caused it

An edible-fats processor spent this summer chasing leakers. Their term covers everything from an actual puncture to an improper liner fold to a case that only got stained because something next to it on the pallet ruptured. Three lines carry the problem: two producing fifty pound cubes of tallow, lard and oil-fry, and one filling thirty five pound jibs, a blow-molded bottle with a screw cap seated inside a corrugated case.

They did the work you would want them to do. Complaint data converted into pounds, plotted by month, showing a real step up through June and July of 2026 that a hot summer does not explain. A fishbone with causes that reach well past packaging: palletizing alignment, nitrogen levels, transport directionality, case integrity across two different suppliers. A Pareto by establishment number that pinned most of the cube leakers on the line where the liner fold is done by hand rather than the one where it is automated. Standard work written for that fold in late July. Board stock that failed compression and crush testing pulled out. Pallet configuration changed so the rigid face of the case sits outboard instead of the open side.

Buried in the middle of that meeting, the operations lead said the thing that reframes all of it. In the last four or five months, the jib maker reduced the plastic volume in the bottle. Separately, the case supplier pulled a lot of the fluting out and the paper volume changed. Both changes date to February or March. The bottle supplier had sent notice back in November. The processor tested at the time by sending pallets around and saw nothing.

The arbitrage is in the interval, not the material

Two clocks run at different speeds here, and the gap between them is where the money moves.

The supplier's clock is short. Material comes out of the spec and the savings book on the next shipment. The buyer's clock is long, and it is not one number. On the cubes, a complaint arrives inside about a week, and past six weeks the claim gets denied. On the jibs, a complaint landing in July can trace to production in February. The failure needs multiple transport legs, a stay in an outside warehouse, a leg to a distribution center, then weeks sitting in the customer's warehouse before the case gives up.

Now put the qualification against that. The processor validated the new bottle by sending pallets around. That is a round trip measured in days, run against a failure mode that takes four to six months to express. The test was not weak. It was shorter than the physics it was supposed to catch, which is a different and worse problem, because a short test returns a clean result and a clean result closes the file.

The second clock is the one nobody prices. When a case gets stained, the box is finished and the product is reworked entirely, and their rework rate moved with the leaker rate one for one. That is not a scrap number. Rework consumes the same line hours as first-pass production, on lines whose output was already promised. A single ruptured case stains a pallet, so a one-case defect converts into a pallet-level claim on arrival. The material takeout on the supplier side is a clean, immediate, provable number. Your side of the same transaction shows up as three unrelated-looking line items, none of which anyone ever adds together.

Notice, too, how the second change surfaced. The processor found the fluting reduction themselves, because their shrink wrap runs at a low tensile compared to industry and the cardboard started crushing. Nobody sent a letter about that one. The tell was a downstream symptom in a completely different operation.

Put every moving variable in a field

The countermeasure is not vigilance. It is structure, and it belongs in three places.

Make the bid sheet carry every variable that can move. On a resin-based RFP for a workspace-products manufacturer, the design decision was to publish one specific resin level and require every supplier to bid the current spec against that baseline, with movement handled through an agreed escalator and de-escalator rather than through each supplier's own assumption. Alternates were not banned, they were given their own fields: current price for the current spec, then a separate alternate price with a separate alternate spec proposal, filled in where the buyer can compare it. The rule underneath is the one I would hand anyone building a bid field: anything that can move gets a field, because anything left in the notes is a blank someone else fills in against you. The same review turned up a label the buyer buys, ships to the supplier, and pays to have applied hot, where it falls off roughly half the time, and it carries the product liability warning. It was never a spec. It was a drawing, with no adhesive and no dimension, which is why no supplier could price it and why it stayed out of round one.

Treat a change notification as a change order with a clock attached. A November letter saying the bottle will be formed with less plastic is not information, it is a request to re-qualify. Set the re-qualification window off your own complaint history, not off convenience: if your longest observed complaint lag on that package is five months, a two-week pallet loop proves nothing and should not be allowed to close the item. Where the calendar will not permit that, say so explicitly and run the change as a monitored risk with a named owner and a date-coded sample plan, which is at least honest.

Refuse savings that are conditional on somebody else's willingness to be interrupted. In a supply agreement negotiation this week, the supplier's structure was a percentage now plus another four points if a plant trial passed, with the fallback written so the conditional tier looked routine. It is not routine. Every conditional point is a claim on plant time that no plant wants to spend, and lowering the conditional percentage makes the trial less attractive internally, which makes the money less collectible, not more. The counter was to strike the conditional clause and take eleven percent across the board up front, backdated to the first of the month, with trials offered only on the low-risk items the supplier itself had already flagged as easy. If the supplier needs the headline number to stay at nine, let them put the difference in the rebate tier. Where it sits is their problem. Whether it is conditional is yours.

What a well-run version reads like

Every supplier material change enters as a change order with a named owner and a re-qualification window at least as long as the longest complaint lag on record for that package. Complaint pounds trace to line, establishment and date code inside a shift, not inside a quarter. Rework hours appear on the capacity plan as consumed line time, and they are reviewed next to first-pass output rather than in a scrap report. The bid sheet has a field for every variable that can move, and the count of blanks a supplier can fill on their own terms is zero. Conditional savings are a small, stated share of the contracted number, and someone can name the plant hours each conditional point requires.

The bottle supplier's takeout was real from the first shipment in February. The buyer's version of the same decision arrived in June and July as reworked pallets, stained cases and denied claims. Both numbers were true. Neither one was ever in the same model as the other.