Classify the event at the acceptance boundary
The acceptance boundary determines which operational record owns the event. Rework before the customer receives the accepted output is an internal failure. A repair, replacement, repeat service or callback after delivery is an external failure. Prevention and appraisal are separate quality-control investments.
| Cost class | When it occurs | Evidence to capture | Decision use |
|---|---|---|---|
| Prevention | Before failure | Training, process, supplier or design cost | Compare with avoidable failure cost |
| Appraisal | Inspection or testing | Inspection hours, test and acceptance record | Evaluate detection effort |
| Internal failure | Before accepted delivery | Rework, scrap, retest and delay records | Fix process or quote recovery |
| External failure | After accepted delivery | Callback, replacement, travel and complaint record | Review warranty and prevention exposure |
Event-ledger fields
- Event ID, date, job/product/service and accepted-output link. (not complete)
- Internal or external classification and detection point. (not complete)
- Rework or callback hours and the loaded labour rate basis. (not complete)
- Replacement materials, subcontractors, travel and disposal. (not complete)
- Realised recovery, supplier credit or customer charge, recorded separately. (not complete)
- Capacity effect only when displaced or recovered work is evidenced. (not complete)
- Cause and avoidable share as evidence or a clearly labelled assumption. (not complete)
Value internal and external failures separately
Internal failure cost = rework hours × loaded labour rate + replacement materials + retest/disposal cost - realised recovery
- rework hours
- Worker-hours spent correcting the failed output before acceptance (hours per event) — time or job record
- loaded labour rate
- Included business labour cost on a consistent basis (CU per hour) — business cost record
- replacement materials and retest/disposal
- Other internal event costs inside the boundary (CU per event) — job, inventory or supplier record
External failure cost = callback hours × loaded labour rate + travel + replacement/subcontract cost + other included event cost - realised recovery
- callback hours
- Worker-hours spent after accepted delivery (hours per event) — service or job record
- travel
- Incremental travel cost inside the callback boundary (CU per event) — vehicle, expense or user record
- replacement/subcontract cost
- External materials or delivery cost caused by the event (CU per event) — supplier or subcontractor record
Lost reputation and future demand are not inserted as invented monetary values.
- Define accepted output and the review period.
- Classify each event by whether it was detected before or after acceptance.
- Value labour, material, travel, subcontract and recovery rows once.
- Multiply event costs by observed event counts and reconcile to records.
- Divide by accepted output only when a per-output view helps the decision.
- Estimate an avoidable share for a named prevention action and expose its evidence.
- Compare prevention and appraisal cost with avoidable failure cost under downside cases.
- Reconcile the same ledger after action before treating a reduction as realised.
Worked example: one internal and one external failure class
Fictional monthly CU scenario, tax excluded. The example values failure events; it does not estimate a warranty provision or legal obligation.
| Intermediate row | Internal rework | External callback | Combined result |
|---|---|---|---|
| Observed events | 18 | 6 | — |
| Hours per event × loaded rate | 1.5 × 42 = 63 CU | 2.5 × 42 = 105 CU | — |
| Material/retest or replacement | 22 CU | 65 CU | — |
| Travel/other included cost | 0 CU | 30 CU | — |
| Recovery per event | 5 CU | 0 CU | — |
| Net cost per event | 63 + 22 - 5 = 80 CU | 105 + 65 + 30 = 200 CU | — |
| Period failure cost | 18 × 80 = 1,440 CU | 6 × 200 = 1,200 CU | 1,440 + 1,200 = 2,640 CU |
| Combined accepted-output denominator | — | — | 400 accepted units |
| Combined failure cost per accepted unit | — | — | 2,640 ÷ 400 = 6.60 CU |
Test prevention without hiding uncertainty
| Case | Avoidable failure cost | Action cost | Bounded interpretation |
|---|---|---|---|
| Base assumption | 40% × 2,640 = 1,056 CU/month | 4,000 CU fixed | Simple recovery threshold is about 3.8 months. |
| Downside avoidability | 20% × 2,640 = 528 CU/month | 4,000 CU fixed | Threshold extends to about 7.6 months. |
| Variable inspection cost | 1,056 CU/month | 4,000 fixed + 350/month | Net modelled recovery is 706 CU/month. |
| No cause evidence | Unsupported | Known quote | Do not claim a financial payback yet. |
Avoid the quality-cost blind spots
- Do not mix pre-acceptance rework with post-delivery warranty events. (not complete)
- Do not use wage rate when the decision requires loaded labour cost. (not complete)
- Do not omit travel, retest, replacement and subcontract cost. (not complete)
- Do not count a supplier credit until it is realised. (not complete)
- Do not monetise reputation, demand or released capacity without evidence. (not complete)
- Do not let financial optimisation override safety, quality or legal obligations. (not complete)
Rework and warranty questions
- Should failure cost simply be added to every price?
- Not automatically. First reconcile the event ledger, identify avoidable and unavoidable components, and decide whether price, prevention, supplier recovery or scope control owns the issue.
- Should lost future revenue be included?
- Only when supported by a separate, defensible model. Do not insert an invented reputation or demand value into the event cost.
- Does prevention always pay back?
- No. Compare the action with evidenced avoidable failure cost, and preserve mandatory quality activity regardless of financial payback.
Sources and methodology boundary
- Cost of quality — ASQ: Supports prevention, appraisal, internal-failure, rework, external-failure and warranty classifications.