Lesson 5 of 10
Quality, scrap and rework
The last of the three fractions, and the one plants report most confidently — usually at ninety-eight or ninety-nine per cent. It is also the one where a loss can be made to disappear simply by fixing the part and saying nothing.
Step one
Read the material
Three ideas. What counts as a good part, why rework is a loss even when the customer is happy, and why a defect is the most expensive thing a machine can make.
Concept one
Good means right the first time
Quality is good count divided by total count, and everything turns on the definition of good.
A good part is one that came out of the process correct the first time, with no touching up, no second pass, no adjustment by hand. That measure has a name outside TPM: first pass yield. Anything else is a loss, whatever happens to the part afterwards.
This fraction carries the two QUALITY losses from lesson one. Startup losses are the parts made while the process is coming right — warm-up, the ramp after a changeover, the trial pieces before the first-off is approved. Production defects are the ones made once the process is supposed to be stable.
The two behave differently and want different fixes. Startup loss is proportional to how often you change over, so it shrinks when changeovers get fewer or shorter — which ties it straight back to lesson three. Steady-state defects are proportional to how long you run, and they usually point at a condition: tooling, material, setting, wear.
A plant that only counts scrap at the end of the day will see one number and cannot tell those two apart. Counting them separately is most of the work.
Concept two
Rework is a loss even when the part ships
This is the point people argue with, so here is the case plainly.
A part comes out wrong. Somebody catches it, fixes it, and it goes to the customer. The customer is satisfied. Nothing was thrown away. Where is the loss?
The loss is that you paid for that part twice, and the second payment bought you nothing you did not already intend to have. The machine time, the material and the labour were all spent to produce one saleable unit; you spent them one and a half times. Meanwhile the rework itself consumed capacity that could have made a new part, so the cost is not just the fixing — it is the part that never got made instead.
There is a worse problem than the money. A shipped-units count cannot see rework at all. Scrap leaves a physical trace: a bin, a weight, a variance. Rework leaves nothing. The part ships, the total agrees, and the process fault that produced it is never investigated because on paper nothing went wrong.
That is why TPM insists rework counts against you. Not to be strict, but because a loss that leaves no trace is a loss that never gets fixed. The plants with the best quality figures are quite often the ones with the busiest rework benches.
Concept three
A defect has already paid for the other two fractions
Quality is usually the highest of the three numbers, so it looks like the least interesting place to work. Ninety-eight per cent against an availability of sixty-five — obviously the availability is the problem.
That reasoning is wrong, and this is the most useful idea in the lesson.
A defective part is not cheap because it is rare. By the time a part is finished and rejected, it has consumed run time that survived every breakdown and every changeover, and cycle time that survived every small stop and every bit of slow running. It sits at the very end of the chain. The defect wasted capacity that had already been paid for twice over.
So the real cost of one scrapped part is not its ideal cycle time. It is its ideal cycle time divided by your OEE. At an OEE of 50%, a part with a 30-second cycle consumed a full minute of the planned production time you actually had. Scrap a hundred of them and you have lost an hour and forty minutes, not fifty.
The same logic settles the sequencing question people ask constantly: should we fix quality or availability first? Fix quality first when a defect is discovered late, because those parts are carrying the whole chain. Fix availability first when the process rejects early, before much has been invested. Where the defect is caught matters more than how many there are.
Step two
Work it with your hands on it
One shift's output. Split the bad parts between scrap and rework and watch what a shipped-units count would have reported instead. The last row is concept three — what those parts really cost once the OEE around them is taken into account.
Hands on it, off the screen. Find the rework bench. Every plant has one, though it is rarely called that — it is a corner, a second bench, a person who is very good at saving parts. Stand there for an hour and count what arrives.
Then ask one question: where does that number get written down? In most plants the honest answer is nowhere. Those parts were made, fixed and shipped, and the quality figure reported upstairs never moved.
Step three
Prove you understood it
One question per idea, new numbers every time.
All three solid.
That is OEE complete — all three fractions, all six losses, and the ways each one can be made to look better than it is. The rest of the class is about fixing them. Lesson six starts with the pillar that changes who is responsible: Autonomous Maintenance.
Modern Total Productive Maintenance (TPM)
Lesson five of ten. This class is taught from the Total Productive Maintenance series — the eight pillars, the six big losses, and the roadmap into AI and predictive manufacturing. Written by our founder, Dr. Gene A Constant, and donated to the Foundation.
Read on Kindle The whole class AI + TPMGSU is a 501(c)(3) nonprofit and book sales are its only income. A purchase is not a tax-deductible donation. As an Amazon Associate, Global Sovereign University earns from qualifying purchases. This lesson is free and always will be. Global Sovereign University: Different by Design. Better by Mission.