Kicking off a 500-unit pilot next week, and I need to set incoming/outgoing sampling that protects quality without throttling flow. I’m leaning ISO 2859-1, Level II, AQL 0.65 for criticals with c=0, but our takt is 42s and containment capacity is limited; has anyone staged from stricter initial gates and relaxed after the first three lots hit Cpk > 1.33, or is there a better control plan for early builds?
Did this on a 480‑unit pilot with a ‘42s’ takt: we ran ISO 2859‑1 Level III, AQL 0.65 with c=0 for the first about [redacted], then shifted to Level II with skip‑lot 1‑in‑5 after two lots held Cpk > 1.33 to keep flow. If containment’s tight, make the c=0 gate only the highest‑risk CTQ with a go/no‑go at the bottleneck station.
I’d echo @aaron_j92 but make the first gate time-boxed: at a 42s takt we ran “c=0” only for the first 2 hours (about [redacted]) and relaxed to ISO 2859-1 Level II with skip-lot 1-in-5 once Cpk>1.33 and FPY>98% on two consecutive subgroups. If rework is ugly, , add a short 100% screen on the top CTQ for the next shift instead of keeping full c=0 across three lots.
On a 520‑unit pilot at about 42s takt, we used a belt‑and‑suspenders start tied to FPY and Z1.4: 100% on the two CTQs for the first 120 pieces, then normal AQL with skip‑lot 1‑in‑7 once FPY ≥98.5% and Cpk at least 1.33; any reject snaps back to tightened. If you try @luna7275’s time box, add a hard piece‑count trigger so a slow start doesn’t eat your buffer.
Quick tip: use a truncated sequential plan (Wald) at start — ‘c=0’ on the CTQs with early accept/reject bounds — so clean lots clear fast and you only grow the sample when the data says so. We switched to normal Z1.4 once Cpk > 1.33 for two consecutive subgroups; if your defects cluster, park a tiny quarantine buffer so one blip doesn’t starve the line. @emma39’s time‑box works too, but sequential gets you the same protection with fewer pieces on the table.
I’ve had good results tying ISO 2859‑1 switching to a rolling window instead of fixed lots: at 42s takt we started on tightened Level II with c=0 on the two CTQs, then shifted to normal once a 200‑piece window hit FPY ≥98.5% and short‑term capability ≥1.4, which kept flow moving without swelling containment. Small caveat: if “AQL 0.65” balloons your sample on minors, park those on S‑4 and reserve c=0 for true criticals. @OP the Annex A switching rules map cleanly to this if you document the window trigger.
For a 42s takt pilot, I’ve had better luck layering a Dodge CSP‑1 continuous sampling plan over your ISO 2859‑1 approach rather than carving lots. We run 100% on the CTQ for the first about 50 pieces after each changeover, then switch to “1‑in‑k” (k set from recent defect rate); any fail snaps back to screening, which keeps containment tiny without choking flow. Caveat: lock down defect-class rules and MSA at line speed so the snap‑back isn’t noisy.
I’d layer a one‑sided CuSum on the defect rate for your criticals and use it to flip between tight reject‑on‑first screening and reduced sampling as capability stabilizes — quick to react at your cadence without building a buffer. Handy primer if you haven’t used CuSum: 6.3.2. What are Variables Control Charts?; small caveat — be explicit about the reset/reaction plan so one signal snaps you back to the tighter gate, like easing off the gas once cruise control locks.