Scrap Reason Codes That Actually Get Used

Most scrap reason codes get built in a conference room by a quality engineer, come out at 40 codes, and within a month more than half the scrap is logged as "other" or "operator error." The list is not wrong. It is just not a list anyone will use at 2 PM with a bad part in one hand and a running machine behind them.

Why 40-code lists die and 8-code lists live

An operator holding a scrapped part decides what to write in about five seconds. Nobody scans 40 lines in five seconds. They pick the first one that looks close, or they pick "other," and either way the data is guesswork by the end of the first week.

Eight codes fit on one line of a tally sheet. An operator can recite them after a week. When a code is easy to remember, it gets picked because it is right, not because it was near the top.

The test is simple: if the lead operator cannot say the list from memory, it is too long. Eight codes should cover about 90 percent of what gets scrapped. The last 10 percent is what "other" is for.

One 60-machine machining and assembly shop started with about 30 codes on a spreadsheet, watched the top five carry everything, and cut to eight. Five years later it is still eight.

Build the list with operators, not for them

Take two weeks of scrap tags, put them on a table, and sort them into piles with three operators and a lead. This takes about 45 minutes. The piles that form are your codes. Let the operators name them.

Name codes in operator words, not quality-manual words. "Tool wore out" beats "tool life exceeded." "Dropped it" beats "handling nonconformance." The check is whether a new hire on day three could pick the right code without asking.

Keep the "why" out of the codes. A code says what happened at the machine. The 3C says why. If a code tries to do both, it becomes a debate every time it is used.

An example 8-code list for a machining and assembly shop

1. Setup or offset: first piece was out, the dimension was wrong from the start.
2. Tool wear or breakage: the dimension drifted mid-run or the tool failed.
3. Material: porosity, hardness, wrong stock, bar out of spec.
4. Machine fault: spindle, coolant, chip jam, crash, anything the machine did.
5. Program: wrong program, wrong revision, wrong offset table loaded.
6. Handling damage: dropped, scratched, dinged between operations.
7. Assembly: missing or wrong component, torque, press fit, seal.
8. Inspection or gage: measured wrong, gage out of cal, passed here and failed downstream.

Notice what is not on the list: "operator error." Operator error is a pile, not a cause. The process is broken, it is not an operator issue, and a code that says otherwise teaches people to hide scrap. If the cause really was a person, the 3C will get there, and the countermeasure will be training or a work instruction, both of which are process fixes.

Your list will differ. A stamping shop's eight will not look like this. What matters is that it came from the tags on the table, not from a standard.

The "other" trap

Keep "other." Then watch it like a red metric.

Above 10 percent of scrap count in a month means the list is missing a code. Below 3 percent means it is doing its job.

Make "other" cost something: it requires a three-word note on the tag. "Bent on cart." "Wrong bar loaded." Review the notes monthly. If five of them say the same thing, that is your ninth code, and one of the existing eight has to leave to make room. Eight is the ceiling, not the starting point.

Reviewing the codes quarterly

Thirty minutes, once a quarter, same three operators if you can get them. Three questions:

Which code was under 3 percent all quarter? Merge it into a neighbor.

What is in "other"? Read the notes, not the count.

Which code has been the top code for two quarters with no 3C against it? That is not a code problem. That is a red box nobody owns, and the review should end with a name and a date.

Do not reuse numbers when a code retires. If code 4 goes away, the new code is 9, even if you only have eight active. Reusing a number makes last year's data lie.

Where paper works fine

A tally sheet at the machine with eight columns and a hash mark per part works well, and operators like it. The problem was never the sheet. The problem is Friday afternoon, when someone has to add up 17 sheets by hand, and the count for one cell gets keyed in wrong, and the top code for the month is off by one.

One honest note: ShopBoard won't hold your own list. Its scrap form has seven fixed categories, and Operator is one of them. If you want to see logged scrap tally itself into a Pareto by part and by category, the demo runs on sample data, no signup needed: Try the live demo →