Who Is the Cost-of-Quality Report For: Making COPQ's Four Categories Work

A cost-of-quality version nobody believes

Start with a real case. An automotive parts maker's quality department did its first cost-of-quality count by asking finance for the year's scrap value and sales for the total of customer claims, then adding the two. The number was small, and management's first reaction was that their quality must be fine.

Yet in that same year, the two things that bothered the workshop manager most were: one operation needing rework constantly, with skilled workers pulled off to fix parts every day; and the rush to tidy records before every customer audit. Both cost money and both consume people, and neither landed on that sheet.

The problem is the basis, not the attitude. The standard split of cost of quality is four categories —prevention cost, appraisal cost, internal failure cost, external failure cost, collectively called the COPQ (Cost of Poor Quality) framework. Money in the first two is spent before defects occur, in the last two after. Counting only the two easiest items inside the failure categories turns cost of quality into nothing more than a scrap ledger, and every judgment management makes from it is distorted.

The more practical consequence is that investment goes the wrong way. Of the four categories, what actually brings the total down is usually moving money to the prevention side — error-proofing, process changes, moving inspection upstream. But if prevention cost has never been counted, that spend is pure cost on the report, and whoever raises it looks like they are asking for budget.

The four-category ledger: the difficulty is data, not classification

The four definitions are not hard; what is hard is that each category has items where you can see money going out but cannot find the document. The table below lists what is most often missed in practice.

The omitted items share one trait: unlike scrap, they have no standalone document, but sit inside wages, hours, travel and freight. Getting them out requires splitting after the fact, and splitting requires rules — which is exactly what the account table is for.

One more type is more hidden:Price difference from downgrade and concession. Selling a nonconforming part as second-grade is a real quality loss, but because the money was received, it never appears in most counts.

First, fix the nature: this is a decision ledger, not an accounting head

Agree on one thing before starting: COPQ is not an accounting head, and it does not have to map one-to-one onto finance accounts. It has exactly one purpose —answering whether the money spent on quality is worth it, and where it should move. So the design principles are: a stable basis, drill-down capable, comparable across periods.

The only test isWhy this money was spent: spent to keep defects from happening, that is prevention; spent to find defects, that is appraisal; the defect has occurred and is still inside the plant, that is internal failure; it reached the customer, that is external failure. Under this rule there is no ambiguity about which category a sum belongs to.

Ambiguity usually shows up at two boundaries. First,Inspector hours: routine first-piece and patrol inspection is appraisal, while sorting and full inspection added because of a complaint is external failure — the same inspection hours, a different trigger, a different category. Second,Gauge calibration: planned calibration is appraisal, while the re-inspection triggered by a gauge found out of calibration is internal failure.

The rule for handling boundaries is simple: collect byTriggering documentCollect by triggering document: inspection plan triggered goes to appraisal; complaint ticket or NCR triggered goes to the failure categories. That way every sum traces to a document, and arguments land on documents instead of estimates.

Collection basis for each of the four categories

The basis has to land on the account table, and each line states three things: what to include, what to exclude, where the figure comes from. The table below can be used as a first version directly.

The 'what not to include' column is the one most often dropped, and the most critical. Without exclusions the accounts keep inflating — safety training invoices and capacity expansion equipment get pushed in, and the total looks good but nobody decides anything with it.

The source column must stateSingle source document. If the same money has two sources (it is in the finance ledger and also in the inspection record), one must be designated authoritative and the other used only for cross-checking. Dual sourcing always yields two numbers, and two numbers means no number.

Three places most often miscounted

First,Scrap only. This is the omission from the previous section and the most common shape it takes. The test: divide the total by output and see whether per-unit quality loss is implausibly low; if it is too low to cover even one customer return, it is probably scrap only.

Second,Rework hours counted at standard hours. Standard hours systematically understate internal failure, because rework usually goes to skilled workers whose actual time is below standard, so pricing at standard overstates it; on the other hand, rework brings waiting and setup, so actual occupancy runs above pure operation time. The bias cuts both ways, and the only stable approach is to take the actual start and end times on the rework order.

Third,External failure counts compensation only. After a customer complains, the real money often is not the compensation itself: travel to sort parts on site, expedited freight to ship replacements, third-party re-inspection the customer demands — together these regularly exceed the compensation. That money sits in expense claims, and without a complaint number on the claim it can never be allocated — which is whyThe expense claim must carry a complaint number field, and that is the watershed for whether the total can be complete.

Designing the account table: responsibility must land on a person

The account table is not quality's alone. For every one of the four categories the figures sit with another department, and the responsibility belongs in the table.

The 'entered on the shift' cell decides whether the whole thing runs. If rework hours and downtime may be backfilled at month end, the backfilled numbers will be wrong — nobody remembers how long a hold lasted two weeks ago. So the rule is simple: whoever creates it enters it, and it is entered within the shift.

The point of the finance cell isNet-amount basis: scrap loss takes off residual value, and rework takes off reusable material. Counting gross overstates internal failure, and after a while nobody reads the sheet.

Allocation basis: no one blunt rule

Cost of quality needs allocation because it has to be compared with output value and volume. Pick the wrong allocation basis and the improvement direction drifts.

The four bases serve four purposes, and there is no single correct denominator. The standard approach isOne primary basis, the others as support: management reporting fixes the output-value basis, workshop comparison fixes the hours basis, and the batch basis is used only for improvement ranking. Mixing denominators on one sheet is the most common error in cost-of-quality reporting.

One more boundary to state:Lost orders and goodwill are not counted. They are real, but they cannot be evidenced or recomputed, and estimating them into the total destroys the credibility of the whole sheet. Express that part with countable process metrics such as complaint counts and repeat-complaint rate, not with an estimated amount.

Who is the report for: three dashboards

The most common failure of a cost-of-quality report is showing everyone the same sheet. A workshop manager cannot read the four-category structure, and the general manager does not care about one team's rework hours. Split it by level, one sheet each.

The three levels must share one basis. The scrap total on the workshop sheet must equal the internal-failure scrap line on the quality sheet, and the quality sheet total must equal the management sheet total. When they tie out, the report is trusted; when they do not, every level gets questioned.

What matters at the quality level isStructural change, not the total. Prevention and appraisal spend rising while failure falls is the healthy direction; all four falling together usually means the counting has loosened, not that quality improved. Put this line in the report notes, or management will see the total drop and assume all is well.

Where does this land inside JEDP

None of this needs a new system; it lands inside the configuration scope. The account table and the four-category mapping are built with form modelling, second-level accounts are fields, and changing an account needs no code. Rework orders, concession records and complaint tickets run on the flow engine, with time limits on nodes and approvals on record; the document number is what triggers the allocation.

On the data side, the report engine pulls figures against the defined basis: the four categories and the three dashboards share one source, one basis produces three tables. The integration engine brings in amounts from finance and hours from production out of the existing systems, so the same money is never entered twice. The permission engine sets the scope of quality roll-up and drill-down: who sees the whole plant and who sees only their own workshop, with any overreach rejected on the spot.

The link to nonconforming product handling is the key. A hold decision must be able to freeze stock and block the next station, rework and scrap must enter the work order state machine, and complaint claims must reach accounts payable — if any of the three is broken, cost of quality only captures the half that has documents. The related practices are covered in three other articles: NCR raised but goods still moving, rework order open yet progress counted as complete, and claim raised but the debit never lands.

Rollout sequence, six acceptance checks and cases to leave alone

Four steps are recommended, each with a checkable completion marker:

Do not check the policy document at acceptance; have them demonstrate these six on the spot:

1. Open a rework order and see actual start and end times, hours and material, automatically allocated to internal failure
2. Look up a complaint and see the total of compensation, travel and expedited freight, gathered under that complaint number
3. Click any amount in the four-category structure and drill down to the list of documents, verifiable line by line
4. Take a batch of downgraded sales records and see list price against amount received, with the difference counted into internal failure
5. Switch between workshop and plant-wide views; the data scope follows permissions, and the total ties to the upper-level report
6. Export six consecutive months of four-category data; the change in structure can be charted, and any basis change is marked

Only when all six can be demonstrated has cost of quality entered the system. Any one that cannot be shown is still living in Excel.

Finally, the boundaries. In three situations my advice is not to rush into the full set of statistics:

First,Cause code can be blank. A cause code is the only pointer to the improvement direction; leaving it blank means the collected total points at nothing. Make the field mandatory with a dictionary before you talk about statistics.

Second,No document entry point for rework and downtime. With no entry point it can only be estimated at month end, and the estimate gets torn apart at the first review. Add the documents first, then count.

Third,Tie it to performance before the owner is even assigned. Once it is tied to performance the data gets massaged, and the direction is usually not better, just smaller. The first year has only two jobs: a stable basis and reproducible numbers.

None of the three is a problem with the four-category split; the split only exposes them faster. Conversely, once those three are settled, a cost-of-quality figure with a stable basis that drills down to documents gives you real decision input: how much to move to the prevention side, which operation to fix first, which class of complaint costs most — the answers stop being guesses.

• QMS quality management: one closed loop across incoming, in-process and complaint handling
• Report engine: one definition, one source for metrics
• Process engine: one rule set for approval and business flow
• Form engine: model, fields and page generation
• CAPA closure rate nobody believes: define closed as three verifiable conditions
• Rework order raised yet progress still counts as complete? Scrap and rework must enter the work order state machine
• Claim raised but the debit never lands? Quality chargebacks must reach accounts payable
• IQC and supplier quality as two skins—weld the loop at four nodes
• 8D and CAPA process template pack
• QMS Selection 20-Question Checklist