Capacity Planning and Constraint Management: Matching Supply Capability to Demand Across Every Time Horizon

Capacity definitions and measurement

Capacity has three distinct definitions relevant to planning. Design capacity is the maximum output a facility or process is engineered to produce under ideal conditions. It is rarely achieved in practice and serves as an upper bound for planning. Effective capacity is the capacity that can realistically be achieved given scheduled maintenance, changeovers, breaks, quality losses, and other planned reductions. It is typically 80 to 90 percent of design capacity. Actual output is what is produced in a given period, which may be less than effective capacity due to unplanned downtime, demand shortfalls, material unavailability, or quality problems.

Utilization = Actual Output / Design Capacity

Efficiency = Actual Output / Effective Capacity

The distinction between utilization and efficiency matters for diagnosis. Low utilization can reflect either low demand (acceptable if the shortfall is demand-driven) or operational problems (downtime, quality losses). High utilization relative to effective capacity suggests a genuine capacity constraint. Efficiency below 100 percent identifies operational losses — downtime, changeover time, speed losses, quality defects — that can be addressed without capital investment.

Long-range capacity planning

Long-range capacity planning addresses decisions with multi-year lead times: facility construction, major capital equipment, long-term workforce planning, and strategic sourcing agreements that determine supplier capacity availability. These decisions require demand forecasts over a 2 to 5 year horizon, which carry substantial uncertainty. The standard approaches to managing this uncertainty are: capacity cushion (building in excess capacity to absorb demand variability — typical for process industries and utilities where capacity shortfalls are commercially costly); capacity leading (adding capacity before demand materializes, appropriate for fast-growing markets where being capacity-constrained means losing market share); and capacity lagging (adding capacity only after demand is demonstrated, appropriate for capital-intensive industries where overcapacity is financially damaging).

Long-range capacity decisions are strategic and largely irreversible. A factory built for a specific production technology cannot easily be repurposed. The financial model for long-range capacity decisions must account for the full lifecycle cost of the capacity, including utilization assumptions across the planning horizon, not just the capital cost at commissioning.

Medium-range capacity planning

Medium-range capacity planning, also called aggregate planning, covers a 2 to 18 month horizon and determines how to match aggregate supply to aggregate demand using workforce, overtime, subcontracting, inventory, and demand management levers. The aggregate planning problem is to meet a demand forecast at minimum cost while satisfying workforce and capacity constraints.

The primary levers in aggregate planning: hiring and layoffs (adjusting workforce size to match demand — effective but with lead time for hiring and training, and with real costs including severance and morale effects); overtime and undertime (varying hours to absorb demand fluctuations without changing headcount — limited by legal constraints and workforce fatigue); subcontracting (using external production capacity for peak demand periods — maintains flexibility but at a cost premium and with quality management overhead); inventory build (producing to stock in low-demand periods to smooth production and meet peak demand — applicable only to non-perishable products); and demand management (shifting demand from peak to off-peak periods through pricing or lead-time incentives — effective when customer demand is time-flexible).

Short-range capacity planning and scheduling

Short-range capacity planning covers the 1 to 8 week horizon and translates the aggregate plan into specific production schedules, machine assignments, and labor allocations. The primary tools are the master production schedule (MPS), rough-cut capacity planning (RCCP), and capacity requirements planning (CRP).

Rough-cut capacity planning checks the MPS against key resource capacities — the critical machines, labor centers, or supplier capacities that are most likely to be binding — to identify overloads or underloads before the schedule is released. It is called “rough-cut” because it uses simplified bill-of-resource data rather than the full routing and work center detail used in CRP. RCCP runs in minutes and enables rapid iteration; CRP is more detailed but computationally intensive and used for final schedule validation.

When RCCP or CRP reveals a capacity overload at a work center, the resolution options are: increase capacity (overtime, additional shift, subcontracting); reschedule orders (push some orders to a later period); reduce demand (negotiate extended delivery with the customer); or split the lot (process part of the order on the constrained resource and the remainder in the next period). Each option has cost and customer service implications that must be weighed explicitly.

Theory of Constraints: the five focusing steps

The Theory of Constraints (TOC), developed by Eliyahu Goldratt, holds that every system has at least one constraint that limits throughput — the rate at which the system generates value. Improving any non-constraint resource does not increase system throughput; it only increases idle time or inventory ahead of the constraint. TOC prescribes five focusing steps for managing constraints.

Step 1: Identify the constraint. Find the resource, policy, or market condition that limits system throughput. In a manufacturing context, this is the work center with the highest utilization relative to available capacity — the bottleneck. Identifying the constraint requires measuring actual throughput rates and queue lengths at each work center, not relying on assumptions.

Step 2: Exploit the constraint. Maximize throughput through the constraint with existing resources. This means ensuring the constraint never starves for work (always has a buffer of material ready to process), never processes defective material (quality checks upstream of the constraint prevent wasted constraint capacity), and runs the most productive mix of products (scheduling the products with the highest throughput per constraint unit first).

Step 3: Subordinate everything else to the constraint. All other resources — upstream operations that feed the constraint, downstream operations that process the constraint’s output — should be paced to the constraint’s rate. Upstream operations running faster than the constraint only build inventory ahead of it; downstream operations running faster than the constraint leave it as the pace-setter regardless. The concept of drum-buffer-rope formalizes this: the constraint (drum) sets the pace, a time buffer protects the constraint from starvation, and a rope signals upstream release rates.

Step 4: Elevate the constraint. If exploitation and subordination are not sufficient, add capacity to the constraint through investment — additional equipment, additional shifts, outsourcing, or process improvement that increases constraint throughput.

Step 5: Return to step 1. Elevating one constraint typically causes a different resource to become the new constraint. The constraint shifts, and the five-step process restarts on the new binding limitation.

Capacity buffers and variability

Variability in demand and process performance makes capacity planning more complex than simple demand-versus-capacity comparison. A work center loaded at 90 percent of capacity in an average week will regularly exceed capacity in high-demand weeks, creating backlogs that take multiple periods to clear. The relationship between utilization and queue length is nonlinear: at low utilization, queues are short and lead times are predictable; as utilization approaches 100 percent, queue lengths grow exponentially and lead times become highly variable.

This relationship — formalized in queueing theory — has a practical implication for capacity planning: operating a constrained resource at very high utilization degrades lead time performance disproportionately. A capacity cushion of 15 to 20 percent on critical resources maintains acceptable lead time performance under demand variability. The appropriate cushion depends on the coefficient of variation of demand (how variable demand is relative to its mean), the desired service level, and the cost of lead time variability versus the cost of idle capacity.

Capacity planning in service operations

Service operations face capacity planning challenges that differ from manufacturing: service capacity cannot be inventoried (an unstaffed call center hour is permanently lost), demand variability is often higher and less foreseeable, and the “product” is delivered simultaneously with consumption. The capacity planning tools for service operations emphasize workforce scheduling (matching staff levels to predicted demand patterns, often using historical arrival data and forecasting models), queuing analysis (modeling waiting time at different capacity levels), and demand smoothing (appointment systems, pricing incentives for off-peak use).

Common mistakes

• Measuring and improving every resource uniformly. Efficiency improvements at non-constraint resources do not increase system throughput. Investment in a non-constraint produces no output improvement — it only increases utilization of a resource that was not limiting the system. Resources should be prioritized by their position relative to the constraint, not by their efficiency metrics in isolation.
• Confusing high utilization with high performance. A resource running at 100 percent utilization is a capacity constraint generating queue buildup and lead time variability. It is not a sign of efficiency — it is a signal that capacity needs to be expanded or demand needs to be moderated. Organizations that celebrate high machine utilization without measuring queue length and lead time are optimizing the wrong variable.
• Building capacity plans on design capacity instead of effective capacity. A new production line with a design capacity of 1,000 units per shift does not produce 1,000 units on day one. Startup losses, changeover time, maintenance, and learning curve effects reduce actual output to a fraction of design capacity during ramp-up. Plans built on design capacity systematically overstate available output and create delivery commitments that cannot be met.
• Treating the constraint as fixed. The constraint is not an immutable property of the system; it is the current binding limitation, which changes as the system is improved. After elevating the current constraint, the next constraint emerges and requires the same analysis. Organizations that fix the constraint in their mental model stop improving when that particular resource is addressed, missing the sequential constraint-shifting dynamic that drives continuous improvement.

Frequently asked questions

¿What is the difference between design capacity, effective capacity, and actual output?

Design capacity is the maximum output a facility is engineered to produce under ideal conditions. Effective capacity is the realistic achievable output after accounting for scheduled maintenance, changeovers, breaks, and planned quality losses — typically 80 to 90 percent of design capacity. Actual output is what is produced in a specific period, which may be below effective capacity due to unplanned downtime, demand shortfalls, or material availability problems. Planning and scheduling should use effective capacity, not design capacity, as the baseline. Using design capacity in plans systematically creates delivery commitments that cannot be met once real-world operational losses are factored in.

¿What is the Theory of Constraints and how does it apply to capacity planning?

The Theory of Constraints (TOC) holds that any system’s throughput is limited by a single binding constraint at any given time. In capacity planning, the constraint is the resource with the least capacity relative to demand — the bottleneck that determines the maximum output of the entire system. TOC’s five focusing steps are: identify the constraint; exploit it (maximize throughput through it with current resources); subordinate all other resources to the constraint’s pace; elevate the constraint (add capacity if needed); and return to step 1 as the constraint shifts. The practical implication for capacity investment is that only investment in the constraint produces system throughput improvement. Investment elsewhere increases resource utilization without increasing output.

¿What is rough-cut capacity planning (RCCP)?

Rough-cut capacity planning checks the master production schedule against key resource capacities to identify overloads before the schedule is released for execution. It uses simplified bill-of-resource data — average hours per unit on critical work centers — rather than the full routing detail used in capacity requirements planning. RCCP is fast and enables rapid iteration: if a period is overloaded, the planner can shift orders between periods and recheck in minutes. It focuses on the critical resources most likely to be binding, not every work center. Its output is a period-by-period load profile showing where capacity is adequate, overloaded, or underutilized, enabling proactive schedule adjustments before orders are released.

¿Why does operating near 100 percent capacity utilization hurt lead time performance?

Queueing theory shows that average queue length (and therefore lead time) grows nonlinearly as utilization approaches 100 percent. At 70 percent utilization, a work center can absorb demand variability with short, stable queues. At 90 percent, the same variability creates much longer queues that take multiple cycles to drain. At 95 percent, queues become unstable and lead times highly unpredictable. The practical implication is that a capacity cushion of 15 to 20 percent on critical resources is not wasted capacity — it is the buffer that maintains acceptable lead time performance under demand variability. The cost of the cushion (idle time) must be weighed against the cost of lead time variability (expediting cost, customer service impacts, lost sales).

¿What are the main levers in aggregate capacity planning?

Aggregate capacity planning levers fall into two categories. Supply-side levers: hiring or layoffs (adjusting workforce size, effective but with lead time and cost); overtime or undertime (varying hours per worker without changing headcount); subcontracting (using external production capacity for peaks); and inventory build (producing to stock in low-demand periods for later consumption, applicable only to non-perishable products). Demand-side levers: demand management through pricing or lead-time incentives that shift demand from peak to off-peak periods. The optimal mix depends on the relative costs of each lever, workforce constraints, product shelf life, and customer sensitivity to lead time and price changes.

¿How do you identify the true bottleneck in a production system?

The true bottleneck is the resource with the longest queue of work waiting ahead of it — not necessarily the busiest-looking resource. To identify it: measure actual throughput rates (units per hour) at each work center; measure queue lengths (work-in-process inventory waiting to be processed) in front of each work center; and calculate the ratio of demand rate to throughput rate for each resource. The work center where this ratio is highest, or equivalently where the queue is longest and growing, is the constraint. Purely visual observation and subjective assessment are unreliable — managers systematically overestimate the importance of resources that look busy and underestimate quiet but slow-throughput resources that are the actual binding constraint.