The Hold That Everyone Wants Released
Northline Plastics runs 14 injection molding presses. Dana, the quality manager, reports to plant manager Rick with a dotted line to corporate QA. Last month a Tier-1 automotive customer rejected a lot: cavity 7 on press 9 produced bores 0.06 mm below the low limit on 2.1% of parts sampled after sorting. Dana's data show Cpk of 0.92 for cavity 7 versus 1.4-1.7 on the other eleven cavities. Forty thousand parts from the same run sit on hold awaiting disposition. The customer's launch is in nine days and the contract carries a $280,000 weekly late penalty. Rick wants to ship against an internal deviation, arguing the plant has had no field failure in two years and that the last three quality holds were eventually released anyway. Corporate QA requires notification within 24 hours of any shipment made on deviation.
1. What should Dana do in the next 24 hours? Justify the decision principle, not only the action.
Read the worked answer
Decision principle: disposition authority is bounded by customer risk and by the quality policy Dana is accountable for, not by plant schedule pressure. An internal deviation cannot substitute for customer acceptance of a nonconforming characteristic on a safety-adjacent dimensional feature.
Actions:
- Keep the hold in force and notify corporate QA within the 24-hour window - this is a compliance obligation, not a judgement call.
- Separate the population: cavity 7 parts are not interchangeable with the rest. Sort by cavity signature if the press records it, so only suspect parts stay quarantined. This converts a 40,000-piece problem into a smaller one and is the fastest way to protect the schedule.
- Build a one-page risk package for Rick: 2.1% observed nonconformance beats any modeled estimate, the functional consequence of a 0.06 mm undersize bore (press-fit retention, leak path), and the cost comparison of penalty versus containment/sort/air freight.
- If the customer will consider it, route a customer-approved deviation with cavity-level traceability, a 100% inspection method validated by gage R&R, and a written containment plan. That is a legitimate path; an internal sign-off is not.
- Document the disagreement in writing. Escalation is not insubordination when the escalation route is defined in the quality manual.
Trap to avoid: treating this as a personality conflict with Rick. Frame it as a disposition decision with defined authority and defined evidence.
2. Evaluate Rick's argument that Cpk 0.92 means the process is essentially 99% good, and that two years without a field failure supports shipping.
Read the worked answer
The Cpk arithmetic: a Cpk of 0.92 corresponds to roughly 2.76 sigma of margin on the closer side, about 0.3% beyond that limit, or on the order of 5,800 ppm if the distribution is centered and normal. That is not 99% good in any sense the customer cares about; it is roughly fifty times the 1.33 benchmark.
Why the number is even less reassuring than it looks:
- Cpk is a snapshot that assumes stability and normality. The stratifying variable here is cavity, and a pooled index averages eleven healthy cavities with one sick one. Cavity-level data, not the plant index, is the relevant evidence.
- The 2.1% observed after sorting is direct measurement of the population that matters. When an observed defect rate contradicts a modeled one, the observation wins.
- Two years without a field failure is not evidence of low risk; it is evidence of low detection opportunity. Escape rates depend on how much of the product is inspected, and if inspection has been reduced, the absence of complaints is a statement about the inspector, not the process.
What Dana should say: the index is the wrong lens; the 2.1% is the right one, and it is per-cavity and reproducible. Offer to re-run capability by cavity on the next run with rational subgroups so the argument is settled with data rather than assertion.
3. Six months later, what leadership system changes would keep this situation from recurring?
Read the worked answer
The pattern is systemic, not personal. Rick has learned that holds get released, and everyone in the plant has learned it too. Culture is what the escalation record says it is.
Changes worth making:
- Written stop/hold authority for quality, with a defined escalation clock and a defined tiebreaker (corporate QA). Ambiguity is what let the argument happen at all.
- Scorecard alignment. If Rick is measured only on OTD and cost, quality will lose every trade-off. Add cavity-level Cpk and cost of poor quality to the plant manager's measures, and stop rewarding release-with-deviation.
- Cavity-level SPC on the high-risk tools with a reaction plan owned by the shift leader, so drift is caught before a lot is built.
- Escalation audit. Review the last three released holds. If the releases were technically justified, say so and fix the process that generated the holds; if they were not, that is the culture problem and naming it is the leadership act.
- Management review discipline: a standing agenda item on open deviations with ageing, so decisions are made once, in the open, with the customer's voice in the room.
Test of success: a supervisor stops a line in month three and nobody asks who authorized it.