German Portal



 About RMP

About themanager



The Pull System Mystery Explained:  

Drum, Buffer & Rope With A Computer

By Gerald Najarian

Going forward by moving backward is how one author described a pull system.  Others use the analogies of drums, buffers and rope to explain how to "pull" production through a manufacturing shop.  There are numerous images one can use to visualize pulling goods through a plant.  Since pull systems are frequently contrasted to so-called "push" systems (i.e., MRP II) in which production is master scheduled to push production from one operation to the next whether needed or not, they are often thought of as merely scheduling or shop floor control tools.  In fact, pull systems are much more.  They are the heart of a synchronized factory.  They permit synchronization by working backward through signals or triggers which cause production events to happen.  The experience of backward motion starts at the finished goods warehouse or shipping area and signals the previous operation - final assembly - when more finished goods are needed.  Final assembly in turn signals a previous operation - perhaps a component fabrication department - when more components are needed.  Component fabrication shops may signal a preceding manufacturing department or a raw materials storekeeping area which would signal a vendor to make a delivery.  The signals in a pull system are in fact inventory levels for fabrication and raw materials replenishment and may be either inventory levels or a forecast/schedule for finished goods replenishment.  With such a system in place, shipment of finished goods triggers withdrawal of components for assembly to replenish the shipped goods which triggers withdrawal of raw materials for fabrication to replenish the withdrawn components and so on through the triggering of a vendor shipment to replenish raw materials.  This pull sequence system is known as a Kanban system in Japan and in The Toyota Automobile Corporation where it was developed and refined.  More revealing of its simplicity is the fact that the Kanban system is often still called the supermarket system because the concept originated in observation of American supermarkets by Toyota executives.  The Toyota executives observed that when customers withdraw goods from the small stocks on supermarket shelves, the stocks are replenished in small quantities by a stock clerk who checks the shelves and replaces only the quantity which was taken.  The first pull signal came from the customer who withdrew the inventory and told the stock clerk how much to replenish.  The Toyota executives reasoned that this supermarket concept could be adapted for management of a factory on a simple visual basis.  Since it is impractical to have roving "stock clerks" in a factory, a card is used to communicate to the production foremen the fact that a shelf was empty.  The Japanese word for card is kanban; hence the name Kanban for a pull system.  


In a system which triggers production in backward motion, a system of signals is the means to communicate the replenishment of goods.  The signal media in a "classic" Kanban system are cards and containers.  While there are many variations on the visual Kanban theme, the most instructive is the Toyota system.  The Toyota system utilizes a specifically sized container for each part which cycles back and forth between the producing department and the using department (each may have specific store-keeping areas).  Two cards (kanban) are used: a production kanban and a conveyance kanban.  These kanban specify the part number, the container capacity and other data.  

When a using department withdraws a container of parts, the conveyance kanban previously attached to it by the producing department is detached and placed in a collection box.  When the most recently emptied container for the same parts is ready to be conveyed to its producing department, the conveyance kanban in the collection box is attached to it.  At the time this empty container is received by the production department, the conveyance kanban is detached and attached to a recently manufactured full container of those parts which is then moved to the using department.  The removal of the full container out of the producing department triggers production through removal of a production kanban attached to it which is placed in a collection box.  The production kanban in the collection box are transferred hourly to a dispatch box and serve as the authorization for the foremen to produce those parts within a specific time frame and fill an empty container.  When the container is filed the production kanban is attached to it and the container is placed in a store area awaiting transfer to the using department.  This process repeats itself over and over again.  

There are three simple rules which control this Kanban system:  

     Producing departments may not make parts unless there is a production kanban in the dispatch box authorizing production.  

     There is precisely one conveyance and one production kanban for each container. 

     The number of containers are controlled by manufacturing management and are kept to the smallest possible quantity in size.  (Toyota management must approve the use of a container holding more than a tenth of a day's supply.)  

Kanban systems are the conceptual model for pull systems in other environments and in fact numerous variations exist.  There are single card Kanban systems; some systems use metal plates instead of cards; one company uses numbered ping pong balls; General Motors sends Kanban signals via computer.  No matter what the variation on the signal, the principle is the same -- the using department tells the producing department what to do based upon demand at the beginning of the chain, a sale of the product.  Hence the name pull system.  


The basic concepts of a pull system are the ideas of small lot production in standard lot sizes (the container) signalled by inventory depletion (the production kanban).  For manufacturing companies in which the Toyota Kanban system is culturally cumbersome and with frequent demand fluctuations, there is an alternative.  That alternative is to simulate the container and card system with cycle times, lot sizes and the company's automated perpetual inventory system.  How exactly does this work?  

In the "classic" Kanban system the containers represent the lot size called for by the container size.  In an automated pull system, lot sizes and cycle times reside in computer files.  The automated "system" sends signals when the perpetual inventory, also resident in the computer files, diminishes to a point which would represent the removal of a container from the production department and the receipt of the Kanban card by the production department.  Let's discuss some of the key concepts and then put them together as a system.  

      Cycle times.  A cycle time for a pull system is the realistic amount of time it takes to manufacture or procure a specified amount of goods by a work center or from a supplier.  This is a replenishment cycle time as differentiated from a capacity oriented line speed.  As such, replenishment cycle times closely resemble the lead time that a vendor quotes to a purchasing manager.  Customarily, the vendor doesn't quote the capacity/line speed to the purchaser.  They quote a period of time which takes capacity into consideration but also recognizes projected demand and the mix of items expected in their shop.  It is similar for an internal cycle time -- the work center is treated as a vendor with a capacity, expected load, changeover rate and mix to manage.  These factors are taken into account and a cycle time expressed usually in days is set with which to set inventory levels and to trigger production to be completed within the cycle time.  

     Forecasts.  Sales forecasting is the critical element in all manufacturing management and control systems.  In an MRP II system it is used to set the master schedule and then push product through manufacturing.  In a pull system it is used to turn a cycle time into a targeted inventory level.  The difference is significant.  The unit forecast for a particular end item usually covers a specific period of time -- a month, a quarter, a year and so forth.  In MRP II, the gross unit forecast is used; in a pull system, the gross unit forecast is converted to a rate of sale expressed in the same denomination as the cycle time for final assembly of the end items.  For example, if final assembly cycle time is expressed in days then the forecast will be expressed as a daily rate, if in hours then the forecast will be expressed as a rate per hour.  

     Buffer stock.  In the drum, buffer, rope image of pulled through production, buffer is the stock level that creates balance in the system (simply stated: the drum is the constraint which paces the plant and the rope is the communication system which links the actions of the work centers into a synchronized flow).  At the finished goods level the buffer stock exists to permit product to be shipped in less time than the final assembly cycle time.  For component parts, buffer stocks are designed to permit production of finished goods in less time than the fabrication cycle time.  These buffer inventories are determined by multiplying cycle times by daily forecast rates to provide for the least amount of stock and to protect production and shipment reliability.  

      Lot size.  The lot size for production will usually be expressed as a combination of cycle days and the daily forecast rate.  However, it can never be set at an amount which is unrealistic to a constraint (bottleneck) in the routing.  Also, if desired the lot size can be set to pace production and thereby become a proxy for the drum in the drum, buffer, rope image.  

In constructing a pull system from these concepts, it will be helpful to use a simple one product, two component model.  Starting with the cycle times: we determine, with the foreman and the planner: (i) that the work center our product is made in requires a five day cycle in finished goods assembly, (ii) that the fabrication center requires two and three days to make each of the components for our product, and finally (iii) that the vendor supplying the raw materials has a ten day lead time.  We'll then assume that the forecast for our product converts to ten units per day and that components and raw materials are in a one-to-one relationship to the end item (with no scrap, miraculously!).  Now we need buffer stocks to protect shipability and lot sizes to communicate to the work centers.  With a final assembly cycle time of five days we need to have a finished goods buffer of something more than a five days supply, in this case more than fifty units.  We also need component buffers of something more than two and three days supply -- twenty and thirty units -- and raw material buffer of something more than ten days supply or one hundred pieces.  Now for lot sizes, assuming no bottleneck, we set the lot size at twice the cycle time for finished goods, components and raw materials.  This means that manufacturing or procurement will deliver a quantity of twice the cycle time extended at the daily forecast rate within the agreed upon cycle time.  In our model this means one hundred end items produced in five days; one hundred forty and one hundred sixty each for component parts within their two and three day cycles; and three hundred sixty pieces of raw material within its ten day cycle time.  The lot sizes can be less but should rarely ever be greater than twice the cycle time.  The general lot size rule for customer service protection is twice the cycle time for the operation plus the cycle times of all preceding operations.  


It should now be evident that cycle times and the forecast rate are the building blocks of a pull system and that lot sizes and buffer stock are arithmetic functions of them.  Buffer stock levels now become the trigger for a signal through the system to manufacture or procure goods.  So, as finished goods inventory diminishes to a point below its cycle -- in our model, a five days' supply -- a signal is sent to release a shop order to make a ten day lot size in five days.  At the rate of forecasted demand, inventory will decline to zero in five days at which time it will be replenished in an amount sufficient to service customers for five days until another signal is sent.  The same things happen for fabricated components and raw materials to signal production or procurement in their cycle times.  The signals through the perpetual inventory and the connected pull system files and logic and the shop orders generated are the "rope" in the drum, buffer, rope model of manufacturing.  

A pull system which simulates Kanban with cycle times and forecast rates effectively combines the simplicity of inventory and production management inherent in Kanban with the synchronized vision of drum, buffer, rope.  With the exception of bottleneck corrections the system is self adjusting through the management of cycle times and a dynamic forecasting process.  


The chief implication of a pull system is manufacturing discipline and discipline is also the main benefit.  More specifically a pull system needs:  

     Labor force flexibility to perform small lot production. 

     Quality material handling skills to keep parts moving in synch from department to inventory to department, etc.  

     Continuous forecasting to keep the system constantly self adjusting to market demand with good data.  

     Adherence to shop order due dates to assure that customer service levels and buffer stocks are maintained.  

Among the many noticeable benefits of a pull system combining the best features of Kanban and Drum-Buffer-Rope is a balanced inventory which permits the company to meet customer lead time commitments.  Perhaps as important is the continuous improvement ethic set up by such a system.  The focus of continuous improvement will be on reduction of cycle times which in turn enhances flow and permits reduction of buffer stock inventories.  There are numerous others -- greater WIP turns, less overtime higher throughput. 

Will it work in your plant?  The answer is a resounding yes if you are in a job shop, repetitive manufacturing process or a discrete fabrication assembly shop.  The configuration is slightly different but the pull system works equally well and produces the same benefits. 


This article was provided by Gerald Najarian, The Remington Group, LLC

Copyright   The Remington Group, LLC


Some links for your own research:





Send to a friend
  up back to publications - Management & Strategy back to  

If you have questions or comments to our website, do not hesitate to contact us (comments and questions are always welcomed): webmaster2 AT 
Copyright 2001 Recklies Management Project GmbH
Status: 18. Januar 2008