Defect

A defect is the smallest concrete unit of failure in a manufacturing system. One bracket that came out .003 over tolerance. One bottle with a chipped neck. One assembly with the wrong fastener. Every shop produces defects. The question lean asks is not "do you have defects" but "are you learning from them or burying them." Most shops bury them. The scrap bin gets emptied at the end of the shift and the same defect shows up next week.

"Every defect is a signal. Throw it in the bin and you throw the signal away."

How a defect actually gets made

A defect is the visible outcome of an invisible process variation. Something upstream changed. The tool wore. The material came in a hardness point above the last lot. The fixture shifted by half a thou. The operator was trained by someone who picked up a shortcut. The machine drifted out of warm-up spec. Any of these can produce a part that fails the requirement. The defect is the trailing indicator; the variation is the leading one.

Defects come in three flavors that shops tend to handle differently.

• Functional defects are parts that do not do what they are supposed to. A switch that does not switch. A valve that leaks. These are usually the most expensive because they cause customer returns and field failures.
• Dimensional defects are parts where a measurement falls outside tolerance. They are easy to detect with a gauge and often easy to root-cause back to a specific tool or setup parameter.
• Cosmetic defects are parts that look wrong. Scratches, scuffs, color shifts, mold flash. They are the hardest to standardize because "looks bad" is judgment-dependent and the cost varies by customer.

The lean response to all three is the same. Treat the defect as data. Tag it, log what happened, and look for the pattern. The defects that recur are telling you where the process is unstable. The ones that show up once are usually random and not worth chasing.

Where defects show up on the shop floor

Picture a 20-person plastics injection molding shop running closures for a beverage brand. The shop produces about 80,000 closures a day across three presses. Returns from the customer for cosmetic defects (slight color streaks, faint sink marks) are running about 2.5 percent. The shop owner thinks the molders are getting sloppy.

A defect investigation looks at the data and finds something else. The defects cluster on Press 2, on Monday mornings, after the weekend. Press 2 cools down on Saturday and Sunday, and the morning startup runs hot for the first 45 minutes before steady state. The defects are not a worker problem. They are a process problem: the startup procedure is producing borderline parts for the first ninety shots, and nobody has been pulling and scrapping them because the shift schedule treats the startup window as productive time. Fixing the startup procedure cuts the cosmetic return rate from 2.5 to under 0.5 percent inside a month.

Common mistakes with defects

• Blaming the operator. Operators rarely cause defects on their own. Process inputs do. Treating the operator as the cause kills the learning loop.
• Untagged scrap. A scrap bin without tags is data thrown away. Even a one-line note on each part turns the bin into a source of insight.
• Chasing every defect equally. Eighty percent of the cost usually comes from a handful of failure modes. Going after every defect at once dilutes the fix.
• Skipping the post-mortem. A defect that gets fixed without anyone asking why it happened will come back next week. The cost of the post-mortem is small compared to the cost of the recurrence.

Defects and related Lean tools

A defect is one observed instance of the broader waste category lean calls defects. The closest QMS sibling is nonconformance, which covers product and process failures together. The two best measurements of how often defects happen are first-pass yield, the percent of parts that complete every operation without rework, and rework rate, the percent of parts that need a second pass to meet spec. Together, these three frames give a shop the vocabulary to actually talk about defects without arguing about whether it is a "quality problem" or a "process problem."