Scrap Rate: How to Calculate It by Units and Cost

Scrap rate is the share of units a process discards as defective beyond economical repair. Measured by units it is scrapped units ÷ units produced; measured by cost it is scrap cost ÷ total production cost. The two can point in very different directions, because a unit scrapped at the last operation carries far more accumulated cost than one scrapped at the first.

Most plants report scrap rate as a single tidy percentage, "we run 2% scrap", and then wonder why the finance number and the floor number never agree. The gap is almost always about where the scrap happens. Two units per hundred thrown away at receiving is a rounding error; two units per hundred thrown away after final test is a budget crisis, because by then the part has soaked up material, machine time, labor, and every operation in between. This post covers both ways to calculate scrap rate, why the cost version is the one that matters, and how to stop booking your most expensive scrap at the worst possible place.

What is scrap rate?

Scrap rate measures irrecoverable loss. Scrap is defective product or material that cannot be repaired, used, or sold, the American Society for Quality classifies it, along with rework, as an internal failure cost one of the four categories in the cost of quality. The distinction from rework matters: rework is a unit brought back to spec with extra work; scrap is a unit written off entirely. Scrap rate counts only the write-offs.

The simplest form is a unit count: of everything the process made, what fraction went to the dumpster instead of the dock. That number is easy to gather and easy to misread, because it treats a scrapped casting and a scrapped finished assembly as equal losses. They are not. A finished assembly represents every dollar the casting cost plus everything spent turning it into an assembly. Scrap rate by units hides that; scrap rate by cost exposes it.

One definitional choice trips people up: units produced versus units started. If the denominator is good units produced, scrap sits on top of a smaller base and the rate reads slightly higher; if it is units started into the process, scrap is measured against everything that entered. Neither is wrong, but they are not interchangeable, and a plant that switches between them mid-year will see a "trend" that is really just a definition change. Pick one, write it down, and apply it the same way to every line you compare, the same discipline that keeps any plant KPI honest.

How do you calculate scrap rate, by units and by cost?

By units, scrap rate = units scrapped ÷ units produced (or started) × 100. By cost, scrap rate = value of scrapped units ÷ total production cost × 100, where each scrapped unit is valued at the cost accumulated up to the operation that scrapped it. A worked example with hypothetical numbers shows why the two diverge:

By units, scrap rate is a flat 2.0%. By cost, the picture is upside down from what the unit count suggests: the 40 finished assemblies scrapped at final test ($1,360) cost more than the 120 raw units scrapped at receiving ($1,200), even though there are a third as many of them. If you drive improvement off the unit count, you will spend your energy on the 120-unit receiving problem and ignore the 40-unit final-test problem that is quietly costing more. The cost-weighted rate points you at the money.

Why is scrap booked at the last operation the most expensive kind?

Because value accumulates and cannot be recovered. Every operation a part passes through pours in material, machine time, energy, and labor, and when that part is scrapped, all of it is written off at once. A defect that was present in the raw material but not caught until final test doesn't just cost the final test, it costs the entire journey the part took while the defect rode along undetected. The plant paid to add value to something it was always going to throw away.

This is also why scrap rate and first pass yield belong on the same page. FPY tells you which step first produced the defect; the cost-weighted scrap rate tells you how expensive it became by the time someone finally threw the part away. The goal is to move detection upstream so the gap between "defect created" and "defect caught" shrinks, every operation you close that gap by is an operation of accumulated cost you stop writing off.

By the numbers. Scrap is not a small line item. The American Society for Quality reports the total cost of quality commonly runs 15–20% of sales revenue and scrap sits squarely inside the internal-failure category driving that figure. ASQ's guidance on how to reduce scrap and rework costs treats both as prime targets precisely because they are pure loss, money already spent on product that will never generate revenue.

What causes a high scrap rate?

Scrap almost never has one cause, but the drivers fall into a short, familiar set, and naming which one dominates is how you stop guessing. The usual sources:

• Incoming material variation. Out-of-spec raw material seeds defects that surface operations later, often at final test, which is why supplier quality problems show up as your most expensive scrap. Catching them at receiving is the cheapest scrap you will ever avoid.
• Process drift. A machine slowly wandering off setpoint produces good parts until it doesn't. Without statistical process control watching the driving variable, the first sign is a batch of scrap, not a warning.
• Startup and changeover. The first parts after a changeover are prime scrap candidates, purge material, warm-up rejects, first-piece failures, which is why startup rejects are their own category in the six big losses.
• Handling and setup error. Dropped parts, wrong tooling, a misload. These are the failure modes error-proofing was invented for, and the ones most responsive to a well-placed poka-yoke.

The reason the causes matter is that each points to a different fix, and a single blended scrap percentage hides which one you have. A plant chasing "scrap" in the abstract tends to exhort operators to be careful; a plant that knows 60% of its scrap dollars come from material drift at one operation goes and fixes that operation. The cost-weighted, coded scrap number is what makes that difference possible.

How do you measure and reduce scrap rate?

Count scrap where it happens, value it by accumulated cost, and attack the expensive end first. The sequence:

1. Define scrap versus rework in writing. A unit written off is scrap; a unit brought back to spec is rework. Fuzzy rules let expensive scrap hide as "rework that didn't work out," so the categories have to be clean before the numbers mean anything, see scrap vs rework for the decision itself.
2. Book scrap at the operation that scrapped it, not in a single end-of-month bucket. You cannot cost-weight scrap you cannot locate.
3. Attach a cost to each scrap point using the accumulated value up to that operation. This turns a unit count into a money map and reveals where the losses actually concentrate.
4. Attach a defect code to every scrapped unit. A scrap number without reasons ranks nothing; codes feeding defect tracking and a Pareto turn it into a target list.
5. Work the highest-cost scrap point first, with root-cause tools and detection moved upstream, an inspection or poka-yoke before value concentrates catches the defect while it is still cheap to throw away.
6. Re-measure both rates weekly. Track scrap by units and by cost side by side; a falling cost-weighted rate is the one that shows up in the P&L.

How does scrap rate connect to OEE and cost of quality?

Scrap is the hard core of the Quality factor in OEE and one of the six big losses specifically the reject losses, split between startup scrap after a changeover and production scrap during steady running. Every scrapped unit is capacity spent for nothing: the machine time that made it is gone, so scrap quietly caps real throughput as surely as a breakdown does. And because scrap is an internal-failure cost, it is a direct input to the plant's cost of quality the reason a falling scrap rate shows up not just on the quality board but on the income statement.

The honest version of the metric depends on scrap being logged where and when it happens, with a cost and a reason attached, and that is exactly what paper loses. Scrap swept into a monthly total, uncosted and uncoded, can tell you the rate is high but never where the money is or why. Plants that capture scrap at the station, valued and coded, the way Harmony turns paper quality logs into live records feeding root-cause analysis (see the platform or the CLS field story), get a scrap rate they can actually act on: by units, by cost, by operation, and by cause, which is the difference between knowing you have a scrap problem and knowing where to spend the next hour fixing it.