Quality Cost and Scrap Reduction

Executive summary

Quality cost and scrap reduction starts with a measurement problem. The money a plant loses to poor quality does not appear on one line of the income statement. Scrapped material sits in cost of goods sold. Rework hours hide in direct labor. Extra inspection sits in overhead. Warranty claims, returns, and credit memos sit in revenue deductions and SG&A. Customer penalties and lost orders appear nowhere at all. Because nobody sees the total, nobody owns it.

The cost of quality (or the cost of poor quality, COPQ) framework puts these costs in one table. It divides them into prevention and appraisal, which are the costs of conforming, and internal and external failure, which are the costs of not conforming. Once the table exists, three useful things become possible. You can size the problem in dollars and as a share of revenue. You can rank causes with a Pareto, so effort goes to the few causes that produce most of the loss. And you can compare the cost of a fix against the cost of the failure it removes, in the same units.

This guide shows how to build the table from records you already have, how to compute first pass yield and rolled throughput yield, how to run a scrap reduction cycle that fixes causes instead of reclassifying symptoms, and how to avoid the common failure of counting savings that never reach the bottom line. Numbers in the worked example are illustrative.

This is an operating guide, not a quality standard. Product safety, industry regulation, customer quality requirements, and certification obligations (for example, a quality management system standard your customers require) are set by your own standards, your customers, and your regulators. Follow those, and involve your quality and compliance advisers where they apply. Nothing here replaces them.

When to use this

Use this guide when:

• Scrap, rework, and returns are growing, but no one can say what they cost in total.
• Gross margin is below plan and material variance or labor variance is unexplained.
• Customers are sending back product, issuing corrective action requests, or threatening to resource.
• You are considering capital for new inspection equipment and need to know what it must prevent to pay back.
• The plant says quality is "fine" because final inspection catches problems, but the cost of catching them is high.
• A buyer or lender is examining margin quality and asks how much of cost of goods sold is recurring waste.

If you have no record of why units are scrapped or reworked, start by recording that. A scrap code list and a daily count are enough to begin.

Diagnostic questions

1. What did the plant scrap last month, in units and in dollars at cost, by product and by work center?
2. How many units were reworked, how many labor hours did the rework take, and where did that time get booked?
3. Is there a fixed list of scrap and defect reasons, and does every scrap transaction carry one?
4. What are the top five reasons by dollars (not by count)?
5. What did customers return or claim over the last 12 months, and what did each cost to resolve, including freight, credit, and labor?
6. How much inspection and testing do you do, and does it catch defects early or only at the end of the line?
7. What is the first pass yield at each major step, and does anyone track it?
8. Which defects reach the customer, and which are caught internally?
9. How many scrapped units were scrapped because of setup, first-piece, or changeover, rather than steady-state running?
10. What share of scrap is material (supplier lot, wrong spec) versus process (machine, method, operator) versus design?
11. Who owns the cost of scrap today, and is anyone measured on it?

The framework: the cost of quality table

The cost of poor quality is internal failure plus external failure. The cost of good quality is prevention plus appraisal. The total of the four is the cost of quality. A mature plant typically wants failure costs to fall while prevention spending rises modestly, because a dollar spent keeping a defect out is cheaper than a dollar spent finding it, and much cheaper than a dollar spent when the customer finds it. That ordering is a reasoning rule, not a statistic. Test it with your own numbers.

Yield formulas

First pass yield (FPY) at a step = Units passing the step the first time, with no rework or scrap ÷ Units entering the step
Rolled throughput yield (RTY)     = FPY(step 1) × FPY(step 2) × ... × FPY(step n)
Scrap rate                        = Scrapped units ÷ Units started
Cost of a scrapped unit           = Material cost + conversion cost added up to the point of scrap − any salvage value

RTY matters because a line with four steps that each look like a good yield can still have a poor overall yield. Final yield, measured only at the end, hides internal rework loops, since reworked units that finally pass count as good.

Scrap cost must include what was added

A part scrapped at the first step costs its material. A part scrapped at the last step costs its material, all the labor and machine time added to it, and the overhead absorbed. Valuing scrap at material only understates the loss and makes late-stage scrap look cheap. Use standard cost up to the point of scrap, and say so in the report. Check with your controller how scrap and rework are valued in the ledger, because your accounting policy decides where these costs show up.

Pareto before projects

Quality problems follow an uneven pattern. A small number of causes usually produce most of the loss. Count by cost, not by occurrence. A defect that is frequent and cheap may matter less than a rare one that scraps an expensive unit late in the process. Build the Pareto first, then choose projects from the top of it.

Step-by-step process

Weeks 1-2: Build the baseline

1. Create the reason code list. Eight to fifteen codes, mutually exclusive, with a definition and an example for each. Include "setup or first piece," "material defect," "machine," "method," "operator," "design," and "damage in handling." Owner: Quality lead with plant supervisors. Output: a one-page list.
2. Pull 12 months of scrap and rework transactions by part, work center, and reason where it exists. Owner: Controller with the quality lead. Output: a transaction extract.
3. Value each scrapped unit at material plus conversion to the point of scrap. Owner: Controller. Output: a cost-per-unit table by part and step.
4. Pull customer-side costs: returns, credits, warranty, expedite freight, corrective action labor. Owner: Customer service lead and controller. Output: an external failure table.
5. Estimate appraisal and prevention costs from headcount, equipment, and training budgets. Owner: Quality lead and controller. Output: first draft of the full cost of quality table.

Weeks 3-4: Find the causes worth fixing

6. Build the Pareto of internal failure cost by reason, and a second by part number. Owner: Quality lead. Output: two ranked lists.
7. Compute FPY at each step and RTY for the main product families. Owner: Quality lead or industrial engineer. Output: a yield table.
8. Separate steady-state from transition scrap. Count units scrapped in setup, startup, and changeover separately from units scrapped mid-run. Owner: Shift supervisors. Output: a split by reason.
9. Choose the top three causes by dollars where a fix is feasible. Name one owner for each. Owner: Head of Operations. Output: a project list with owners.

Weeks 5-8: Fix causes

10. Contain, then investigate. For each chosen cause, put a containment action in place (additional check, quarantine, a stop rule), then find the root cause using a structured method such as five whys or a cause-and-effect diagram, with evidence from the floor. Owner: Cause owner.
11. Fix at the source. Typical fixes are standard work and first-piece approval for setups, preventive maintenance on a drifting machine, error-proofing a step that depends on attention, tighter incoming inspection or a supplier corrective action, or a drawing and process correction. Owner: Cause owner with engineering.
12. Verify the fix with data. Track the reason code for four weeks after the change. Do not close a project on the strength of a meeting. Owner: Quality lead.
13. Move detection earlier. Where a defect is found at final inspection but made at step two, add a check at step two. Owner: Quality lead and the process owner.

Weeks 9-12: Hold the gains

14. Update standard work and training for each fix. Owner: Supervisors.
15. Put scrap cost into the operating review. Report cost of quality by category, scrap by reason, FPY and RTY monthly. Owner: Controller and Head of Operations.
16. Reconcile the savings. Compare the claimed saving to material usage, labor hours, and credits in the ledger. A saving that does not show up in the numbers is a forecast, not a result. Owner: Controller.

Templates

Cost of quality table (copy into a spreadsheet)

Scrap log with reason codes

Yield table

Corrective action sheet (one per top cause)

Metrics and KPIs

Setting targets without invented benchmarks. Use your own trailing best: the best quarter of scrap rate and FPY for each product family, which proves the process can do it. Use contractual requirements where a customer sets a quality floor in the agreement. For an outside reference, use a named public peer's disclosed data, or a standard your customer specifies. Do not apply a generic parts-per-million target from a conference slide to a different process. Set the first-year goal as a share of the current top-cause loss that your root cause work supports.

Worked example (illustrative)

The figures are made up to show the method. They are not benchmarks.

A manufacturer with $60.0M of annual revenue assembles its first cost of quality table from a year of records.

Cost of good quality (prevention plus appraisal) is 150 + 400 = $550K. Cost of poor quality (internal plus external failure) is 900 + 450 + 350 + 200 = $1,900K. They sum to $2,450K, or 4.08 percent of the $60.0M of revenue. Failure cost alone is about 3.17 percent of revenue. Prevention is $150K, roughly 6 percent of the total, which is a signal worth examining: the plant spends most of its quality money finding and paying for defects rather than keeping them out.

The scrap Pareto

Scrap is $900K per year, or $75,000 per month. The plant scraps 1,200 units per month at an average cost of $62.50 per unit. The cost per unit is $40.00 of material plus $22.50 of conversion added up to the point of scrap. The check: 1,200 × $62.50 = $75,000.

Setup scrap is the largest single cause at 35 percent of units. Suppose a standard-work and first-piece approval project, costing a one-time $30,000 in fixtures and training, halves setup scrap. The avoided units are 210 per month. At $62.50 per unit, the saving is 210 × $62.50 = $13,125 per month, or $157,500 per year. Against a $30,000 one-time cost, the payback is under three months on paper. The word "on paper" matters. The saving is real when material purchases fall or output of good units rises, so the controller reconciles it to the ledger in week 16 of the plan.

First pass yield and rolled throughput yield

A four-step process has these first pass yields:

RTY = 0.98 × 0.96 × 0.99 × 0.97 = 90.3%. Each step looks healthy, yet roughly one unit in ten needs rework or is scrapped somewhere along the way. This is why step-level yield alone is not enough. Where the rework loops are hidden, a plant can report a final pass rate near 100 percent while spending real money on the loop.

Common mistakes

1. Valuing scrap at material only. It understates the loss and makes late scrap look cheap.
2. Counting rework as free. Rework consumes labor and constraint time, and it can raise scrap.
3. Pareto by count instead of dollars. A cheap, frequent defect can crowd out an expensive rare one.
4. Relying on final inspection. It catches defects after the cost has been added, and it is imperfect.
5. Reclassifying instead of fixing. Moving scrap into rework or into "other" improves a report and changes nothing.
6. Blaming operators first. Operator error is often a design, training, or standard work problem. Ask what made the error easy.

AI prompt assistance

AI can tabulate, rank, and draft. Your people hold the floor knowledge. Give the model your data and check every ranking against what you can see.

1. Build a Pareto from a scrap log.

I will paste a scrap log with date, part, step, units, reason code, and cost per unit. Compute total cost by reason code, share of total, and cumulative share, sorted by cost. Then give the same view by part number. Flag any row with a missing reason code or cost.

Check: recompute the total of one reason by hand and confirm the grand total ties to the scrap account in the ledger.

2. Normalize free-text defect notes.

Here are 60 free-text notes from operators describing why a part was scrapped. Map each to one code from this list: [list]. Return a table with the original note, the code, and a confidence of high, medium, or low. Do not invent a code that is not on the list.

Check: review every low-confidence row with a supervisor and sample ten high-confidence rows.

3. Compute yields.

Given units in, units passed first time, scrapped, and reworked at each of four steps, compute FPY per step and RTY. Show the multiplication. Then compute the final pass rate if reworked units are counted as good, and explain the difference.

Check: verify the unit counts against the traveler or MES records, not estimates.

Implementation guide

Downloadable assets

The cost of quality table, scrap log, yield table, and corrective action sheet above are designed to be copied into a spreadsheet.

Related resources

• Operations guides: the full operations library.
• Capacity utilization and throughput analysis: the cost of scrap at the constraint.
• Lean visual management for plants: making abnormal conditions visible.
• S&OP process setup: where quality yield enters the supply plan.
• Margin expansion playbook: where quality savings land.
• On-time delivery root cause attribution and the OTD root cause calculator: when rework and scrap are the reason an order ships late.

Frequently asked questions

What is the cost of poor quality?

The cost of poor quality is the total of internal failure costs (scrap, rework, retest) and external failure costs (returns, warranty, credits, replacement freight, corrective action response). It sits in several places in the P&L, which is why most companies cannot state it. Build one table from your records, and express it as dollars and as a share of revenue.

How do you calculate first pass yield?

First pass yield at a step is the units that pass the first time, with no rework or scrap, divided by the units entering the step. For a multi-step process, multiply the step yields to get rolled throughput yield. Rolled yield is often lower than any single step suggests, because the losses compound.

Why should scrap be valued at more than material cost?

A unit scrapped late has absorbed labor, machine time, and overhead along with its material. Valuing it at material only understates the loss and makes late-stage scrap look cheap. Value scrap at standard cost up to the point where it was scrapped, less any salvage, and confirm the treatment with your controller.

Where should a scrap reduction project start?

Start with a Pareto of scrap cost by reason code, because a few causes usually dominate. Count by dollars, not occurrences. Then separate setup and first-piece scrap from steady-state scrap, since the fixes differ. Pick the top cause where you can name an owner and a verification metric.

How do you know scrap savings are real?

Reconcile the claimed saving to the ledger. Real savings appear as lower material purchases, fewer labor hours, or more good output from the same inputs. A reduction in the scrap log that never reaches those lines may be a change in coding. The controller should sign off on the reconciliation.

How this works in a System of Performance

Scrap and rework data are rarely the problem. The problem is that the scrap log, the labor records, the credit memos, and the corrective actions sit in different systems, so the cost of quality never appears as one number and the fix never gets traced back to its effect. A Performis routine can take on a recurring job like this, with the scope and fit agreed with you: it works from the records your team supplies, prepares the work for your named reviewer, and keeps each approval on record. If you want to see what poor quality is costing in your own records, bring it to us as a launch partner.