Pull vs Push: a Discussion of Lean, JIT, Flow, and Traditional MRP Part Two: Challenges and User Recommendations


Still, while some tout that flow manufacturing principles can be implemented successfully, regardless of the industry, type of manufacturing environment, or product volumes, the concept has not been everything to all people so far. It is still challenging or even unsuitable to use in a jobbing shop producing highly configure-to-order (CTO) or engineer-to-order (ETO) products with high setup and long lead times, although when it has been occasionally deployed there, it has met almost as much success as in high-volume, more repetitive make-to-demand environments.

The fact is also only a minority of all ERP vendors properly supports the ETO environments, let alone flow manufacturing concepts. Cincom is one notable exception, given its Flow Manager product handles kanban replenishment and demand smoothing (but not line design and operation method sheets (OMS), because these features do not bring much benefit to ETO manufacturers). Their customers often specify product families that include products requiring one or two unique and expensive components in addition to common parts that could benefit from flow methods of smoothing spikes in demand.

Also, only certain industries like high-tech, aviation and defense (A&D), and consumer packaged goods (CPG) have been the best candidates to adopt flow manufacturing, because they can most feasibly achieve feedback from reconfiguring the shop floor into dedicated production lines for product families, around which flexible and cross-trained work teams are established. However, for other organizations, that feat will not happen overnight, since achieving flow manufacturing takes more than moving equipment into product family production lines, creating flexible workstation teams, establishing quick changeovers, or introducing kanban signals. It also requires specific flow manufacturing training, continuous discipline and process improvement mindsets (e.g., zero defects, zero setup, the use of standardized components, and zero inventory).

Until relatively recently, there have not been many off-the-shelf software applications to help manufacturers institute flow processes, other than cumbersome spreadsheets and internally developed packages or practices to handle some flow techniques, such as line design, line balancing, kanban management, and mixed-model production. In addition to the lack of consensus regarding what exact set of features constitutes the flow manufacturing software, there has also been a philosophical debate about what flow manufacturing software should do in relation to ERP/MRP.

On one hand, John Costanza Institute of Technology (JCIT) has been advocating the misfit of ERP's concepts of planning and scheduling, which do not allow a manufacturer to use a demand forecast as the basis for a reliable materials purchasing plan. Conversely, flow-oriented manufacturing starts with the earlier-mentioned process called demand smoothing, which technique involves looking at a forecast over a certain period of time (anywhere from one week to several weeks) to determine how many products must be built each day to fill the total amount of orders expected over that time. The flow manufacturer will then ask its suppliers to deliver parts every day, with each shipment amounting to only enough parts to satisfy that day's quota.

On the other hand, there are some indications that flow systems cannot handle demand variability, variable product mix, shared resource constraints, or complex products with long lead times, thus limiting flow for items where variability is only at the end item mix, and not with frequent content variations of option mixes. For this, and all the above reasons, most manufacturers implement this method gradually and use flow to make one product family, which necessitates ERP, MRP, or APS (advanced planning and scheduling) for the rest of the business. While lean/flow manufacturing leverages practices to stay ahead of actual demand, the traditional approaches better coordinates secondary, back-office systems like accounting and human resources (HR) management. Moreover, flow should be a company-wide strategy that does not only impact manufacturing.

Thus, many prospects will be more amenable to the flow manufacturing product designed to complement an MRP II or ERP system rather than replace it. To that end, for example, Oracle's and American Software's system offer full support for demand flow techniques on the shop floor while also providing a standard interface to the rest of the business systems for activities like purchasing, accounting and order management. A further example would be QAD that uses MRP for mid- and long-term planning, on the execution side however, it also caters to both JIT that is used for build-to-order or final assembly manufacturing, and to lean manufacturing, which is suitable for build-to-demand pull for finished and component items. Namely, most companies will still use MRP to ascertain longer-term dependent demand (i.e., based on forecasts), but replenishment will be based on independent demand (i.e., actual customer orders), with closely managed inventory buffers. Customer demand cannot be predicted months ahead, since nothing is that certain. Thus, many ERP systems are employing heuristics, algorithms, and even simulations to estimate the uncertainty in demand and effects of constraints on the shop floor. Understanding these variations then allows enterprises to assign flexible boundaries around material and capacity plans.

This is Part Two of a two-part tutorial.

Part One defined lean flow manufacturing.

Illustrating Differences

To illustrate the differences between these methods, let us compare the AMPM Convenience stores with Wendy's and McDonalds burger chains modes of operation.

  • For one, AMPM, whose slogan is "Always there to serve your needs," is an MRP-based planning operation, whereby hamburgers are made in daily batches, which often results with scrap leftovers and lost sales.

  • McDonalds, with the "Serving 46 million customers each day" motto, has done well at lean execution and dishing burgers out fast (often in one minute), but the problem occurs with make-to-order (MTO) requests like "I would like no onions, with five tomatoes instead, and I want American cheese instead of Swiss cheese, please." Thus, the franchises still have to stock finished goods buffers that cater for the demand pull, and the replenishment tact is adjusted to the buffer. Consequently, standard finished products are served from stock, while any special orders will take some time to be delivered.

  • Finally Wendy's, whose slogan is "Made fresh, tastes best," has done a pretty good job at made-to-demand because the final assembly of the final item does not happen until the actual order. The single customer line therefore creates pull, there are no stocked finished goods, and the buffer is only at the component level (e.g., pickles, cheese leaves, etc.).

The above burger examples would be quite applicable, for example, to computer or automobile manufacturers, since the issue becomes much more complex when the customer would say, "I don't want a beef burger, but rather an ostrich burger" or "I would like a car with a sunroof and painted with the US star spangled banner motif." These peculiar instructions would have to be ordered and propagated down the supply chain.

Consequent Vendor Responses

More recently, most vendors have further focused on their solutions for supply management and visualization. In other words, while still adding to their ERP capabilities (for example lean manufacturing and JIT management modules), many ERP and SCM vendors have been fleshing out the components of the extended enterprise (distributed order management, flexible schedules, customer self-service, consignment inventory, vendor managed inventory [VMI] and supplier managed inventory [SMI], replenishment management, and business-to-business [B2B] e-commerce). They have also been exploring the manufacturing community (i.e., consolidated visibility via a portal of inventory, production, product life cycle, payments, and invoices).

While forecasting might have had a poor reputation in manufacturing, recently there has been an increased awareness that with good collaborative planning and forecasting software, which would support collaborative sales and operations planning (SOP) processes, many manufacturers could improve their business performance (see Sales and Operations Planning). Like with the production planning, manufacturers need to remain on top of forecasting by leveraging much shorter review intervals than traditional quarterly updates. By taking forecasting more seriously and supporting it with smart, interactive tools, all the parties within the manufacturing businesses should be on the same page at the end of the day, which should result with agility. For manufacturers in volatile markets or with products with short life cycles, forecasts based on history only, often mean missing the true demand signals from customers or distribution channels.

Any supply chain planning (SCP) endeavor starts with a demand forecast based on a consolidated view of sales and operations plans from across the organization, remote manufacturing sites, and distribution centers. These demand plans are common to all manufacturing environments alike, and are instrumental for annual planning of resources, to monitor demand mix variations, and even as data for preliminary/rough-cut line balancing and kanban planning. Enterprises use SCP for strategic purposes such as to plan for resources across an organization, prepare for promotions, negotiate long-term contracts, establish objectives, and coordinate multisite operations, whereby sales and operations, inventory, distribution, collaborative demand management, transportation planning and other departments are all involved. Flow manufacturing does not address synchronizing around the supply chain, multiple partners, and suppliers, since it is merely a shop-floor execution tool.

Thus, it would be too nave to dismiss the need for proper planning, because regardless of how responsive an execution system may be, waiting for a chaos to happen and only then trying to act, would be as much of a disaster as it has been with compiling nearly ideal plans (through cumbersome algorithms) and never doing anything about executing or obtaining feedback about the plans' outcomes. As supply chains become more dynamic and operate in near real-time, the lines between planning and execution continue to blur, which bode well for their functional convergence. Companies need real time information from execution systems to develop and adjust optimal plans, while the execution side should benefit from more realistic plans for the sake of readiness, rather than to merely react after the fact in a firefighting fashion.


Manufacturers should understand the part and parcel of a more complete flow package so that they can decide how much functionality they need for their business. Although many ERP vendors have been professing lean or flow functionality, most of them still support some nuggets of pseudo JIT ways of accommodating mass customization. Just the support for kanbans or VMI to push inventory elsewhere (e.g., suppliers) rather than to reduce it across the supply chain is a far cry from a true support for lean or flow manufacturing. On the other hand, MRP should not be discounted as useless just like that, since it will often have an important purpose that could and should be used together with lean or flow practices (i.e., MRP will typically handle planning, while lean would deal with the execution). Manufacturers must also be fully aware of whether their systems use actual demand, sales forecasts, or a combination of the two in order to populate their MPS.

Also, manufacturers need to do some preliminary work before even thinking about deploying flow manufacturing software, such as adapting their plants to a flow production model. In other words, they will have to operate in work cells that build families of products (rather than functional work centers producing large batches of components/products), and they will have to establish rules for sending replenishment signals to their internal (i.e., preceding work station) and external suppliers. These changes will not happen overnight. The process should begin with the conversion of a few appropriate products with relatively simple production processes, and then progress to other product lines. Thus, many manufacturers happen to be in a hybrid production mode, with a part of the plant running according to the flow principles and the rest using traditional MRP methods. Consequently, the flow manufacturing application should provide reasonably easy integration with ERP systems.

More complete flow manufacturing systems should show most of the following capabilities (this is by no means a complete list):

  • Line design tools — for calculating the optimal line design, mixed-mode sequences, and daily rates or takt time of the production line for a given product family, and for balancing the sequence of events (SOE) in the line to keep the pace moving.

  • Demand smoothing — for calculating daily production requirements and the resources needed on each line, based on customer orders and on available capacity.

  • Kanban resizing and replenishment techniques — for quickly and automatically calculating the optimal size of kanbans and how often they should be replenished.

  • Operation method sheets (OMS) — graphical instructions explaining each step in the SOE defined for a given line.

  • The ability to reconcile the completion of a product with the back-end ERP system by back-flushing data about inventory, material, and labor, also with operator instructions and production reporting capability.

  • The ability to define engineering change orders (ECOs), to apply them to materials requirements throughout the supply chain, and to quickly reflect product engineering changes in the line design.

  • Multiple manufacturing styles can be used concurrently to manufacture a single product or support complex manufacturing processes.

  • The easy conversion of multilevel bills of material (BOMs) and routings into flat BOMs and SOE documents.

  • The product technology that fosters collaborative agreements and signals with suppliers and trading partners, and the requisite data collection, as the actual products flow through the supply chain.

  • The software supports mixed-model MPS, as it mitigates the risk of producing in large batches.

  • The software either has its own or links to a product configurator for real-time capable-to-promise (CTP), demand loading, and line sequencing when products are configured-to-order (CTO).

While the importance of the above varies with the manufacturing environment, most ERP vendors lag their flow manufacturing counterparts with regard to the dynamic creation and update of kanbans as well as in-line design tools. If opting for an ERP system, one should ensure that the auto-release function for scheduling MRP-recommended orders based on specific promise dates can be turned off, so that the pull-controlled parts are not subject to the auto-release (i.e. push) system.

comments powered by Disqus