Table of contents:
In the context of internal logistics in manufacturing, a pull system is an approach where production and material movements are driven strictly by actual demand rather than forecasts. In such a system, work-in-progress (WIP) and inventory replenishment occur only in response to consumption further along the production line. This stands in contrast to a push system, which relies on schedules and predictive planning to initiate production and material flows.
Baudin and Netland's Introduction to Manufacturing frames internal logistics as part of "making materials flow," and emphasises the importance of aligning production with real demand signals through well-designed information and control flow management. Although the book does not provide a verbatim definition of pull vs. push systems, it presents a coherent "systems perspective" in which materials must move only when needed—aligning with the core principle of pull.
Characteristically, a pull system is demand-driven. Components, subassemblies, and raw materials are only replenished when an actual downstream operation consumes them. A common mechanism for this is the Kanban system: visual signals such as cards, bins, or empty containers flag the need to replenish parts, thereby pulling inventory through the process only when necessary (2). Some systems, like CONWIP (Constant Work-In-Process), further define pull control by using a limited number of system-wide authorisations (cards) that allow a new order to enter the system only when a finished part leaves.
Reference (1)
By contrast, a push system schedules production based on forecasts and planning. Manufacturing and logistics upstream of the demand point push materials forward regardless of whether downstream processes are ready or customers have ordered. This can lead to overproduction, excess inventory, higher storage costs, and inefficiencies (3).
The key differences between pull and push systems can be distilled as follows:
Reflecting Baudin and Netland's system-oriented approach, a pull system can be seen as aligning manufacturing operations with the actual "central nervous system" of information flow and event responses, ensuring that material handling and logistics respond dynamically to real-time needs rather than predetermined plans.
In summary, a pull system in internal logistics is a demand-driven approach where production and material flows are triggered by actual usage and demand, often operationalised via Kanban, WIP limits, or CONWIP. It contrasts with push systems, which rely on forecasting and scheduling to drive production, often resulting in overproduction and inventory buildup. The pull system, as integrated into the systems perspective offered by Baudin and Netland, underscores lean principles centred on flow efficiency, responsiveness, and waste reduction.
A pull system is often considered counterintuitive because it challenges deeply rooted managerial instincts about control, planning, and productivity. In traditional manufacturing logic, which underpins a push system, managers rely heavily on forecasts and detailed schedules to ensure that production resources are fully utilised and that goods are always available ahead of demand. The belief is that by producing in advance and pushing items downstream, one can minimise idle time, keep machines running, and maximise economies of scale. From this perspective, it seems risky and even inefficient to wait for actual consumption before replenishing materials or starting new production. The counterintuitive element of a pull system lies in its insistence that less control in terms of forward planning—combined with reliance on simple signals from the shop floor—can actually yield better flow, higher efficiency, and shorter lead times.
Baudin and Netland, in Introduction to Manufacturing: An Industrial Engineering and Management Perspective, stress that manufacturing systems must be designed as coherent wholes in which flow is prioritised over local optimisation. Push systems, by trying to maximise utilisation at each workstation based on forecasted demand, tend to create bottlenecks, excess work-in-process, and long lead times. These inefficiencies are often hidden because the system appears busy and productive, while in reality, most of the effort is tied up in non-value-adding inventory. The push approach generates waste in the form of overproduction, excess handling, and storage requirements, all of which contribute to higher costs and less responsiveness.
Pull systems, by contrast, restrict production to what is immediately needed by the next process step. This reduces work-in-progress dramatically and exposes inefficiencies that are otherwise masked by excess inventory. Time savings occur because materials flow smoothly with fewer interruptions, and throughput time is reduced since items are not queuing unnecessarily between operations. From an efficiency perspective, pull systems minimise the mismatch between supply and demand, aligning resources with actual requirements instead of predictions that are invariably uncertain. They also reduce the likelihood of producing the wrong items in the wrong quantities, a common outcome in push environments where forecast errors cascade into excess stock or shortages.
The superiority of pull systems, as Baudin and Netland argue, lies in their systemic efficiency rather than in local utilisation. By lowering inventory, shortening lead times, and aligning production with consumption, pull systems free capacity, improve delivery reliability, and make the entire manufacturing operation more resilient. The counterintuitive insight is that doing less in advance and producing only when needed results in more effective performance overall.
Kanban plays a central role in the functioning of a pull system, acting as the mechanism that translates actual consumption into replenishment signals. At its core, a pull system requires a way to trigger production or material movement only when something has been consumed downstream. Kanban provides this trigger in a simple, visual, and standardised form. Traditionally implemented with cards, bins, or electronic signals, Kanban communicates the need to produce or move a specific quantity of material, thereby ensuring that resources are mobilised only in response to real demand.
The strength of Kanban lies in its ability to make material flows transparent and self-regulating. Instead of relying on forecasts or complex scheduling systems, each Kanban signal authorises a defined quantity of work. For example, when a downstream process uses up a container of parts, the empty container or associated card is sent upstream as a signal to replenish exactly that quantity. This not only synchronises production stages but also enforces limits on work-in-process inventory, since the number of Kanbans in circulation effectively caps the amount of material in the system.
In Baudin and Netland's view of manufacturing as a system of interdependent flows, Kanban embodies the principle of decentralised control that underpins pull. Restricting production to demand-driven authorisations prevents the accumulation of excess stock, reduces lead times, and helps uncover inefficiencies. Thus, Kanban is not merely a scheduling tool but the operational backbone of a pull system, ensuring alignment between material flow and consumption.
In a pull system, the effectiveness of material flow depends on how pull signals are issued and controlled. A pull signal is essentially an authorisation for upstream processes to produce or move a specific quantity of items. These signals are triggered only when a downstream process consumes material, thereby ensuring that replenishment is tied to actual demand. The Kanban method is the most widely recognised way of structuring these signals, as it provides a tangible and standardised communication system that is simple to operate on the shop floor.
Pull signals can take various forms, from physical cards and empty bins to electronic messages in advanced manufacturing execution systems. The principle, however, is always the same: when an operator or machine consumes materials, a signal is sent upstream to replenish that exact amount. Control is achieved by limiting the number of active signals circulating in the system. This limitation prevents overproduction and stabilises work-in-progress inventory. Kanban cards or containers act as tokens of authorisation—no production is allowed without one, which enforces discipline and synchronisation across operations.
For a Kanban system to function effectively, several rules are necessary. First, production may only occur in response to a Kanban signal; forecasts or assumptions cannot be used to justify extra output. Second, each Kanban must correspond to a fixed, standard container size to simplify material handling and ensure consistency. Third, defective products should never be passed along, as Kanban aims to synchronise both quality and quantity. Fourth, every Kanban must be attached, returned, or logged correctly; losing a card means breaking the control loop. Finally, continuous monitoring of demand and replenishment cycles is necessary because a Kanban system must adapt to changes in takt time, product mix, or process reliability.
Calculating the number of Kanbans in a loop is a critical design step, as this number determines the amount of inventory that circulates between two processes. The standard formula often cited in lean manufacturing literature is:
Number of Kanbans = (Demand rate × Lead time × (1 + Safety factor)) ÷ Container size
Here, the demand rate is the average consumption of parts per unit of time, and the lead time is the replenishment time needed from signal issuance until parts are available again. The safety factor accounts for variability in demand or supply, providing a buffer against shortages. Container size fixes the standard quantity that each Kanban represents. For example, if a process consumes 100 units per hour, replenishment lead time is two hours, safety factor is 10 per cent, and each container holds 50 units, then the calculation yields:
(100 × 2 × 1.1) ÷ 50 = 4.4, rounded up to 5 Kanbans.
This means that at most five containers, each authorised by a Kanban, will circulate in the loop, capping the work-in-progress inventory while ensuring supply continuity.
Baudin and Netland emphasise that such calculations are only the starting point. Continuous observation and refinement are necessary to adjust Kanban numbers as conditions change, ensuring the pull system remains lean, stable, and responsive.
A pull system aligns production directly with actual demand rather than forecasts. This reduces overproduction, excess inventory, and waiting times. The result is shorter lead times, lower storage costs, and greater responsiveness to customer needs. Push systems, in contrast, often generate waste because they rely on predictions that may not reflect real consumption.
Kanban provides the signaling mechanism that makes a pull system work. Each Kanban card or container represents authorisation to produce or move a fixed quantity of material. When a downstream process consumes parts, the Kanban signal is sent upstream to trigger replenishment. This enforces discipline, limits work-in-progress, and ensures that materials flow smoothly in response to actual demand.
The number of Kanbans is calculated using demand rate, replenishment lead time, container size, and a safety factor for variability. The formula is:
Number of Kanbans = (Demand rate × Lead time × (1 + Safety factor)) ÷ Container size.
This calculation caps inventory levels while ensuring a stable supply. Adjustments are made over time as processes, product mix, or demand patterns change.
Real-time locating systems (RTLS) enhance the effectiveness of pull systems and Kanban by automating the flow of demand signals. Instead of relying solely on manual cards or bins, RTLS solutions provide precise, real-time visibility of material consumption, inventory levels, and container movement across the shop floor. This transparency reduces errors, accelerates replenishment, and ensures that Kanban loops operate with minimal friction. By integrating RTLS into intralogistics, manufacturers strengthen demand-driven control, streamline workflows, and uncover hidden inefficiencies, creating a more resilient and adaptive pull system that responds instantly to actual production needs.
Delve deeper into one of our core topics: Real time locating system
CONWIP (Constant Work-In-Process) is a pull-based production control method that limits the total amount of work circulating in a system. Unlike Kanban, which ties replenishment signals to specific parts or stations, CONWIP uses a fixed number of system-wide authorizations (cards). A new job can only enter the line when a finished product exits, keeping WIP stable while allowing product mix flexibility. (5)
References:
(1) M. Baudin & T. Netland (2023). Introduction to Manufacturing. An Industrial Engineering and Management Perspective. Routledge.
(2) https://resources.duralabel.com/articles/pull-system
(3) https://www.interlakemecalux.com/blog/push-pull-system
(4) https://en.wikipedia.org/wiki/Push%E2%80%93pull_strategy
(5) Spearman, M. L., Woodruff, D. L., & Hopp, W. J. (1990). CONWIP: A pull alternative to Kanban. International Journal of Production Research, 28(5), 879–894.
Note: This article was partly created with the assistance of artificial intelligence to support drafting. The head image was generated by AI.