Three Changes in One Week: Containing a Defect Rise Without Stopping the Line
Monthly defect rate on a production line was steady at 1.8 percent for six months, then rose to 3.1 percent over the last two weeks across roughly 40,000 units. Three things changed in the same week: a new incoming lot of raw material from the same supplier, a move from 8-hour to 12-hour shift rotations, and a seasonal humidity rise from 35 to 58 percent. The line produces about 2,000 units per shift. Stopping the line costs about 28,000 dollars per shift. Scrap cost is 6 dollars per unit caught in-house, but a defect that escapes and fails downstream costs about 220 dollars, and the existing inspection catches only about 60 percent. Your operations manager wants to stop the line, and your finance partner wants to keep running and sort later.
1. All three candidates are plausible. Which do you test first, and what is the cheapest test that discriminates between them?
Read the worked answer
The cheapest and most discriminating move is not an experiment at all: stratify the existing two weeks of defect data by hour of shift, by day, and by material lot. This costs nothing and separates the hypotheses sharply.
If defects cluster in hours 9 through 12 of the rotation, human factors are implicated, because that pattern cannot be produced by humidity or by raw material. If they track the new material lot's run dates, material is implicated regardless of hour. If they rise monotonically with humidity across both shifts and both lots, humidity is implicated. Run the stratification before spending anything.
Only if the data are too sparse to stratify do you run a controlled trial, and even then the cheapest design is a short material swap on one line at normal conditions, not a full stop. Jumping to a line stop tests nothing; it produces a clean period whose improvement you cannot attribute to any of the three causes.
2. Make the containment decision. Do the arithmetic and state what to do in the next 48 hours.
Read the worked answer
Per shift, the excess defect rate is 1.3 percent of 2,000 units, about 26 units. Caught in-house, that is roughly 156 dollars of scrap; escaped, the arithmetic worsens, since 40 percent of 26 units, about 10 units, at 220 dollars is roughly 2,200 dollars. Total exposure is a few thousand dollars per shift. Against a 28,000 dollar per shift stop, stopping the line is not justified by direct cost alone, and the operations manager's instinct, though understandable, is expensive relative to the harm.
The finance partner is also wrong to propose sorting later, which defers the escape cost and multiplies it. The middle path is containment at the suspect operation: add inspection there, targeting the 40 percent escape gap, at roughly 1,400 dollars per shift, and quarantine the suspect material lot. Keep running, contain locally, stratify the data today, and decide on a root cause within 48 hours. Cheap containment buys the time the investigation needs.
3. What would change your decision to keep running? Name the specific condition that flips it.
Read the worked answer
The economics flip if the defect touches a characteristic where the consequence is not a 220-dollar rework but a safety, regulatory, or warranty event. Nothing in the cost figures captures that, and neither manager framed the question that way, which is the real gap in the discussion. The first question is whether this defect is on a critical characteristic, not whether 28,000 dollars exceeds a few thousand.
Operationally, three conditions would justify stopping: stratification showing defects concentrated in a narrow window you cannot inspect fast enough; evidence that the escape rate is materially worse than 40 percent, for example a downstream failure in a customer return; or inability to staff the added inspection within one shift.
State the flip condition in advance and in writing. A containment decision without a pre-agreed exit condition tends to be re-argued every shift on whoever's preference is loudest.