production-scheduling

Rank work centers by utilization ratio (load hours divided by available hours) over the trailing two weeks by shift. Verify causally: if adding one hour of capacity at that work center increases total plant output, it is the true constraint, not just a WIP accumulation point.

First determine if the broken machine is the constraint and estimate repair time. Freeze jobs in process or within 2 hours of start, re-sequence remaining jobs by priority (past-due, constraint-feeding, then EDD), communicate within 30 minutes, and lock the schedule for 4 hours.

Use campaigns when changeovers exceed 60 minutes or $500 in scrap and lost output, since fewer changeovers outweigh higher WIP. Use mixed-model when changeovers are under 15 minutes or customers demand short lead times. Compute the crossover where marginal changeover cost equals marginal carrying cost.

Use Earliest Due Date for high-variety short-run mixes to minimize maximum lateness, Shortest Processing Time for long-run few-product environments to minimize flow time and WIP, and setup-aware EDD when sequence-dependent setups exist.

MRP assumes infinite capacity, fixed lead times, and perfect BOMs, so its output is not executable. Run it through finite-capacity scheduling that respects machine availability, shift patterns, sequence-dependent setups, and yield losses before releasing to the floor.

World-class OEE is 85% or higher, while typical discrete manufacturing runs 55-65%. If constraint OEE drops below 65% for three consecutive shifts, trigger a focused improvement event with maintenance, engineering, and scheduling.