A kitchen that can prepare 100 lunches in a day does not necessarily need to prepare 100. Receiving 65 requests at noon does not mean it can deliver 65 lunches within one hour, either.

Operational capacity connects resources, time, work sequence, and quality requirements to a promise the business can fulfill. Adding machines or headcount is not enough: the end-to-end flow has constraints, and some activities share resources. MIT OpenCourseWare uses process flow diagrams, capacity and utilization calculations, and bottleneck analysis to examine waiting times and resource limits [R01]. Sales and operations planning calls for matching demand with available resources before making commitments [R02]. Here we develop a starting calculation for a small business rather than present a universal optimization model.

1. Distinguish nominal, effective, available, committed, and achieved capacity

Use the same units and time horizon before comparing figures:

Concept Question it answers Hypothetical daily example
Nominal or theoretical capacity How much could the system produce under ideal conditions without interruptions? 100 lunches
Effective capacity What is plausible after planned setup and cleaning? 80 lunches
Capacity available to promise How much can we confirm while retaining our chosen buffer? 72 lunches
Committed capacity How many slots are already reserved by confirmed orders? 40 lunches
Achieved fulfillment How many units were completed under the agreed conditions? Recorded at close, not assumed

These terms may be defined differently in different industries; always state the formula. Committed capacity is not another engineering limit: it is part of existing capacity already reserved. Achieved output is not the maximum capacity either. Low sales can leave resources idle even when the system works correctly.

A unit might be lunches/day, orders/hour, appointments/shift, or active contracts per specialist. Do not compare “30 customers” with “30 orders” if each customer requests multiple units.

2. Identify the bottleneck in the full workflow

Map the stages from order to fulfillment and calculate each relevant resource's rate. Imagine three independent stations dedicated solely to this product line, each staffed by one person:

To identify which activity truly limits delivery, map the complete operating workflow, not merely the speed of one machine or person.

Stage Cycle time per lunch Theoretical capacity per hour
Preparation 3 minutes 60/3 = 20
Packing 2.5 minutes 60/2.5 = 24
Dispatch 2 minutes 60/2 = 30

In steady operation, this sequence cannot exceed its slowest stage's rate: 20 lunches/hour, under the stated assumptions. Do not add 20 + 24 + 30; these are sequential stages, not three independent lines producing complete lunches. If the same employees alternate between preparation and packing, actual capacity could be lower and must be measured differently. If delivery relies on one vehicle with a slow route, the bottleneck may move outside the kitchen.

Bottlenecks can change with the time of day, product mix, breakdowns, or order types. A three-minute task does not mean an entire order arrives at the customer three minutes later: queues, coordination, batches, and transportation take time.

3. Calculate capacity using time, availability, and yield

Assume a nominal 5-hour shift during which the preparation line could sustain 20 lunches/hour:

If the calculation reveals missing equipment, staff, or essential skills, document the gap with the resource and capability inventory before committing investment.

Nominal capacity = 5 h × 20 lunches/h = 100 lunches.

One of those hours is used for station setup, cleaning, and other scheduled activities, leaving 4 effective hours: 4 × 20 = 80 lunches. The team decides, illustratively, to reserve 10% of those 80 units for contingencies: 80 × (1 − 0.10) = 72 lunches available to promise that day. The 10% is a decision within this case, not a standard or universally optimal utilization level.

If a process produces scrap or rejects, do not treat every started unit as sellable. For example, if a different machine makes 50 parts/hour and only 96% pass inspection, expected conforming output is 50 × 0.96 = 48 conforming parts/hour, under stable conditions. Do not apply the same availability loss or scrap rate twice.

Check the time horizon: 72 daily units do not authorize 72 deliveries at 12:30. Verify capacity for each time slot and for transportation, not just the full day.

4. Compare demand, utilization, and fulfillment for each period

If capacity available to promise is 72 lunches/day, with 40 confirmed orders distributed across feasible windows, committed load is 40/72 = 55.6% of that capacity. If effective capacity of 80 is the denominator, effective utilization would be 40/80 = 50%. Neither figure proves economic efficiency: their denominators and meanings differ.

A promise-fulfillment rate could be orders complete and on time / orders evaluated for fulfillment, with explicit rules for cancellations and changes. If 38 of 40 orders are fulfilled completely and on time, the rate is 38/40 = 95%, assuming the original 40 promises were still active. Do not hide the two failures by changing the denominator afterward.

For an indivisible service such as a professional appointment, calculate available time blocks and the duration of each visit type. For a digital app that requires human support, the number of concurrent users and the number of cases the team can resolve are different constraints.

5. Understand why 100% utilization can damage service

Variability does not behave like an average: two orders arrive at once, an employee must resolve a problem, or a delivery takes longer than planned. When every resource is committed with no spare capacity, a small disturbance can generate queues or overtime. MIT material on capacity and queueing illustrates that waiting times can increase nonlinearly as utilization approaches a limiting rate in systems with variability [R03]. This does not mean every business has the same “safe” percentage.

It depends on whether you can prepare in advance, store safely, reschedule, distribute work, or tolerate waiting. Automated software can scale transactions faster than a kitchen handling perishables, but both may depend on critical services. A capacity buffer is a service, risk, and cost decision, not automatically wasted capacity.

Do not treat a queueing formula as precise without checking its assumptions. A small team can begin by observing waiting times, work backlogs, and cases that required additional help.

6. Account for cost, contribution, and unused capacity

Available capacity is not the same as profitable capacity. For each sold unit, calculate price and variable delivery cost. Suppose, hypothetically, the price is S/ 40, the total variable cost per sold lunch is S/ 27, and unit contribution is S/ 13. For 40 sales, total contribution is 40 × 13 = S/ 520, before fixed expenses and other items.

A shift may incur fixed costs even if no customers arrive. But producing in advance just to keep the kitchen busy can be worse: spoiled ingredients, wasted preparation, storage requirements, and quality risks. Distinguish the cost of keeping resources available from the cost of processing materials without confirmed demand.

When comparing decisions, include additional shift costs, commissions, overtime, support capacity, and cash timing. Do not confuse greater volume with greater profit. Detailed contribution-margin and break-even calculations belong in the Atlas's existing finance article; here we apply them to an actual operating-capacity decision.

7. Respond differently when demand is below, near, or above capacity

Situation Initial operating response What must be checked first
Demand below the limit Schedule to orders, reduce spoilage, preserve reasonable availability Shift cost, preparation lead times, minimum service
Demand near the limit Close or limit slots, protect contingency capacity Order variability, staffing, reservations, quality
Demand above the limit Offer another slot, a waitlist, feasible outsourcing, or transparent refusal Customer consent, total cost, third-party control, deadlines

Rescheduling need not breach a promise if the person knows and accepts the new terms before the order is confirmed. Unilaterally changing an accepted promise is different and may require correction, a refund, or another remedy depending on agreed terms and applicable law. A wait the customer did not accept is not extra capacity.

Before stimulating more orders, align the marketing channel plan with verified slots and available delivery windows.

The right response may vary by segment: a recurring B2B contract requires capacity reservation for future commitments, while an occasional sale may depend on today's slots. Document why a commitment has priority instead of deciding solely based on who complains loudest.

8. Choose buffers according to risk, not a magic number

Buffers can take different forms: uncommitted units, time between appointments, backup supplies, alternative suppliers, or a rescheduling process. Consider variability, criticality, ability to recover, carrying cost, and consequences of failure.

If waiting and flexible appointments are acceptable, less unused capacity may be required. If health, safety, perishables, or critical processes are involved, protective controls and allowances must be justified by specific obligations and risks. Do not extrapolate the kitchen's hypothetical 10% reserve to a health clinic or computing infrastructure.

Make the rule verifiable: “Reserve up to 8 units of daily capacity for contingencies; release that capacity only after authorized review.” Revisit it when new data or risks emerge. Do not hide the reserve from the sales function: everyone needs a shared view of availability.

9. Compare alternatives under multiple criteria and constraints

Do not evaluate an operating alternative only by how much it can produce. Consider orders actually requested, expected contribution, investment, quality, fulfillment time, slot capacity, risk, and liquidity. First eliminate options that violate mandatory conditions; then compare the feasible choices.

Illustrative alternative Possible advantage Cost or risk Condition before considering it
Prepare 65 lunches without confirmed orders Production is ready earlier Waste and cost of unsold units Sufficient demand evidence and safe preservation
Prepare 40 against orders Less waste Requires schedule and confirmation discipline All 40 can be delivered in agreed windows
Open later windows Spreads the load Extra staff and operating hours Customers accept and resources are available
Outsource delivery May release a logistics constraint Commissions, quality, coordination Service quality and costs verified by zone and window

Do not turn a decision matrix into artificial numerical certainty. When two options depend on uncertain demand, document scenarios and the information that could change the choice. Formal optimization and long-term scaling decisions belong in Pillar 7.

10. Set up daily and weekly load control

At minimum, track time slot, sellable capacity, confirmed orders, remaining slots, waiting orders, actual delivery time, incidents, rejected units, and reason. Review upcoming commitments daily and patterns weekly. If one window concentrates declined orders, investigate whether the restriction comes from preparation, packing, delivery, or an overly narrow sales promise.

An operating alert must lead to an action: close reservations for a slot, warn the commercial team, propose another time, or request approval for additional help. At close, document the decision and observed outcome. Avoid using the weekly average alone: it can hide four quiet days and one impossible lunch hour.

Worked example: capacity for 100, demand for 40, and a peak of 65

Hypothetical assumptions, not actual restaurant data. During a nominal five-hour shift, the line could start 100 lunches (20 per hour). Four effective hours allow up to 80, and the business reserves eight units for contingencies, leaving daily available capacity of 72. Normal confirmed demand is 40 lunches/day spread over feasible slots. A promotion attracts requests for 65 lunches during the 12:00–13:00 window.

Although 65 is below 72, it does not fit in that hour: preparation can sustain 20 lunches/hour under the assumptions. Without verified safe pre-preparation and additional fulfillment capacity, the gap is 45 units in that window. The business must not confirm all 65 deliveries for that hour. It offers later windows with validated capacity and explicit acceptance, limits advertising, or records unmet inquiries; it does not assume all would convert into sales.

Now consider a separate decision from the peak: the company prepares 65 units in advance but sells only 40. If the variable cost of each sold unit is S/ 27, and the already incurred ingredient and preparation cost per leftover unit is S/ 19, there are 25 unsold units, with an illustrative cost of 25 × 19 = S/ 475. Contribution from the sold units is 40 × (40 − 27) = S/ 520; subtracting the waste leaves S/ 45 before fixed expenses. If fixed expenses allocated to the day were S/ 400, the simplified outcome would be −S/ 355. By contrast, producing and selling only 40 units under the same assumptions would produce S/ 520 − S/ 400 = S/ 120. This is not a formal income statement: taxes, unexpected expenses, and other possible costs are omitted.

Lesson: making 65 was not better merely because it approached maximum output. The objective was to serve real orders in promised windows with viable economics. The business may prefer 40 reliable deliveries and investigate peak demand before assuming new fixed costs. An operational maximum does not guarantee an economic maximum.

Template: capacity and order-admission calculator

Input Formula or record Case example
Nominal time and limiting rate hours × units/hour 5 h × 20 = 100
Effective hours nominal − scheduled losses 5 − 1 = 4 h
Effective capacity effective hours × limiting rate 4 × 20 = 80
Chosen reserve units justified by risk 8
Daily available capacity effective − reserve 72
Per-slot capacity bottleneck rate × effective hours in the slot Hypothetically 20 per hour
Committed load confirmed orders compatible with the slot 40/day across feasible windows
Uncommitted daily headroom available capacity − commitments 72 − 40 = 32
Declined or rescheduled commitments quantity, reason, customer acceptance, owner Record without assuming a sale

Do not subtract 32 daily slots from the limit of a single hour: the time unit must match. If another resource becomes more restrictive, replace the assumed available capacity with the documented real limit.

Common mistakes and limits

Adding rates of sequential stages; using daily capacity to promise one time window; confusing maximum capacity with actual orders; assuming employees never take breaks; counting the same loss twice; filling the line with output nobody ordered; comparing revenue without costs; imposing a universally “optimal” utilization rate; and removing backup capacity without understanding the consequences of a failure.

What you should have by the end

A map of stages and bottlenecks, an auditable capacity estimate for the day and each window, a buffer rule, a way to control committed orders, and documented choices for peaks. You will be able to decide how many promises to make without confusing busyness with sustainable value creation.

References