Incoming Coplanarity Failure, a 35% GRR, and a Line-Down Deadline
A dock-to-stock program accepts a 50,000-piece lot of board-to-board connectors on the supplier's Certificate of Analysis plus a 30-piece incoming sample. Incoming inspection measures coplanarity with a new automated vision system; 4 of 30 parts exceed the 0.10 mm maximum, reading 0.12–0.15 mm. The supplier's CoA reports a maximum of 0.06 mm from its own laser system. A gauge R&R study on the vision system gives 35% GRR, and its correlation with the supplier's laser is unknown. The lot is urgently needed: the line stops in five days, and requalifying a second source takes eight weeks. The supplier's scorecard shows 99.7% conformance over two years with no previous coplanarity escapes. The purchase order invokes the supplier's ISO 9001 certificate and a quality agreement requiring a SCAR for any lot rejected at incoming. The receiving inspector wants to reject and return the lot.
1. Should the lot be rejected on this evidence? Analyze the measurement and decision risk.
Read the worked answer
Not yet, on this evidence. The decision hinges on measurement capability, not on the four observed readings alone.
A 35% GRR is above the commonly used 30% marginal threshold and well above the 10% target. For a 0.10 mm one-sided limit, measurement variation of that size can create apparent failures from conforming parts, and can also mask real failures. The observed 0.12–0.15 mm readings exceed the limit, but without a verified reference method the magnitude of the true deviation is unknown.
The sample is also small. Four failures in 30 gives an observed fraction of 13.3%, with a wide confidence interval extending to roughly 30%. That is enough to justify concern and containment, but not enough to declare the lot nonconforming when the gauge itself is suspect.
The supplier's CoA of 0.06 mm maximum is not proof of conformance either, because the two measurement systems have never been correlated. A 0.06 mm versus 0.12 mm gap could be a real process difference, a bias between systems, or edge-definition differences in how each system detects the seating plane.
Given the two-year 99.7% history, the prior probability of a bad lot is low but not negligible. The correct first step is a referee measurement and a correlation study, with the lot quarantined. Rejecting and returning the lot now risks a false rejection and a line stop; accepting it risks a latent field failure. Neither is justified before the measurement question is settled.
2. Outline a five-day plan that protects the line without accepting unverified material.
Read the worked answer
Day 0–1: quarantine the lot and stop dock-to-stock release. Select 10–12 parts spanning the observed range (including the 0.12–0.15 mm readings) and measure them on three systems: the vision system, a CMM or optical reference, and the supplier's laser, with the same parts, same fixturing, and blind part IDs. Compute bias and agreement against the reference. Re-run the GRR properly with the production parts, not the calibration artifacts used before.
Day 1–2: if bias is found, either correct the vision system or apply a guard band, accepting only parts clearly inside the limit (for example 0.08 mm on the corrected scale). Sort the lot or a sufficient sub-lot with the reference method, and segregate conforming material with lot and inspection traceability. If sorting capacity is insufficient, ask the supplier for an expedited replacement or a certified sort at its site.
Day 2–3: issue the SCAR now, not after the crisis, and request the supplier's containment data, its own GRR, and a correlation study with our system. This starts the supplier's clock while the line is protected.
Day 3–5: release only sorted, verified material under a documented deviation approved by engineering and quality, with a defined quantity and expiry. Notify the customer if the quality agreement requires it. The line stop is avoided by sorting and correlation, not by waiving the requirement.
3. What should the SCAR and supplier development actions require, and how should the acceptance decision be documented?
Read the worked answer
The SCAR must cover two distinct problems: the coplanarity of the product and the measurement disagreement between the two labs. Requiring only a process fix would leave the next lot disputed again.
Product-side requirements: root cause of the coplanarity variation (lead forming, plating stress, carrier warpage, packaging), corrective action with process data, and capability from the actual production stream, not a hand-picked sample. Require a control plan update and evidence that the corrective action was verified on at least three consecutive lots.
Measurement-side requirements: a GRR study from the supplier's laser system using production parts, a formal correlation study with our vision system, agreement on a referee method for disputes, and acceptance of a guard band. Dock-to-stock should be conditional on periodic correlation checks, not on the CoA alone; the skip-lot rule should include a periodic verification sample.
Documentation: record the disposition of this lot as sorted and released under deviation, with quantity, date, deviations, and traceability to the sort records. If any material is accepted without full conformance, the deviation must be approved by the design authority and by the customer if the quality agreement requires it. Record the SCAR number, the correlation results, and the revised acceptance criteria in the supplier file.
The commercial urgency is real, but it must not convert a measurement dispute into a waived specification. A 35% GRR system should not be the basis for rejecting a two-year, 99.7% supplier, and it should not be the basis for accepting the lot either.