How to reduce scrap and rework: a worked month from 12 reason codes

Step 1: set the counting rules, then read the month

This example is illustrative, not from a real plant. A machining cell makes aluminium valve housings: a casting is bought in, machined, deburred and washed, a bushing is pressed in, and each housing gets a final inspection and a pressure test. The month is 20 shifts and 6,000 housings started. Every cost below is made up for the example, so replace them with your own.

Two rules are written down before the month starts. First, scrap and rework are counted apart. Scrap is a housing that cannot be made good and is thrown away. Rework is a housing made good by extra work. They cost differently (the whole housing against some minutes) and they have different fixes, which is why one combined defect count hides the answer. The definitions are ours, in plain words; the scrap rate glossary entry and the first pass yield calculator draw the same line. Second, every affected housing gets one code, the one that decided its fate: a housing reworked and then scrapped is scrap. The same codes and the same rule apply every month, or next month will not compare with this one.

That table is the start of every cost sum below. A housing scrapped at final test has cost $29, not the $14 of its casting. Here is the month: 12 reason codes, six for scrap and six for rework, with the step where the defect was made and the step where it was found.

Step 3: put a cost on each

Counts say how often, cost says how much it hurt. Two formulas cover the whole month. Scrap cost is the housings times what one housing had cost by the step where it was found. Rework cost is the housings times the extra minutes each times the labor rate.

• Scrap, S1 bore oversize: 74 housings × $29 (cost up to final test) = $2,146. Counting only the $14 casting would give $1,036 and miss $1,110 of machining and labor already in the part.
• Rework, R1 burr at the cross hole: 312 housings × 4 minutes × $1 a minute = $1,248.
• Month: scrap $6,249 and rework $3,375, together $9,624.

Three in four affected housings were reworked, but two dollars in three went on scrap. The rework still took 3,375 minutes of extra labor, just over seven eight-hour shifts of one person, and that time was not spent making new housings. A report that only counts defects would send the team after rework first. Costing the codes first is what changes the order.

This is the internal failure part of the cost of quality: scrap, rework and the like, found before the product reaches the customer. The COPQ calculator and the COPQ worksheet add the other three categories (prevention, appraisal and external failure).

Step 5: where each defect is made against where it is found

Every defect has two places: the step that made it and the step that found it. The cost of scrap grows with the distance between them, because each step in between adds work to a housing that is already bad. For the scrap codes where the origin is known, the table prices that distance.

Two codes carry all of it. The oversize bores (S1) are made at machining but found at final test, because the bore gauge at the machine is checked one housing in 50 and the rest wait for the end of the line. The porosity (S2) is made at the foundry and found only when the cut opens it. Moving the check changes the cost without making a single extra good part: gauge the bore more often at the machine, and check the first cut of each casting lot at receiving.

Shigeo Shingo's three kinds of inspection, as a university paper summarises them, name the pattern. Judgement inspection compares a finished product with the standard at the end of the process and rejects what fails; it does not prevent the defect. Informative inspection feeds what it finds back to the process, by statistical checks, by the next process or by the operator checking their own work. Source inspection catches the error and gives feedback before the next step, so the error does not become a defect, and it is done by people or by error-proofing devices. S1 and S2 are being found by judgement inspection. The aim is to move each check toward the source.

Rework is not in this table, because its cost is the minutes whenever it is found. Finding a burr late does not change the 4 minutes; it means more housings carry the burr by the time anyone sees it, and a few may slip through. The table cannot price that, so it is listed as a limit below.

Step 7: contain this week, fix the cause, then remove the containment

Containment protects the next process and the customer while the cause is still there: sorting suspect stock, adding a check, holding a lot. It does not remove the cause. Root cause work does, and when it works the containment is taken away. Plan both for each of the two top codes, with the date the containment ends.

Containment costs money every month it stays. For S1, 6,000 housings at 1 in 50 is 120 gauge checks a month, and at 1 in 10 is 600: 480 more. At an assumed one minute a check and $1 a minute that is $480 a month, about the same as the $444 that finding S1 late cost. A stopgap that costs as much as the problem is a reason to hurry the root cause work, not a fix.

The order of work in the cost of quality primer read for this guide is the same: find the largest contributors, use a Pareto analysis to find the main reasons for each, ask why until the root cause of each is clear, start corrective action, then keep monitoring the cost element to confirm it went down. The monitoring is not optional, and it is the next step.

What this method cannot tell you

• The counts are only as good as the codes. If two people use R2 and R4 for the same fault, the Pareto is wrong however neat the arithmetic. Write each code down with a boundary sample, train the counters, and check a few by eye every week.
• Costs here use a flat labor rate and an average of extra minutes. Time a few real reworks and use your own cost per step from your accounts; do not reuse these figures.
• Scrap that is sold or recycled brings some money back. This example does not subtract it. Do, if your records show it, and say so.
• Hidden costs are real and are not in the example: inspection time, expediting, extra material handling, overtime, the capacity taken up by rework, and the risk that a defect found late slips through. One university paper on the cost of quality names the hidden factory, where a worker keeps correcting a defect so that a good part goes downstream and the cause is never looked at; another is a scrap allowance built into a cost standard, which keeps scrap from being counted as a cost of its own.
• Defects that reach the customer are external failure costs and are not in this month, which only counts what was found inside the plant.
• One month is one sample. Small counts are noisy, and a fix that shows after one month may be luck. That is why the plan reads 30 and 90 days.
• If a defect could be put in either category (say, an item that is reworked but then downgraded), pick one rule and keep it. Consistency matters more than which choice was made.