Carry the capacity gap forward, not the machine-speed claim

Part 1 converted monthly sellable demand into a required hourly output. Its illustrative example required 31,731 qualified pieces per hour, while the selected planning record used 24,000 qualified packed pieces per hour. The resulting gap was 7,731 pieces per hour.

A common mistake is to apply a generic OEE percentage to the 24,000 figure again. That would double-count losses because the number already represents qualified packed output. Today’s task is the reverse: open that combined result and identify where the missing output actually went.

Set one measurement boundary before calculating OEE

OEE combines availability, performance and quality, but the calculation is meaningful only when every team uses the same start and end points. Decide whether the counted good unit is accepted at final cutting, after automatic rejection or after packing. Do not mix these boundaries in one report.

Actually, packing should remain a separate constraint when the OEE boundary ends at the main line. Otherwise a slow bagging or stacking process can appear as poor machine performance, while a full downstream buffer can force the line to slow even though the drive system is healthy.

• Freeze one SKU, material set and pack format for each record.
• Define planned production time after approved breaks and planned maintenance.
• Count good units at one named quality gate.
• Record packing release separately if it sits outside that gate.

Availability loss: find where scheduled time disappeared

Availability asks how much planned production time became actual runtime. A 40-minute stop caused by a material wait and a 40-minute mechanical repair have the same duration, but they do not have the same countermeasure. The event code matters as much as the clock.

Record the stop start, stop end, station, reason code and recovery action. Planned cleaning, SKU changeover and preventive maintenance should be visible but separated from unplanned breakdowns. If every stop is coded “machine fault,” the loss tree will send investment toward the machine even when materials or scheduling caused the wait.

Performance loss: separate reduced speed from short stops

A line may show high runtime and still miss output because it spends the shift below the approved benchmark. Performance loss includes deliberate reduced-speed operation and micro-stops too short to appear in a manual downtime log.

In most cases, the useful benchmark is not catalogue maximum speed. It is the validated rate for the same SKU, BOM, pack format and quality boundary. Compare actual gross count during runtime with that benchmark, then group losses by station and duration. Five seconds repeated hundreds of times deserves more attention than one visible ten-minute stop.

Quality and packing losses: count what the warehouse can release

Quality loss is the difference between total units produced and units accepted at the agreed gate. Classify rejects by defect and operating phase: normal run, startup, restart, material splice or size change. A single waste percentage hides whether the problem belongs to process stability or changeover recovery.

The real question is whether accepted products leave the end of line at the same rhythm. Record stacker waits, bag shortages, manual handling, full buffers and blocked discharge separately. Packing is not a footnote. It is the last capacity boundary before a good product becomes a sellable packed product.

A worked OEE decomposition that reconciles to 24,000 pieces per hour

Assume the approved gross benchmark for one SKU is 30,000 pieces per scheduled hour. The shift record shows 90% availability, 95% performance and 97% quality. Standard OEE is therefore 82.94%. A separate packing release factor of 96.5% brings end-to-end output to approximately 24,010 qualified packed pieces per scheduled hour.

This decomposition does not solve the 7,731-piece hourly gap by itself. It shows which layer owns each part of the loss. Management can then test recoverable hours and output before approving another shift, retrofit or production line.

Build one loss sheet by shift and SKU

Daily averages conceal changeover and product-mix effects. Use one row per shift and SKU, then reconcile gross count, rejects and packed release. The first objective is not a perfect dashboard. It is a record that production, maintenance, quality and packing can all sign.

The acceptance and ramp-up framework can supply the same discipline during FAT, SAT and early production. One definition should follow the project from supplier test to normal shift reporting.

Decision gate: recover capacity before buying capacity

The loss tree should end with four outputs: the constraint station, recoverable hours, recoverable qualified output and the unresolved capacity gap. If removing the largest verified loss closes the gap, another line is premature. If the gap remains after realistic recovery, the investment case becomes stronger.

Part 3 will use this result to test shift allocation, campaign length and changeover sequence. Hardware should enter the discussion only after operational recovery has a measured limit.

Related machinery planning routes

Conclusion: one percentage is not a bottleneck analysis

OEE becomes useful when it explains the output gap. Availability identifies lost runtime, performance identifies lost pace, quality identifies rejected output, and the packing record shows whether good products actually reach a sellable format.

If your best shift repeated for an entire month, would it close the capacity gap calculated in Part 1? If yes, the next problem is how to reproduce that shift. If no, the remaining gap belongs in the retrofit-versus-new-line decision.

FAQ