Capacity planning

Why it matters

A schedule that ignores capacity is a wish list. Capacity planning decides how many machines, people and shifts are needed, checks whether the master schedule and MRP orders can actually be produced, and shows which work centre is the bottleneck. Too little capacity means late orders and lost sales; too much ties up capital in idle equipment.

Key ideas

Capacity is the maximum rate of output of a facility, measured in output units (cars per day, tonnes per month) or, for mixed products, in input units (machine-hours or labour-hours available per period).

Three levels of capacity.

• Design (theoretical) capacity – the maximum output under ideal conditions, as designed.
• Effective capacity – the maximum output possible given product mix, set-ups, scheduled maintenance, breaks, quality standards and similar planned allowances. It is always ≤ design capacity.
• Actual output – what is really produced, after unplanned losses (breakdowns, absenteeism, material shortages, rework). Always ≤ effective capacity.

Utilisation and efficiency.

• Utilisation = actual output ÷ design capacity.
• Efficiency = actual output ÷ effective capacity. Efficiency can look good while utilisation is poor, because effective capacity already excludes planned losses.

Capacity cushion = (capacity − average demand) ÷ capacity. Plants with uncertain or variable demand, or many products, keep a larger cushion; continuous process plants keep a small one because idle capacity is very expensive.

Hierarchy of capacity planning (matches the planning hierarchy).

• Resource requirements planning (long range) – checks the aggregate plan against plant, equipment and workforce; leads to expansion decisions.
• Rough-cut capacity planning (RCCP) (medium range) – checks the MPS against critical work centres using overall factors (CPOF), bills of capacity or resource profiles.
• Capacity requirements planning (CRP) (short range) – takes the MRP planned and released orders, routings and set-up/run times, and builds a time-phased load profile for every work centre.
• Input/output control (very short range) – compares planned with actual input and output at each work centre to control queues.

Responding to a load imbalance. Increase capacity (overtime, extra shift, alternate routing, subcontracting, more operators) or reduce load (reschedule orders, split lots, change the MPS). Underloads are handled by pulling work forward or reducing hours.

Long-term expansion strategies.

• Lead – add capacity ahead of demand; protects service and market share, risks idle capacity.
• Lag – add capacity after demand has grown; high utilisation, risks lost sales.
• Match (average) – add in small steps around the demand trend. Economies of scale, bottleneck balancing, and break-even and decision-tree analysis are used to choose the size and timing of increments.

Bottleneck. The resource with the highest load relative to its capacity limits the output of the whole system; an hour lost there is an hour lost for the plant. Adding capacity elsewhere does not raise output.

Formulas

Utilisation = actual output / design capacity Efficiency = actual output / effective capacity Actual output = design capacity × utilisation = effective capacity × efficiency

• All in units per period (or hours per period); results as fractions or %.

Capacity cushion = (capacity − average demand) / capacity

• Fraction or %.

Available hours per machine = days × shifts × hours per shift × efficiency (or availability)

• In hours per period; efficiency or availability is a fraction.

Number of machines N = (Q_good / (1 − s)) · t / H_avail

• Q_good = good parts required per period (units), s = scrap fraction, t = standard time per part (h/unit), H_avail = productive hours per machine per period (h). Round N up to the next whole machine.

Load on work centre j (period t) = Σ (set-up time + quantity × run time per unit) over all orders at j in t Load (RCCP, overall factors) = Σ MPS_i × total hours per unit of i × historical share of work centre j

• Compare load (h) with available capacity (h).

Worked examples

Example 1 (standard) – utilisation, efficiency and machines required. (a) A line has design capacity 1000 units/week, effective capacity 850 units/week and actual output 800 units/week.

1. Utilisation = 800/1000 = 80 %.
2. Efficiency = 800/850 = 94.1 %. (b) A part is needed at 60,000 good pieces per year. Standard time 6 min per piece, expected scrap 4 %, plant works 250 days × 2 shifts × 8 h, machine efficiency 85 %.
3. Pieces to start = 60,000/(1 − 0.04) = 62,500 pieces.
4. Hours required = 62,500 × (6/60) = 6250 h.
5. Productive hours per machine = 250 × 2 × 8 × 0.85 = 3400 h.
6. N = 6250/3400 = 1.84, so 2 machines are needed (utilisation about 92 %).

Example 2 (GATE level) – rough-cut check by overall factors. Product A needs 0.8 h and product B needs 1.2 h of total direct labour per unit. Historically 60 % of hours fall on work centre WC1 and 40 % on WC2. Weekly capacity: WC1 = 100 h, WC2 = 70 h. MPS for weeks 1–3: A = 100, 120, 80 units; B = 50, 60, 90 units.

1. Total hours per week = 0.8 × A + 1.2 × B.
   • Week 1: 0.8 × 100 + 1.2 × 50 = 80 + 60 = 140 h.
   • Week 2: 0.8 × 120 + 1.2 × 60 = 96 + 72 = 168 h.
   • Week 3: 0.8 × 80 + 1.2 × 90 = 64 + 108 = 172 h.
2. WC1 load (60 %): 84.0, 100.8, 103.2 h. WC2 load (40 %): 56.0, 67.2, 68.8 h.
3. Compare: WC2 is within 70 h every week. WC1 is within 100 h in week 1 but overloaded by 0.8 h in week 2 and 3.2 h in week 3.
4. Since WC1 has 16 h spare in week 1, the scheduler can pull about 4 h of work forward or authorise 4 h of overtime; the MPS is then feasible. Over three weeks WC1 total load = 288 h ≤ 300 h available, so the problem is timing, not total capacity.

Common mistakes

• Swapping utilisation and efficiency, or dividing by the wrong capacity.
• Forgetting scrap: machines must process the good quantity divided by (1 − s), not multiplied by (1 + s) (the second is only an approximation).
• Rounding the number of machines down; capacity must be at least the requirement.
• Mixing minutes and hours, or days per year with shifts per day.
• Treating total load over a horizon as enough when individual periods are overloaded.
• Adding capacity at a non-bottleneck station and expecting system output to rise.

For GATE PI

Expect short numericals on design and effective capacity, utilisation and efficiency, number of machines needed with scrap and efficiency, and load versus capacity at work centres for a given schedule. Conceptual items cover RCCP vs CRP, lead/lag strategies and bottlenecks. Practise unit conversions and always round machine counts up.

Quick check

1. Effective capacity 450 units/day, actual output 405 units/day, design capacity 500 units/day. Find utilisation and efficiency.
2. 9000 good parts are needed per month with 10 % scrap. How many parts must be started?
3. A machine has 400 h per month available and a job needs 1500 h. How many machines are needed?
4. Which capacity check uses MRP planned orders and routings?

Answers: 1. 81 % and 90 %. 2. 9000/0.9 = 10,000 parts. 3. 1500/400 = 3.75, so 4 machines. 4. Capacity requirements planning (CRP).