Scrap Rate: Formula, Benchmarks & the Honest Number

What is scrap rate?

Scrap rate is the proportion of parts produced that fail to meet specification and are discarded as unsellable. In its simplest form it is a ratio — scrap quantity divided by total quantity produced — expressed as a percentage. Every plant manager knows the number. Very few plants know the right number.

Here is the sentence that changes how most manufacturing engineers look at this metric after they first hear it: the scrap rate reported by your ERP is almost always 30–50 % lower than the actual quality loss in the plant. The gap is not the result of anyone lying. It is the result of how scrap gets recorded in practice — and once you understand the recording mechanism, you understand why reducing the number printed on the monthly report is a much smaller achievement than it looks.

The scrap-metric family — and why they're not interchangeable

Before getting to the measurement-honesty problem, a quick disambiguation. "Scrap rate" circulates alongside four related metrics that are routinely confused. Each measures something different; treating them as synonyms is how quality conversations go in circles.

The distinction that matters most commercially is scrap rate vs. First-Pass Yield. A plant can have a scrap rate of 0.5 % — sounds excellent — and an FPY of 82 %. The 17.5 % gap is rework. Rework consumes the same material, the same labour, and the same machine-hours as scrap does, but doesn't show up in the scrap number because the parts eventually shipped. For OEE and capacity purposes, rework is a quality loss. For the ERP scrap report, it is invisible. Six Sigma programmes rightly focus on FPY for exactly this reason: it is the honest version of what scrap rate pretends to measure.

How scrap rate is calculated — and where the numbers start lying

The textbook formula is unambiguous:

Scrap Rate (%) = (Scrap Quantity ÷ Total Production Quantity) × 100

The arithmetic is trivial. The problem is what gets counted on each side of the division. After running quality programmes in automotive plants in China, Mexico, the US, Tunisia, Macedonia, France and Russia — and watching scrap numbers reported up the chain at each of them — the pattern of invisible exclusions is remarkably consistent:

• Start-up scrap is booked to a separate cost centre. The first 50–200 parts after every changeover are typically off-spec while the process stabilises. In most plants these do not appear in the scrap rate; they are charged to "set-up waste" as a process cost. Real quality loss, wrong accounting bucket.
• Station-level rework is not recorded. An operator at a press notices a dimensional defect, pushes the part back one step, and the station corrects it. The ERP sees a good part come out; the quality system sees nothing.
• Sorting loss is not always included. 100 % inspection pulls out defective parts before shipment. In some plants these are booked as scrap; in others the cost is absorbed into inspection labour and the scrap rate stays clean.
• Downgrade is not scrap. Parts that fail Grade A specification but meet Grade B are sold at lower margin. That is revenue loss and quality loss, but the scrap rate never sees it.
• Destructive testing and samples are excluded. Reasonable — they were never meant to ship. But in some plants the sample quantity grows quietly until it is a meaningful fraction of output.

Add these up honestly and a plant that reports 0.8 % scrap is typically carrying 2–4 % real quality loss. That is not a small correction. In a plant producing 10,000 parts per day at €8 each, the difference between 0.8 % and 3 % is €1,760 per day of invisible loss — about €440,000 per year on a single production line. Multiply by the number of lines. This is why the first OEE number an honest measurement system produces is usually 15–20 percentage points lower than the one reported on paper — and why that lower number is the more useful one.

The true cost of a defective part

The second place scrap-rate conversations go wrong is cost calculation. "The part cost us €8 to make, so scrapping it cost €8." That is the direct material and labour cost, and it is a minority of the real cost in almost every case I have seen.

• Capacity consumed. Every scrap part consumed the same machine-time as a good part. If the plant is capacity-constrained — which it almost always is in some bottleneck — the lost capacity has an opportunity cost equal to the contribution margin of the sale that could not be made. In a tight market, this frequently exceeds the direct cost of the part itself. See the production capacity article for the arithmetic of this bridge.
• Downstream handling. The part got inspected, routed to the scrap bin, recorded, segregated, and eventually disposed of. Each step has labour attached.
• Quality-system overhead. Root-cause analysis, 8D reports, customer complaints if the defect escaped — all of it is driven by scrap events and is typically charged to indirect quality labour rather than to the specific defect.
• Warranty and field cost. For the fraction of defects that escape — the DPPM number — the cost multiplier jumps by 10–100× versus catching the defect in the plant. This is the well-known 1:10:100 rule of quality economics.

A useful rule of thumb for internal business-case work: the true cost of a scrap part is 2.5–4× its direct material-plus-labour cost when the plant is capacity-constrained, and 1.5–2× when it is not. Any scrap-reduction programme costed against direct material alone is understating its payback by a factor of 2 or more.

Benchmarks — what "good" actually looks like

Scrap rate varies by two orders of magnitude across industries, which means cross-industry benchmarking is meaningless. The table below gives rough order-of-magnitude ranges from plants I have worked in or audited. These are real-world numbers, not textbook ideals.

The numbers should be read carefully. A 1 % scrap rate in stamping is not the same as a 1 % scrap rate in casting, because the underlying process capability is different. Compare within your industry and within your process, not across.

Reducing scrap rate — the hierarchy that actually works

Most scrap-reduction programmes stop at the first step of what should be a three-step hierarchy. That is why they produce a 10–20 % reduction and then plateau.

1. Detect faster. Shorten the time between defect creation and defect detection. In-line inspection, SPC, automatic alarm classification when process parameters drift. The plants we work with typically recover the first 30–40 % of their scrap-reduction potential here. The Neoperl implementation — 15 % scrap reduction, 8 % higher availability — was mostly at this level: PLC alarms correlated with quality defects through automatic process-data capture, which made the signature of an impending defect visible before the defect happened.
2. Prevent. Move from detection to process control: SPC on the process parameters that drive the defect, closed-loop intervention before the part is made. This is where Six Sigma and SPC earn their reputation. Typical further gain: another 30–40 % reduction on top of the detection work.
3. Design out. The defect cannot occur because the design, tolerancing, or fixture makes it geometrically impossible. DfM, Poka-Yoke, tolerance-stack analysis. This is the domain of engineering change, not of production floor improvement, and it produces the largest and most durable gains — but also requires the longest time and the most organisational will.

Plants that skip past level 1 to invest directly in level 2 usually discover that they didn't have enough process data to make level 2 work. Plants that skip past level 2 to demand level 3 changes usually discover that engineering has no idea which defects are actually occurring, because level 1 was never properly instrumented. The hierarchy is a hierarchy for a reason.

FAQ

What's the relationship between scrap rate and OEE?
Scrap rate directly drives the Quality factor of OEE. If you produce 10,000 parts and 300 are scrap, your quality factor is 97 %. The connection is mechanical, but watch the definitional trap: OEE quality typically includes rework as a loss, whereas scrap rate typically does not. OEE is therefore the more honest number. See the OEE article for the full bridge.

What is considered an acceptable scrap rate?
Industry-dependent, as the benchmark table shows. More useful than absolute targets: a plant that has never measured automatically should expect to discover the real rate is 1.5–2× what the ERP reports, and should target a 30–50 % reduction over 12 months from the honest baseline. Chasing a number that was wrong to begin with produces theatrical improvements and no real gain.

Why is rework not counted as scrap?
Accounting convention. The part eventually shipped, so revenue was preserved and direct material was not wasted. But rework consumed labour, capacity, and quality system overhead — which is why quality-focused metrics (FPY, OEE quality factor) include it and pure scrap rate excludes it. For internal improvement, use FPY. For external reporting, use scrap rate and explain the footnote.

How often should scrap rate be reviewed?
Daily at line level, weekly at plant level, monthly at plant-manager level, with Pareto analysis of top defect codes. Reviewing scrap rate monthly without daily visibility is how plants miss two-week drift events entirely. The review cadence matters more than the KPI target.

Can software reduce scrap rate?
Software does not reduce scrap. Faster detection of defect causes reduces scrap, and good software makes that possible. The mechanism in practice: real-time defect tagging at the source, automatic correlation with process parameters and defect causes, Pareto ranking by shift and by machine. Typical improvement from turning an opaque manual process into a transparent automatic one is 5–15 % scrap reduction within 6–12 months — that is the SYMESTIC portfolio range across quality-focused implementations, and it is consistent with what I saw running global MES programmes at Visteon and Johnson Controls.

What's the single most common mistake in scrap management?
Managing the reported number instead of the real number. I have watched plants celebrate a scrap rate falling from 1.2 % to 0.8 % while the real quality loss — scrap plus rework plus start-up waste plus downgrade — stayed flat at around 3 %. The reported number improved because the coding conventions tightened, not because the process did. Before optimising any quality metric, spend one quarter making sure the metric is honest. Then improve it.

Related: OEE · Production Defect · Process Quality · Process Control · Production Efficiency · MES

About the author

Christian Fieg

Head of Sales at SYMESTIC. Over 25 years in manufacturing — Johnson Controls (Six Sigma Black Belt, headliner production, global MES for 900+ machines and 750+ users across four continents), Visteon, iTAC Software, Dürr. Author of OEE: Eine Zahl, viele Lügen (2025), a book on the structural dishonesty of manufacturing metrics and why a low-but-real number is worth more than a high-but-false one. · LinkedIn