A quality manager at a pump manufacturer can tell you the scrap rate to two decimal places. Ask what poor quality costs the plant in a year and the number goes soft. Scrap is in one report, rework hours sit inside labour, the sorting contractor is booked under operations, the customer credit note lives in finance, and the line that stopped for four hours while somebody traced a bad batch was recorded as downtime, not as quality.
None of that is bad accounting. Each number is correctly filed. The problem is that no one adds them up, so the total never appears in front of anyone who could act on it. This guide covers what COPQ is, the four categories it splits into, how to calculate it from numbers your plant already has, and where the biggest share usually hides.
The cost of poor quality (COPQ) is the total cost a business carries because things were not made right the first time: scrap, rework, re-inspection, sorting, warranty, recalls and lost customers. The American Society for Quality puts it at 15 to 20% of sales revenue for many organisations, and in manufacturing the average lands near 15%, ranging from about 5% to 35% depending on product complexity.
What is COPQ?
COPQ stands for cost of poor quality. It is the money a business would not have spent if every unit had been made correctly the first time. That includes the obvious costs like scrapped material, and the ones that rarely get labelled as quality: the overtime spent re-running a batch, the contractor hired to sort suspect stock, the freight on a replacement shipment, the engineer pulled off a project to run a root cause analysis.
The distinction that matters is between money spent making quality and money spent because quality failed. Training an operator is the first. Paying that operator overtime to rework what a different operator got wrong is the second. Only the second is COPQ.
What is COPQ in manufacturing specifically?
In manufacturing, COPQ is unusually easy to underestimate because the failures are distributed across departments that do not talk to each other in the same units. Scrap is measured in kilograms or pieces. Rework is measured in hours. Sorting is an invoice. Warranty is a provision. A recall is a board-level event. They are all the same underlying defect, priced at four different points in its life.
That fragmentation is why plants with genuinely good scrap numbers can still be carrying a heavy COPQ. The material loss is controlled. The labour, logistics and customer costs downstream of it are not being counted against the same defect.
What are the four categories of quality costs?
Quality costs split into four categories, two of which are investments and two of which are losses. COPQ is the second pair.
| Category | What it is | Examples | Counts as COPQ? |
|---|---|---|---|
| Prevention | Spend to stop defects occurring | Training, design reviews, process capability studies, poka-yoke devices | No |
| Appraisal | Spend to find defects before they escape | Incoming inspection, in-process checks, final QC, test equipment, calibration | No |
| Internal failure | Defects caught before the customer sees them | Scrap, rework, re-inspection, sorting, downgrade, line stoppage | Yes |
| External failure | Defects that reached the customer | Warranty, returns, credit notes, field service, recalls, lost accounts | Yes |
Prevention and appraisal are what you choose to spend. Internal and external failure are what the process spends on your behalf. The whole discipline is about shifting money from the second pair to the first, because the exchange rate strongly favours it.
Does COPQ include appraisal costs?
No, and this is the most common mistake in a first COPQ calculation. Appraisal is the cost of looking for defects, not the cost of having them. Inspection salaries, gauge calibration and test rigs belong in the cost of quality, but not in the cost of poor quality.
The exception is appraisal you only do because the process is unreliable. A 100% manual sort added to a line because it keeps producing escapes is not routine appraisal. It is containment, and containment is an internal failure cost. If you would stop doing the check tomorrow if the process were stable, count it as COPQ.
Which COPQ category is the worst?
External failure, by roughly an order of magnitude at each step it survives. The rule of thumb is the 1-10-100 rule, codified by George Labovitz and Yu Sang Chang in Making Quality Work (1992): a defect costs about 1 unit to prevent, about 10 to correct inside your own process, and about 100 once it reaches the customer.
Treat those figures as an order of magnitude, not an accounting law. The point is not that the multiplier is exactly ten. It is that the cost of a defect climbs in steps rather than inches, and every step it survives multiplies what it will eventually cost. In regulated sectors the final step is far steeper than 100, because a recall carries penalties and reputational cost that scrap never does.
This is also why COPQ reduction and inspection accuracy are the same conversation. Every defect moved from external failure to internal failure is a large saving. Every defect moved from internal failure to prevention is a larger one.
How do I calculate COPQ?
Add internal failure costs to external failure costs, then express the total as a percentage of sales revenue so it can be compared across periods and sites:
COPQ = Internal failure costs + External failure costs
COPQ as % of revenue = (COPQ / Total sales revenue) x 100
A worked example. Take a plant turning over 50 crore a year:
| Line item | Category | Annual cost |
|---|---|---|
| Scrapped material and components | Internal | 1.80 crore |
| Rework labour and machine time | Internal | 1.10 crore |
| Third-party sorting and containment | Internal | 0.45 crore |
| Line stoppages traced to quality | Internal | 0.60 crore |
| Warranty claims and credit notes | External | 1.30 crore |
| Return freight and field service | External | 0.40 crore |
| Total COPQ | 5.65 crore |
That is 11.3% of revenue, which would sit at the better end of the range the ASQ describes and still represents 5.65 crore that the process spent without being asked. Note what the table does not contain: no inspector salaries, no calibration, no training. Those are appraisal and prevention.
What metrics are used to measure the cost of quality?
Money is the reporting unit, but you cannot manage a rupee figure directly. These are the operational metrics that move it, and each one should have an owner:
| Metric | What it exposes | Feeds which COPQ category |
|---|---|---|
| First pass yield | Share of units made right with no rework | Internal failure |
| Scrap rate | Material lost per batch or shift | Internal failure |
| Rework hours per unit | Hidden labour absorbed by defects | Internal failure |
| Defect escape rate | Defects that got past final inspection | External failure |
| Parts per million defective | Escapes in the customer's own language | External failure |
| Warranty claims per thousand units | Field performance against build quality | External failure |
| Cost of quality as % of revenue | The roll-up that finance recognises | All four |
Two of these deserve special attention. First pass yield is the cleanest single indicator of internal failure, because it captures rework that scrap rate misses entirely. Defect escape rate is the cleanest indicator of external failure, and it is the one most plants cannot measure, for the reason in the next section.
Why don't companies track the cost of quality?
Because no single department owns the whole number, and the parts of it that hurt most are the parts nobody is measuring. Scrap has an owner in production. Warranty has an owner in finance. The four hours a line lost while a supervisor decided whether a batch was safe to ship has no owner at all, so it is filed as downtime and disappears.
There is a second reason, and it is more uncomfortable. Defect escape rate is only knowable if you know what left the plant defective, and most plants only learn that when a customer tells them. Sampling inspection makes this structural: if you check 1 in 500 units, you are measuring the sample, not the population, and the escapes you never hear about are invisible by construction.
So the number that would justify the investment is the number the current process cannot produce. That is the trap, and it is why COPQ programmes usually start with better detection rather than better reporting.
How do I reduce the cost of poor quality?
By moving defects earlier in their life, which means catching them closer to the operation that created them rather than at the end of the line or at the customer. Four moves do most of the work, roughly in order of return:
- Make detection continuous rather than sampled. Inspecting every unit turns escape rate from an estimate into a measurement, which is what makes the rest of the programme provable.
- Move the check upstream. A defect found at the station that caused it is an internal failure cost of one operation. The same defect found at final inspection carries the value added by every operation since.
- Verify the process, not only the product. Many defects are the predictable output of a step done in the wrong order or with the wrong component. Catching the process error prevents the product defect rather than sorting it.
- Attach evidence to every rejection. Without a record of what the defect looked like and when it occurred, root cause analysis is opinion, and the same defect returns next quarter.
This is where automated inspection changes the arithmetic rather than just the workload. Jidoka's KOMPASS inspects every unit at production speed instead of sampling, at accuracies above 99.5% and up to 12,000 parts per minute, which converts defect escape rate from a guess into a figure you can put in a report. NAGARE works on the other half of the problem, verifying that the operation itself was performed correctly so the defect is prevented rather than detected. Both run on cameras already installed on the line, and every reject carries a timestamped image, which is what turns a rework decision into a root cause.
See how continuous inspection changes the escape rate: AI defect detection for zero escapes
Where to start
Spend one afternoon building the number before you spend anything on fixing it. Pull twelve months of scrap value, rework hours at loaded cost, sorting and containment invoices, quality-caused downtime, warranty and credit notes, and return freight. Add them. Divide by revenue. Almost every plant that does this for the first time finds a figure two to three times larger than the one they would have guessed, because four of those six lines were never filed under quality.
Then look at the split. If external failure is a meaningful share of the total, the problem is detection, and sampling is why. If internal failure dominates and external failure is small, detection is working and the problem has moved upstream to process control. The two need different interventions, and the split tells you which one you are buying.
If you want to see what continuous inspection would surface on your own line, talk to our team. Related reading: poka-yoke examples in manufacturing, what OEE actually measures, AI assembly line inspection, and what quality control inspection covers.
