Capacity, Throughput & Utilisation Analysis
Stand near a busy pizza counter on a Saturday night: ovens are hot, riders are waiting, orders are flashing, and yet customers still see delays. The problem is not always “too little effort” - it is often one narrow step where capacity, throughput and utilisation stop matching each other.
- Capacity is the maximum sustainable output possible in a defined time under stated conditions.
- Throughput is the actual completed output leaving the system per unit of time.
- Utilisation = actual output or busy time ÷ available capacity; high utilisation can signal efficiency or hidden congestion.
- The bottleneck is the step with the lowest effective capacity; it sets the system’s maximum throughput.
- Do not average capacities across steps. In a process, the weakest stage governs the whole flow.
- Good analysis separates design capacity, effective capacity and actual output.
- The interview-safe answer: map the process, calculate capacity by step, locate the bottleneck, quantify utilisation, then recommend a fix.
Think of operations as a pipeline. Demand enters from one side, but completed output exits only as fast as the tightest stage allows. That is why a business can have idle people in one area and long queues in another at the same time.
The Big Idea: Capacity Is Potential, Throughput Is Reality
Capacity analysis answers one practical question: “How much can this system realistically deliver, and what is stopping it from delivering more?” It is used in factories, hospitals, restaurants, airlines, call centres, dark stores, consulting teams and SaaS support desks.
The core distinction is simple but interview-critical:
In one line: capacity tells you what could happen; throughput tells you what actually happened; utilisation tells you how heavily the system was loaded.
Definitions You Can Say in One Breath
- Capacity: Maximum sustainable output a process can deliver in a period under stated operating conditions.
- Throughput: Actual rate at which completed units exit a process.
- Utilisation: Actual output or busy time divided by available capacity.
- Effective capacity: Design capacity adjusted for downtime, staffing, product mix, setups and quality losses.
- Bottleneck: The process step with the lowest effective capacity, limiting total system output.
The Capacity-Throughput-Utilisation Framework
Use this five-step framework whenever you see a process, queue, plant, service operation or resource-planning question.
Key Metrics and What Good Looks Like
Do not discuss capacity with vague words like “good”, “low” or “busy”. Use numbers. The right benchmark depends on the industry, but the logic below is safe for most interview cases.
The most important warning: 100% utilisation is not automatically excellent. In high-variability services, it often means the system has no breathing room. A hospital bed network, airport security lane, contact centre or restaurant kitchen operating at full load will quickly generate waits when arrivals fluctuate.
Worked Example: A Quick-Service Kitchen
Suppose a quick-service kitchen has three steps in one hour:
- Order taking: 60 orders per hour
- Cooking: 45 orders per hour
- Packing and handoff: 50 orders per hour
The system capacity is not 60, 45 or 50 averaged together. It is capped by the bottleneck: 45 orders per hour at cooking.
Interpretation: cooking is heavily loaded and controls output. Adding one more order-taking counter will not improve throughput unless cooking capacity also increases. The better move is to reduce cooking time, add cooking equipment, pre-prep ingredients, simplify the menu during peaks or shift some demand to off-peak periods.
Capacity Levers: What Managers Can Actually Change
Once you know the bottleneck, there are only a few broad ways to improve the system. Strong candidates name the lever, the trade-off and the risk.
This logic also applies to people-heavy businesses. For example, utilisation is a core economic driver in consulting firms because revenue depends on how much billable capacity is converted into client work; if that angle interests you, revise how consulting teams are staffed and utilised.
Case Study: Domino’s India and Capacity at the Dinner Rush
Domino’s India shows capacity analysis in a familiar setting: orders arrive in bursts, but throughput depends on the tight coordination of kitchen, oven, packing and delivery resources.
Situation: In food delivery, demand is lumpy. Friday evenings, weekends, rain and local events can create sharp order peaks. A store may look fully staffed, but if the oven, makeline, packing table or rider availability becomes the constraint, orders wait.
The move: Domino’s operating model is built around repeatable products, standard store processes and a delivery-focused fulfilment system. The primary driver is process standardisation: predictable recipes, defined preparation steps and repeatable kitchen routines reduce variation at the bottleneck. Supporting drivers include store-level demand forecasting, ingredient pre-preparation, delivery-zone discipline, rider coordination and technology-enabled order visibility.
The lesson: The business does not win merely by “having more staff”. It wins when each capacity node is balanced against the others. Extra riders do not help if pizzas are waiting in the oven queue; extra oven capacity does not help if packing and handoff are slow. The manager’s job is to protect the bottleneck and make the whole flow stable.
How AI Changes Capacity, Throughput & Utilisation Analysis
AI does not remove the need for capacity logic. It makes the measurement faster, more granular and more predictive.
- ML demand forecasting: AI models can predict demand by hour, location, weather, event and promotion. This helps managers schedule capacity before the rush instead of reacting after queues form.
- Digital twins and simulation: Operations teams can simulate process changes - such as adding one packing station or changing rider dispatch rules - before spending money on physical capacity.
- Computer vision and sensor analytics: In factories, warehouses and service counters, AI can detect idle time, queue length, rework loops and machine downtime more continuously than manual observation.
Use ChatGPT or Claude to practise cases: paste a process description, ask it to create a capacity table by step, then challenge its answer by asking, “Which step is the bottleneck, what assumption could be wrong, and what would you measure on-site?”
Interview Relevance
“A restaurant receives 100 orders per hour during peak time, but customers are waiting despite high staffing. How would you diagnose whether this is a capacity, throughput or utilisation problem?”
Say “effective capacity” before you say “utilisation”. It shows you understand that real operations lose capacity to downtime, mix, staffing, changeovers and quality problems.
Common Mistake
The mistake is treating high utilisation as automatically good. It costs candidates because interviewers know that near-100% utilisation in variable systems creates queues, delays and service failures. The one-line fix: always connect utilisation with variability, bottlenecks and waiting time.