This post is about production processes and throughput. Throughput is the rate at which a system produces finished output, and the question I want to answer here is why some firms increase it easily while others hit a ceiling they cannot see.
Table of Contents
Process Types and When Each Applies
Definition
Process types classify production systems by the volume of output and the degree of customisation. Nigel Slack, Alistair Brandon-Jones, and Robert Johnston, in Operations Management (now in its ninth edition), and Robert Hayes and Steven Wheelwright in Restoring Our Competitive Edge (1984), developed the taxonomy that operations scholars still use: project, job shop, batch, line, and continuous flow. Each type is suited to a distinct combination of volume and variety, and matching process type to product strategy is a foundational decision in operations management.
The five process types are:
- Project — one-off, highly customised output (construction, consulting engagements)
- Job shop — low volume, high variety, made to order (custom furniture, bespoke software)
- Batch — moderate volume, moderate variety, produced in lots (specialty chemicals, bakeries)
- Line — high volume, low variety, sequential operations (automotive assembly, electronics)
- Continuous — very high volume, very low variety, non-stop flow (oil refining, utilities)
I want you to notice that the choice of process type is a strategic decision, not a technical one. High-volume, low-variety products require line or continuous processes to be economically viable. Low-volume, high-variety products require job shop or batch processes because line processes cannot accommodate variation without excessive changeover costs. Firms that mismatch process type to product strategy face a structural cost disadvantage that cannot be remedied by operational improvements alone. The classic example is a firm that tries to manufacture customised products on a line process. Either the line breaks down under variation, or the customisation is lost.
The volume-variety trade-off produces four consequences:
- Volume-variety trade-off — increasing volume requires reducing variety, and vice versa
- Capital intensity — line and continuous processes are capital-intensive; job shop and project processes are labour-intensive
- Flexibility — lower-volume processes are more flexible; higher-volume processes are more efficient
- Unit cost — cost per unit falls as volume increases, but only if variety is reduced
Process type constrains what is possible operationally. A firm can improve within its process type, but changing process type requires a strategic decision that affects capital investment, workforce skills, and product strategy simultaneously. Firms that treat the process type as fixed and try to improve their way out of a mismatch waste years. Firms that recognise the constraint and redesign around it move faster.
Capacity, Throughput, and Bottlenecks
Definition
Throughput is the rate at which a system produces finished output. Capacity is the maximum throughput a system can achieve under normal operating conditions. A bottleneck is the stage in a process that limits the total throughput of the system. Eliyahu Goldratt, in The Goal (1984) and Theory of Constraints (1990), developed the constraint-based logic that operations scholars still use. Goldratt held that no system can produce more than its bottleneck stage allows, and that improving non-bottleneck stages produces no increase in total output.
Throughput management rests on four concepts:
- Throughput — units of output produced per unit of time
- Capacity — maximum possible throughput under defined operating conditions
- Bottleneck — the process stage with the lowest capacity; it sets the ceiling for the whole system
- Capacity utilisation — the ratio of actual throughput to maximum capacity
Here is the mechanism in one sentence: every system has at least one bottleneck, and the bottleneck determines the throughput ceiling for the entire system. If the bottleneck runs at full capacity, the system reaches its ceiling. If the bottleneck loses time to downtime, changeover, or defects, the whole system loses throughput. The managerial implication is counter-intuitive. Firms should manage the bottleneck with the utmost care, protect it from disruption, and avoid over-investing in non-bottleneck stages. Once a bottleneck is relieved, a new bottleneck emerges elsewhere, and the process repeats. Capacity expansion decisions should always target the current bottleneck, not the stage that is easiest to improve.
Goldratt’s five focusing steps are the standard sequence:
- Identify — find the stage with the highest utilisation or longest queue
- Exploit — extract maximum value from the existing bottleneck without major investment
- Subordinate — align non-bottleneck stages to support the bottleneck rate
- Elevate — invest to increase bottleneck capacity when exploitation is exhausted
- Repeat — once the bottleneck is elevated, identify the new bottleneck and continue
Throughput improvement is a systems problem, not a local problem. Improving the wrong stage produces no throughput gain. Improving the bottleneck produces immediate improvement. That logic is what makes the theory of constraints operationally valuable, and it is the logic that most firms fail to apply because their performance metrics reward local efficiency over system throughput.
Lean Flow and Waste Elimination
Definition
Lean production is an operating philosophy that focuses on eliminating waste — any activity that consumes resources without creating customer value — while maintaining flow. Lean originated in the Toyota Production System and was formalised for Western audiences by James Womack, Daniel Jones, and Daniel Roos in The Machine That Changed the World (1990). The seven classic wastes were later expanded by some practitioners to eight. Masaaki Imai’s work on kaizen provided the continuous improvement engine that sustains lean over time.
The eight forms of waste are:
- Overproduction — producing more than is needed, more than is required, or before it is needed
- Waiting — idle time caused by unbalanced process stages or delayed inputs
- Transport — unnecessary movement of materials or products between stages
- Over-processing — doing more to the product than the customer requires
- Inventory — excess stock at any stage of production
- Motion — unnecessary movement of people or equipment within a stage
- Defects — work that has to be re-done, scrapped, or corrected
- Unused talent — underutilising the skills of the workforce
I want you to see why lean is easier to state than to sustain. The mechanism is simple: identify value from the customer’s perspective, map the value stream, make the value-creating steps flow, let the customer pull value, and pursue perfection continuously. Lean produces improvement by eliminating waste, which reduces lead time and cost simultaneously. Just-in-time delivery, kanban scheduling, andon problem signalling, and standard work are the tools that support these principles. Lean is not a one-time project. It is a management system that requires continuous attention from leadership and continuous involvement from front-line workers.
The five lean principles are:
- Value — define what the customer actually values and is willing to pay for
- Value stream — visualise every step, delay, and information flow in the process
- Flow — arrange value-creating steps so product moves without interruption
- Pull — produce only what is needed, when it is needed, by the next downstream stage
- Perfection — pursue continuous improvement as a permanent discipline
Lean is a systemic commitment, not a toolkit. Firms that adopt lean tools without changing management philosophy typically see initial gains that erode over time. Firms that adopt lean as a management system sustain improvement for decades. That distinction is the one that matters most in practice, and it is the one that most firms underestimate.
Throughput Metrics and Operating Discipline
Definition
Throughput metrics are the operating measurements that reveal how well a production system is performing. They include cycle time, lead time, throughput rate, capacity utilisation, first-pass yield, and overall equipment effectiveness. Goldratt’s operational triad of throughput, inventory, and operating expense provides the accounting foundation for constraint-based measurement. Slack, Brandon-Jones, and Johnston extend the metric set to include customer-facing measures such as delivery reliability and quality conformance.
The core metrics are:
- Cycle time — the time taken to complete one unit of output at a given stage
- Lead time — total time from order to delivery across the whole process
- Throughput rate — units produced per unit of time
- Capacity utilisation — actual output as a percentage of maximum capacity
- First-pass yield — percentage of output that passes quality checks without rework
- Overall equipment effectiveness — availability multiplied by performance multiplied by quality
I will not pretend that metric selection is straightforward, because it is not. Throughput metrics must be aligned with system objectives, not local efficiency. A plant that maximises machine utilisation may produce excess inventory that increases holding cost and hides quality problems. A plant that measures throughput rate without measuring first-pass yield may report high output while generating rework that consumes capacity downstream. Best practice is to measure the system holistically, including throughput, lead time, and quality, rather than optimising any one metric in isolation. The metrics should also reflect the bottleneck logic. Utilisation of the bottleneck matters more than utilisation of non-bottleneck stages.
The five system-level metrics that should anchor the dashboard are:
- System throughput — finished goods produced per period, measured at the system level
- Inventory — work in process and finished goods held at any point
- Operating expense — the cost of converting inventory into throughput
- Quality yield — the proportion of output that meets specification without rework
- Responsiveness — the ability of the system to change output in response to demand shifts
Metrics are a management choice. Different metrics produce different behaviours. A firm that measures the right combination of throughput, quality, and responsiveness will make better operating decisions than a firm that measures only local efficiency. And the metric set itself should change as the operating environment changes, because a metric that drove the right behaviour in a stable market can drive the wrong behaviour in a volatile one.
Case Study
The two clearest public illustrations of process and throughput design are Toyota, whose production system is the benchmark for line-based throughput, and Safaricom, which operates a mixed process architecture that combines job shop, batch, and continuous flow elements in a single service business. Reading them together shows what the process-type framework and the throughput metrics look like across very different operating environments.
Toyota operates a hybrid process architecture. Core models are assembled on line processes with high volume and standardised platforms, while customisation is layered in through modular design and option packages. The system runs on line balancing and heijunka, which deliberately smooths production to reduce the variability that would otherwise disrupt the bottleneck. Kaizen operates as a daily practice, and employee suggestion systems generate hundreds of thousands of improvement ideas annually. Toyota produced more than ten million vehicles in 2025, and its throughput performance reflects decades of refinement on the same process architecture. Safaricom, by contrast, operates several process types simultaneously. Network infrastructure maintenance resembles a job shop. M-PESA transactions run on continuous flow. Customer service operates as a batch process. The firm uses statistical process control principles in network operations and transaction processing. Transaction success rates, network uptime, and fault restoration times are tracked against specification limits. Its Brand Strength Index above 90 in 2025 reflects the operational reliability that this mixed process architecture supports.
I want you to see what the two cases establish together. In both, process type was chosen to match the volume and variety profile of the service or product. Toyota standardised where variety could be absorbed through modular design, and Safaricom segmented different functions into different process types rather than forcing them all through one. In both, throughput was managed as a system property rather than a local one. Toyota protects the line’s bottleneck by smoothing demand. Safaricom segments its operations so that a fault in one service does not cascade through the others. In both, the metrics were designed to surface problems, not hide them. Toyota treats the surfacing of problems as a healthy signal. Safaricom’s brand strength depends on the reliability that transparent metrics support. The cases also show what the frameworks do not provide. Neither firm achieved its throughput performance by adopting a template. Toyota built the Toyota Production System over decades. Safaricom built its mixed process architecture around the specific constraints of a Kenyan telecoms market. The frameworks describe the principles. The design discipline is what makes them work.
Conclusion
The Toyota and Safaricom cases together establish a proposition that is more useful than either case alone: throughput is a design property, not an operational achievement. Firms that treat throughput as something to be optimised within an existing process architecture hit a ceiling. Firms that treat throughput as a design property and redesign the process architecture around the volume and variety profile of their product or service find throughput gains that were not available through operational improvement alone.
My detailed conclusion is this. The process-type framework from Slack and the volume-variety trade-off from Hayes and Wheelwright are the correct starting point, but the standard operations curriculum undersells three things. First, it undersells how rarely firms operate a single process type. Safaricom operates job shop, batch, and continuous flow simultaneously, and most large service firms do the same. The framework is a design vocabulary, not a classification system, and firms that treat it as a classification miss the design opportunity. Second, it undersells how much throughput is determined by bottleneck management. Goldratt’s constraint logic has been taught for forty years, and yet most firms still optimise non-bottleneck stages because those are the stages whose metrics they can see. The bottleneck is where the leverage is, and the leverage is invisible without the right metric set. Third, it undersells the trade-off between throughput and responsiveness. A firm that maximises throughput in a stable market will lose responsiveness when the market shifts, and a firm that maximises responsiveness will lose throughput in normal times. The design discipline is to hold both in balance, and the balance point moves as the market moves. What the frameworks ask of you, as a manager or a student of management, is the discipline to treat throughput as a system property, to design around the bottleneck rather than around the metrics, and to keep the design tuned to the volume and variety profile of the market you actually serve rather than the one you wish you served.
My recommendations follow from that conclusion. First, match process type to product strategy explicitly, and recognise that changing process type is a strategic decision that affects capital, workforce, and product simultaneously. Second, identify the bottleneck and manage it as the system’s throughput ceiling; protect it from disruption and do not over-invest in non-bottleneck stages. Third, treat lean as a management system rather than a tool kit; the firms that adopt the culture sustain improvement, and the firms that adopt the tools alone do not. Fourth, design the metric set to surface problems rather than hide them, and update it as the operating environment changes. Fifth, when you operate multiple process types in one organisation, segment the metrics and the operating disciplines accordingly rather than forcing every function through the same framework.
The final argument of this post is that throughput is not what you optimise. It is what you design. The process types describe the options, and the design choices determine whether you reach the throughput your market allows.
Read Also on Kat-Syd Resources Hub
Management Principles Series: Planning, Organizing, Staffing, Directing & Controlling — Foundational management functions that govern how firms plan, organise, and control their operations.
Functional Management — The Pillars — How core business functions including operations and production are organised and measured.
Revenue Growth Playbook Series — Analytical series on the strategies and operating levers that drive sustainable revenue expansion.
Written by Kateule Sydney — Researcher and Writer
© 2026 Kat-Syd Resources Hub. All rights reserved.
Kat-Syd Resources Hub — Educational case studies and analytical reference