12.6 Capacity Types, Safety Capacity, and Labor Scheduling

The Four Capacity Definitions

"Capacity" on its own is never an answer on this exam. Domain VII names four distinct measures, and items turn on which one a planner should load against.

Worked calculation

A grinding cell has 2 machines on one 8-hour shift, 5 days a week.

Available capacity = 2 x 8 x 5 = 80 hours.

Utilization is the fraction of scheduled time the resource is actually working; the loss is breakdowns, absence, material outages, and meetings. Here it is 90%. Efficiency is standard hours earned divided by hours actually worked - how the pace compares to the standard. Here it is 95%.

Rated capacity = 80 x 0.90 x 0.95 = 68.4 standard hours per week.

Now the history. Over the last five weeks the cell earned 66, 62, 65, 63, and 64 standard hours: 320 / 5 = demonstrated capacity of 64.0 standard hours.

Load against 64.0. Demonstrated capacity is measured output; rated capacity is a calculation resting on two assumptions, and the utilization and efficiency factors in most enterprise resources planning (ERP) systems were set once at implementation and never revisited. When rated exceeds demonstrated, the factors are stale - the shop is not underperforming, the arithmetic is optimistic. That is why demonstrated capacity usually wins the exam question: it requires no assumption to be true.

The 4.4-hour gap is not rounding. Load 68.4 hours a week into a cell that proves 64.0 and you inject 4.4 hours of backlog every week - 44 hours after ten weeks, more than a full week of queue - which surfaces as lead-time growth that no dispatching rule can fix. Use rated capacity only where there is no history: a new work center, a new part with no run experience, or a shift-pattern change that makes past output non-comparable.

Safety Capacity and the Capacity Cushion

Safety capacity is capacity deliberately left unloaded to absorb variability, preventive maintenance, rework, engineering trials, sample orders, and demand spikes. Keep the two related ASCM terms distinct: protective capacity is the extra capacity carried at non-constraint resources above the constraint’s rate so they can recover and keep the constraint fed, while a capacity cushion is that reserve expressed as a percentage — 100% minus planned utilization. Plan a resource to 100% and the first breakdown, quality hold, or hot order has nowhere to go: it becomes queue.

The reasoning is queueing, not accounting. Waiting time at a resource scales with the utilization factor rho / (1 - rho), where rho is planned utilization:

Between 85% and 98% loading the queue multiplier grows roughly 8.6-fold (49.00 / 5.67) while output rises about 15%. That is the entire argument: a plant loaded at 98% has far longer lead times than one loaded at 85%, and because the relationship is non-linear, the last few points of utilization are the expensive ones. Capacity cushion = 100% minus planned utilization; the right cushion grows with process and demand variability and shrinks with equipment reliability and short changeovers.

This is also why the theory of constraints subordinates non-constraints. The constraint is protected with a time buffer; every other resource is given protective capacity so it can catch up after a disruption and keep the constraint fed. A non-constraint loaded to 98% is no longer protective - it becomes an interactive constraint, and system throughput falls while every local efficiency report still looks excellent.

Process Flow Scheduling

Discrete plants schedule work order by work order through a routing. Process industries - chemicals, food, pharmaceuticals, pulp and paper, refining - cannot, because the plant is not a set of independent work centers but a connected process train of stages that must run together. Process flow scheduling builds the schedule stage by stage from the structure of the process rather than order by order, and it operates in batch, discrete, or continuous mode.

Its defining features:

• The plant is modelled as process stages and units. The schedule is generated first for the stage that governs flow - typically the constraint or the finishing stage - and then propagated to feeding and receiving stages.
• Scheduling is by rate and run length, not by individual order. The question is "how many hours of grade B do we run," not "which of these 40 orders goes next."
• Campaign scheduling groups all demand for a grade into a single run so that cleanouts, purges, and transition yield loss are incurred once instead of repeatedly.

Run-length economics is order-quantity logic in another costume. If a grade change costs $4,000 in cleanout, yield loss, and downtime, a four-grade campaign wheel turned every two weeks incurs 26 x 4 = 104 changeovers a year ($416,000) and carries low cycle stock; turning the same wheel every four weeks incurs 13 x 4 = 52 changeovers ($208,000) but roughly doubles cycle stock. Changeover cost sets the wheel length; the fixed sequence around the wheel - light to dark, low to high viscosity, allergen-free before allergen - is set by the setup-matrix logic from section 12.5.

Labor Scheduling

Machine capacity is a ceiling. Qualified labor is usually the binding constraint. Staffing recommendations come from three inputs: human resources policies (shift patterns, overtime limits and premiums, collective agreements, rest and break rules, maximum consecutive days worked), the available labor pool (headcount, absence rate, vacation calendar), and a labor skills matrix.

A skills matrix records who is qualified on which operation:

The coordinate measuring machine (CMM) has two shifts of machine availability - 16 hours - but only Dube is certified, so labor-limited capacity is 8 hours, and on the day Dube is absent it is zero. Rated capacity computed from machine hours would report 16 and the schedule built on it would be fiction. Always calculate rated capacity against the binding resource, which is frequently a person rather than a machine.

Two levers close that gap, and they run on different clocks:

• Cross-training creates flexible capacity permanently. Qualify Chen on the CMM and the resource gains a second shift plus absence cover. It takes weeks and consumes capacity while training is under way, so it must be planned ahead of the peak, not during it.
• Contract labor is a surge lever for a known peak. It is quick to contract and slow to become useful: onboarding, safety training, and part or process qualification create a qualification lag, so contract labor rarely relieves a certified or constrained operation in the current week. Use it to backfill general unqualified work and redeploy your certified people onto the constraint - subordination applied to people.