Production Line Restart Yield

A machine stops for maintenance and then resumes; it may turn out units immediately, but only a portion pass first-time quality checks. Production line restart yield measures the proportion of output made during a declared restart window that meets requirements without rework.

Name the restart, because a planned cleaning stop, breakdown, safety shutdown and product switch can have different causes and should be segmented, and set the window, since the first batch, first 30 minutes or first 100 units give different results and the choice should be made before viewing the outcome. APQC defines finished-product first-pass quality yield as the share meeting specifications at a final test point, and restart yield applies a similar quality idea to output after a line resumes.

The Lean Enterprise Institute discusses reaching stable production after a start, so running again does not by itself mean the line is at normal rhythm. Throughput and conformance should be tracked separately.

Choose the denominator as all units completed in the restart window, including failures, which forms a clear unit-based basis, and choose the numerator as units that meet the approved specification at the defined test point without rework or repeat processing. Do not count quarantine as good, since untested or held units are pending disposition and a provisional rate should say so.

Check rework, because a repaired unit may be saleable but did not pass the first time under this definition, and track scrap separately, since restart output can include both scrap and recoverable rework so scrap alone cannot measure yield. Use a stable test point, because moving the quality check upstream or downstream changes comparability, and identify batch links so each unit connects to a restart event and its own quality decision.

Include delayed rejects: if a unit fails the same first-pass test after the shift, attribute it to its recorded restart cohort. Capture setup conditions, as temperature, calibration, tool wear and material lot can influence early output, and investigate defects, since a restart-related calibration issue and a bad material lot lead to different corrective actions.

Separate downtime, because a high yield after a very long restart delay may still be costly, so report both measures, and review first-good, since one passing unit does not guarantee the next hundred will pass and the window should capture the practical stabilisation phase. Avoid cherry-picking by not starting the window only after the first defects have been discarded, and compare like assets, since a packaging line and chemical process can have very different startup validation requirements.

Use equivalent units in continuous processes, where weight or volume may be more meaningful than item count, and state the conversion basis; if a sample represents a batch, say how the batch outcome is assigned and what remains untested. Show small samples, because a restart that produces five units can show a dramatic percentage shift from one defect, so pair the rate with the count, and track repeated restarts as distinct events if each triggers a new declared window.

Separate planned engineering trials, which may use different specifications, and pair restart yield with normal-run yield, since the gap can reveal whether startup work needs attention. Protect safety, as a restart following a safety stop requires the approved clearance and operating procedure even if waiting worsens the metric, and keep the metric subordinate to the approved quality and safety plan, not a shortcut around it.