Implementation of lean production allows many companies to shift from producing goods to a forecast to production based on actual customer requirements; in the mature lean environment, "production equals sales". Reductions in work-in-process inventories and lead times permit the application of simplified visible inventory signals to schedule your manufacturing resources.

On a tactical level, pull systems seek to provide the information necessary to control and to accelerate the flow of materials throughout the supply chain. Changing the scheduling of the supply chain from a forecast-based, infinite reverse push system to a consumption-based pull system will result in significant and permanent reductions in inventories with simultaneous increases in customer service.

What are the prerequisites to pull systems? Certainly not the implementation of techniques such as kaizen, setup reduction or cellular manufacturing, or the advance buildup of supermarkets. In fact, pull systems should be the first step toward filling the supermarkets and providing the springboard for focusing these other techniques. The only real prerequisite is an internal champion with the leadership and the will to drive the implementation.

So, the basic objectives of pull systems are to:

Synchronous Management has developed a step-by-step approach to repetitive pull system implementation, based on years of experience in materials management, and the successful design and implementation of hundreds of pull systems. If any of the steps are omitted, less than optimal results will be guaranteed. If the steps are followed in sequence, significant and permanent financial benefits will result. The steps are:

The intent of this paper is to provide you with a general outline of our approach, with sufficient explanation of each step that you may be able to relate it to your company. This is not a "cookbook", but a set of guidelines to help you to assess the scope of implementation and the potential for the application of a successful system in your company.


As with any significant program, pull systems implementation must be supported with a time-phased project plan. Clear definition will be required of specific activities, their timing, and their ownership. Although this paper provides you with an outline of our approach, the actual implementation plan will be in significantly greater detail.

Because the purpose of the pull system is to link production process across functional lines, the makeup of the implementation team should reflect the functions affected. Typical functions represented would include:

Production control/scheduling/dispatching

Line operations/key floor leaders

Management information systems

Materials logistics/warehousing/material handling

Marketing/sales/distribution/customer service

Cost accounting/finance

Purchasing/receiving/outside processing

Process/design engineering

Of course, one individual may represent several functions, and the actual makeup of the team will vary depending on your organizational structure.


Often, companies will send a delegation to a "kanban" seminar and expect them to return with the ability to implement a pull system. Our experience is that this approach almost never works - at least not within several years. Many problems are encountered during implementation which cannot possibly be covered in a generic seminar, and the solutions must be tailored to your environment.

For example, on the strategic level, you will be faced with changes in performance measures, defining customer service requirements, and contention for your resources by other productivity improvement efforts. Tactically, you will need to decide on locations and sizes of supermarkets, calculate setup-driven batch sizes, and interface the pull system with other information systems.

Your management and front-line implementation teams should receive training in both the theory of constraints (to help with the strategic issues) and how to implement pull systems (for the tactical issues). In addition, the use of an experienced outside facilitator will help keep your teams on track and show you alternative means by which other companies have successfully dealt with issues similar to yours.


Repetitive pull systems are not applicable across the board. The basic question is: On which items must we carry inventory in order to provide competitive customer service levels with our current supply constraints? A competitive customer service level is one which enables you to capture additional market share or to stave off loss of market share. Current supply constraints include such factors as your current lead times, batching requirements, vendor performance, product structures, information systems, and labor environment. Answering these questions will help you to identify repetitive items:

  1. Is the market demanded lead time for an item shorter than its lead time through the supply chain?
  2. Which 20% of your end items consume 80% of your capacity?
  3. Where is supply chain capacity not flexible enough to adjust to variability in demand?
  4. Do you have common component parts whose parents are non-repetitive, but whose total demands are repetitive?

For the application of pull system techniques to the non-repetitive environment, see MAKE-TO-ORDER KANBAN.


The pull system is driven and production is scheduled based on actual customer demand. In general, two methods are available to provide the top level pull mechanism to drive the entire supply chain - direct consumption pull or mixed model schedules.

A direct consumption pull approach replenishes the supermarket closest to the customer in the actual quantities consumed. This causes the variability in customer demand to propagate through the supply chain. If supply chain capacity is not flexible enough to handle the peaks, replenishment lead times would extend to account for the de facto level-loading which will occur. In addition, because the consumption of components is unpredictable, safety stocks may have to be carried at the component levels; the result will be mis-timing of lower level requirements.

A better approach where demand is highly variable is to schedule to a finished goods supermarket with a mixed-model or rate schedule such as takt time. In this case, the production requirement for each item over a predetermined time period is divided into the smallest possible increments, and the final assembly or configuration process is scheduled at this mix in each time period. The advantages of this approach are less excess capacity required in the supply chain, more predictable replenishment lead times and, most importantly, consumption of each lower level item at a steady rate. Any safety stock would have to be carried at the end item level only. Of course, providing your customers with small-batch, frequent deliveries would automatically drive the pull system with a mixed-model.


The supermarkets should be strategically located - that is, at key points in the product/process flow to protect the availability of material and the smooth flow of product. These are the specific operations whose schedules will be set directly by the pull system. The following are the most common criteria we use to decide on supermarket locations:

Full implementation will also deal with such issues as point-of-use storage, multiple-use items, and secondary priority controls which are beyond the scope of this summary.


Unless a model of the system is developed which projects the results, two common problems will result. First, the supermarkets will be sized inappropriately for your constraints. Replenishment will be triggered either too soon or too late. Second, as market and supply conditions change, the system will be so difficult to maintain that it will deteriorate. Modeling the supermarkets with a spreadsheet enables the user to project changes in inventory and customer service and to identify key constraints before implementation.

To size each supermarket, two quantities are calculated. The first is the trigger quantity (a.k.a. reorder point). This is the supermarket level at which a replenishment order is released - the demand during the replenishment lead time. Thus, if a part has demand of 10 pieces per day with a 3 day replenishment lead time from the prior buffer, the trigger point would be 30 pieces. That is, when the inventory in the buffer drops to 30 pieces, a replenishment order must be started.

The second quantity is the amount of the replenishment order. If no setup is involved in the supply of the item, the replenishment quantity can be one piece. However, if setups are required which would drive a batch greater than one, EVERY PART EVERY INTERVAL should be applied to determine the replenishment lot size. In our example, if the supplier is able to produce the part in 2-day batches, the replenishment quantity would be 20 pieces. So, when the buffer inventory drops to 30 pieces, a replenishment order is released for 20 pieces. The total inventory of work in process and finished goods is therefore limited to 50 pieces. This limiting of inventory levels is essential in any pull system.

In addition, the trigger and the lot size quantities may have to be adjusted for factors such as safety stock, quality yields, packaging multiples, tooling constraints, parent lot sizes, or raw material yields.


Pull systems are, of course, information systems. They answer the scheduling questions: What do we make next, when, and how many? So, they must be interfaced with other information systems, particularly MRP, shop floor reporting, inventory management, and performance measurement.

In most companies, some form of work order or shop order authorizes production. Under the pull system, the pull signal authorizes production. So, the work order becomes merely the means by which the pull system interfaces with shop floor control and cost accounting. Work orders for items on pull should not be firmed and printed in the materials requirements planning system until a pull signal is received authorizing release of the material.

In some manufacturing systems, work orders are not required - production is accounted for at key steps in the process. Materials (and sometimes labor) are automatically issued as parts pass these key steps. We encourage this automated "backflushing" because it simplifies the interface of pull systems with other information systems.


Unfortunately, this is where many companies start. For example, they decide to implement a kanban system. However, kanban is only one form of pull system. Where the supplier is within visual range of the buffer and the batch size is small, an empty shelf at the point of use may be all that is required for a signal. Or, if the supplier is close, but not in visual range, empty containers may suffice. With even greater distance, kanban cards may be necessary. If the batch quantity is larger than one card's worth of parts, trigger boards may be required to accumulate a batch of kanbans. If the material flows are complex, a work in process tracking mechanism may be needed. Finally, if large numbers of parts are implemented, if the supplier is in another facility, or if frequent resizing of buffers is required because of demand volatility or seasonality, electronic signals may be appropriate.

Before the system can be started up, the actual pull mechanisms must be made, such as kanban cards, trigger boards, or dedicated containers. The actual number of signals required for each buffer is calculated during the buffer sizing step. Usually, it is the trigger quantity plus the replenishment batch size divided by the number of pieces per shelf or container. Make only enough signals to cover the inventory authorized by the pull system. Pay particular attention to the kinds of information the pull signal must communicate. Try to keep it simple and flexible.


When the pull system is functioning correctly, the accelerated material flow will be reflected in improved customer service with decreased inventories. The key performance measures we recommend are variations of those proposed by Dr. Eli Goldratt in THE GOAL - increases in output (both the volume and timing), decreases in inventories (and therefore lead times), and reductions in total costs. We also recommend that the performance measures be maintained over time (trends) and be clearly and visibly posted and maintained so that they can be seen by the associates who can actually improve performance. For additional information, see PERFORMANCE MEASUREMENT.

Examples of measures of improved material flow may include, among others: output per person; days of inventory; throughput times; stockouts or fill rates; output per square foot; hands-on vs. total lead time; output linearity. Of course, the specific measurements, and their units of measure, will vary depending on your production processes and demand constraints.


Once the system has been modeled, performance measurements have been agreed upon, and the physical components have been fabricated, you are ready for startup. Generally, the following sequence will apply:


The pull system is designed and implemented based on assumptions as to the nature of demand and the constraints in supply. Reality may be different. So, it is important that the functioning of the system be monitored to ensure that the system is being followed and that it is actually producing benefits. Key questions include: Is the inventory (work in process and finished goods) for each buffer approaching the levels projected with the pull system model? Are the supermarkets in the right locations? Are the buffers too large or too small? Are additional or fewer buffers required? Is the best pull signal mechanism being used?

Responsibilities should be assigned, and procedures documented for operating the system on a day-to-day basis, for monitoring performance of the system, taking corrective action, and for updating the buffers as demand or supply conditions change.


Hopefully, the above has helped to clarify that a quantitative approach does exist for implementing pull systems in any repetitive environment. The supermarkets required to protect and improve material flow and to schedule your operation can be rationalized to provide the maximum output and customer service with the minimum inventories and operating costs. Most importantly, because a successful pull system requires transition to a throughput-oriented (rather than cost-oriented) strategy, it becomes the springboard to focused application and success for your other productivity improvement efforts.