Quality workbench — Season 4 Rollout

One chart per metric with two or more dated readings as of 2026-07-01: the mean, ±3σ control limits worked out from the readings themselves, and the rule-of-seven signal — seven readings running on the same side of the mean, a likely process shift even while every point stays inside the limits.

Colour QC rejects per reel

Only 1 reading so far — a control chart needs at least two to estimate a spread.

Every out-of-tolerance reading as of 2026-07-01, grouped by the metric it measures and ranked largest first, with the "vital few" categories that together reach 80% of the total.

Vital few: Colour QC rejects per reel — fixing these first clears 80% of the failures.

Sample size for estimating a proportion from a finite population, worst-case p=0.5 so the plan never undersizes, finite-population-corrected. GET only; nothing typed here is saved.

Trace one open issue to its likely cause: a fixed six-category (method, machine, material, people, measurement, environment) fishbone shape to fill in by hand, plus a five-whys chain. Nothing here is saved.

Conformance spend (prevention + appraisal) against non-conformance spend (internal + external failure). The budget's own cost categories (labour, materials, services, travel, contingency) carry no conformance/non-conformance split, so there is nothing stored to read this from — type the four bands by hand below; nothing typed is saved.

Read straight off the project's own quality management plan, one line per sentence — a reading list for the audit, never a pass/fail record; nothing here is scored.