Nortegubisian

These two trends, depicted in Exhibit 1, need and support each other. A big benefit of a build-to-order (BTO) supply chain is that large stocks of inventory can be removed. But once these buffer stocks are gone, the supply chain must be agile enough to respond to the variability inherent in the actual order stream. That’s where integration comes in. Real-time visibility and integration back through the supply chain allow suppliers to see the demand in real time and begin mustering the raw materials needed to respond quickly.

Converting the supply chain to BTO requires changing a lot more than manufacturing. Product design needs to change, too. Designs must be created that use standard parts and have fewer saleable configurations. The whole organizational mindset needs to change as well. For a company that is accustomed to planning, purchasing, stocking, kitting, manufacturing, and replenishing based on a forecast, it is a major shock when the forecast no longer drives any of these activities. The forecast is still needed to pre-position raw materials and do long-term capacity planning, but that is where its role ends.

The Business Challenge Faced

NMS Communications is a mid-tier electronics company with significant growth opportunities. The company posted $134 million in sales in the year 2000, which represented a 69-percent increase over 1999 sales. NMS produces printed circuit board assemblies for the telecommunications industry to enable transmission of voice over IP (Internet protocol). Its customers range from large original equipment manufacturers (OEMs) to small start-ups that provide system-level equipment to communications carriers such as telephone companies. NMS Communications’ suppliers include the makers of integrated circuits, chips, and other electronic components. Manufacturing is completely outsourced to SMTC Manufacturing Corp. in Toronto, a contract manufacturer. SMTC also builds the new product prototypes for NMS.

As part of a strategy to become a trusted partner of its customers, NMS has embraced supply chain management as a key differentiator. The company was originally positioned squarely in the middle of the telecommunications equipment supply chain. (See Exhibit 2.) Finished goods were sent from the contract manufacturer to NMS to be tested and stored. Customer orders were then filled from this inventory. NMS had visibility to only one level in each direction: to customers and to SMTC.

NMS recognized that it needed to change its positioning in order to streamline supply chain operations. Instead of being in the middle handling the physical goods, NMS wanted to sit above the supply chain, orchestrating and monitoring its daily performance. The company also wanted to provide visibility across the supply chain for both itself and its trading partners. This would allow NMS to focus on its core competencies—design engineering and supply chain management, not supply chain execution. (Exhibit 3 shows the new supply chain positioning.)

This repositioning would also solve the problem of scalability. Under the old model, growing the company meant adding people and/or facilities in testing, warehousing, distribution, shipping, and transportation. The new model scales more easily, requiring fewer additional resources to support growth.

The telecommunications sector is a highly competitive marketplace with equipment technology leapfrogging itself every few months. There is enormous pressure for rapid time to market and for quick turnaround of customer orders. Although NMS was performing as well as most of its competitors in these areas, it had all the ills of a forecast-driven supply chain: inadequate delivery predictability, high inventories, long customer order cycle time, and raw material shortages. End-of-quarter deals at multiple points along the supply chain and the practice of pulling in orders from the next quarter caused huge peak loads on manufacturing and distribution. Further, the process of building and testing prototypes of new products was slow, unpredictable, and frustrating to the NMS engineering groups. The time was right for a change.

Pursuing a New Vision

NMS and Avicon formed a small project team and drew up a simple project plan in three phases: (1) define a build-to-order vision; (2) design new information systems and business processes; and (3) build and implement new information systems. Phase 1 included only one project—a quick, four-week quantitative assessment of the supply chain. This assessment, which included simulation modeling, helped define the new supply chain vision. Phase 2 lasted 17 weeks and focused on designing new business processes and information systems. This phase included six subprojects:

1. Demand Planning/Collaborative Planning.

2. Manufacturing Scheduling.

3. Materials Management.

4. Order Management.

5. New Product Introduction.

6. Overall Supply Chain Architecture.

Phase 3, still under way, focuses on implementing the new processes and information systems in the areas listed above. Very early on, before beginning Phase 1, the project team laid out the entire 38-week program plan and estimated the cost of each subproject. Sticking to this plan—in terms of both time and cost—proved to be crucial to maintaining corporate sponsorship throughout the program.

Conducting the Quantitative Analysis

“Just how close to build to order can we get?” A quick (one-month) quantitative assessment of the supply chain answered that question. It also helped prioritize which problems to attack first.

The assessment began by looking at customer buying patterns, including both the day-to-day order stream and the quarterly pattern of demands. Instead of showing a consistent flow of large orders, the analysis revealed an erratic pattern of mostly small orders, including many for one unit of one product. The quarterly profile showed the classic end-of-period “hockey stick.” Spikes in demand made building to order extremely difficult because it is not practical to have enough spare capacity to handle these temporary peaks. The assessment also focused on the manufacturability of the products themselves. The team found little use of common parts among the products as this had never been a priority of the design engineers. If products have many raw materials in common, it is much easier to switch from building one to the next. Greater commonality also means there are fewer parts in total to order, stock, and manage. But in the case of NMS, 39 percent of the components were unique to a single product, and fully 70 percent were used on four or fewer end products.

Contract manufacturer SMTC builds the NMS populated printed circuit boards on high-speed surface-mount lines. These state-of-the-art lines can “stuff” boards very quickly, with some operations happening faster than the eye can see. The lines are great for building thousands of the same board in runs that last several days. But the time required to change the line over from building one board to building the next is typically six hours. Simple math tells you that the manufacturer can only build two to three different boards each day. Unfortunately, NMS customers order, on average, 16 different products each day—and some days many times that. A simulation of BTO production determined that line changeover times would need to be under one hour to respond effectively to that order pattern.

The project team studied in detail both forecast accuracy and inventory. Forecast accuracy was poor—under 50 percent at the SKU level. In a BTO system, the forecast is used mainly to pre-position raw materials and to predict rough capacity needs. Forecast accuracy at the raw materials level is affected significantly by part commonality. Parts with high commonality (used on many end products) benefit by the law of large numbers; that is, some end product forecasts are too high, others too low, but the errors tend to cancel each other out. On the other hand, forecasts for unique parts can be, and often are, highly inaccurate.

The team knew from the beginning that creating a build-to-order system with zero inventory would be impossible to achieve. Yet the results of the quantitative analysis showed that NMS could actually get a lot closer than ever thought possible. But for this to happen, a few business practices had to change. The first was leadtimes. Under the old system, NMS would quote a standard leadtime of 30 days for all orders. Delivery would then occur in anywhere from 10 days to 30 days or more depending on the availability of the items. Under the new approach, once system implementation is completed, leadtimes will differ depending on the product and the order size. Old, slow-moving products will continue to have a 30-day leadtime, while orders of new products will have a 10-day leadtime initially, phasing to a six-day leadtime after six months. Large orders will have a 10- or 15-day leadtime. These demand management steps enabled NMS to spread the load on manufacturing from the end-of-quarter peak into the “valleys” on either side. Both the NMS sales team and the customers understood that big end-of-quarter deals hurt delivery performance, and they welcomed the new, faster, more predictable business practices.

The team also backed away from having zero inventory of finished goods. Products were divided into several categories based on cost and velocity. Lot size rules were then established by category. NMS will never build product unless forced to do so by a customer order. But when it does have to build product, it will build enough to both satisfy the order and to leave one “lot” sitting on the shelf. A lot for a low-cost product is seven days of supply, or a minimum of 15 units. A lot for a high-cost product is three days of supply, or a minimum of five units. The new lots range from five for the slowest movers to 100 for the top movers. (Compare this to runs of 500 to 1,000 typical in the electronics manufacturing industry.) Once fully tuned, this approach will yield well over 20 inventory turns per year. Under the old forecast-driven system, inventory turns were in the single digits.

The quantitative analysis also highlighted the importance of having an uninterrupted supply of raw materials, including some parts with one-year leadtimes. With only a tiny buffer stock of finished goods, manufacturing in a build-to-order system is exposed to the full variability of the order stream. Therefore, it is critically important to have adequately pre-positioned raw materials ahead of the manufacturing lines. Based on historical and forecast data, the team calculated the quantity of each part needed to support BTO.

Capturing the Quick Wins

The project team presented the results of the Phase 1 quantitative analysis and outlined the new supply chain vision to the NMS executive staff. (Exhibit 4 on the following page contrasts the new supply chain vision with the old.)

 

The modeling and analysis conducted in this first phase served to dispel many misconceptions about the NMS business environment. For one thing, it showed the true nature of the business and the order profile. The analysis corrected a mistaken belief that a typical order was 25 to 100 units. In fact, the order size was one to 10 units 80 percent of the time. The analysis further showed that the actual number of different permutations of products to be built was 127, not 50 as believed. Interestingly, the findings failed to support the notion that the company had a core group of major customers that bought high volumes of several products. In reality, even the largest customers placed many small orders for a wide variety of products.

Most importantly, the quantitative analysis demonstrated that NMS could streamline its supply chain enormously and get quite close to a BTO model. This ended the debate over whether or not to move forward while generating critical excitement and enthusiasm among the executive staff. This momentum later turned out to make all the difference as the economy faltered and the telecommunications equipment industry retrenched. Only the BTO initiative and new product development survived the company’s budget cuts.

Building on the momentum generated by meeting with the executive staff, NMS began immediately pursuing several “quick wins.” One such opportunity centered on product end-of-life. The company did not have a policy or mechanism governing when to phase out a product. Consequently, many slow-moving old products still were available for sale. Some of these were languishing in finished-goods inventory. Others were out of stock, the parts needed to build them hard to find. The NMS Customer Operations Group quickly identified about one-third of the product line that could be phased out and initiated a process with sales, marketing, and finance to accomplish this. Another quick win focused on common parts. Having learned of the importance of standardized common parts in product design, the NMS engineers immediately reviewed the entire set of raw materials and began reducing the set of acceptable parts. They now have the information system tools that flag any new components being designed into products.

Redesigning the Supply Chain Processes

Moving from a forecast-driven to a build-to-order system meant redesigning three major supply chain processes: new product introduction, demand planning, and order management. In each of these areas, a joint team from NMS (supply chain, engineering, and sales), SMTC, and Avicon laid out the “as-is” process, identified the weaknesses, set the goals for the new process, and then mapped out a new process. The new product introduction process was fraught with delays, missed communications, and frustration. The NMS design engineers would begin work on a new design and seek input from SMTC. Several weeks would pass without hearing anything from the contract manufacturer, and design engineers would just plow ahead with product introduction. The problem: Flaws that should have been caught early in the prototype runs were either built into the products or had to be corrected later at much higher expense. After several “home and away” visits and a “getting to know you” dinner with SMTC, the team designed a new process that featured much earlier and more frequent information exchanges. The goal is now to use a workflow engine to manage these information exchanges. The team expects that the new process will shave six weeks off of the time required to bring a new product through development and testing. Product quality also is expected to improve.

The demand planning process had two major problems. First, operations received the forecast too late in the process for them to have any meaningful input about its feasibility. Second, once operations converted the demand plan into a supply plan, there was no feedback mechanism to sales management or to the executive staff. Often the official demand plan included things that operations knew could never happen. Again, the project team laid out the as-is process, identified the flaws, and jointly created a new demand planning process. This improved process, which will be supported by new demand planning and supply chain planning software, incorporates frequent real-time communications between NMS sales, NMS operations, and SMTC as they iteratively develop the final demand plan. Simulation will be used continuously to determine the feasibility of changes to the demand plan.

Among the three key processes, order management probably changed the most. Before the switch to build-to-order, SMTC produced assemblies to a forecast and shipped them to stock at NMS. When a customer order arrived, NMS would pull the items from inventory (if available) and ship them out. SMTC never saw a real customer order, and NMS had 30 days to ship the orders to customers. In the new process, orders come into NMS. After employees run an instant credit check and apply a due date stamp, the orders are immediately posted to a Web site visible to SMTC. The contract manufacturer can monitor the inflow of customer orders in near real time, enabling it to plan the next day’s production effectively.

With new supply chain planning software and with their databases synchronized, NMS and SMTC can now quickly simulate the impact of taking large or unusual orders. If there is a problem in accepting the order, both companies can see the cause immediately (capacity or material availability), and if raw materials are limited, they can see which other customer orders are competing for those parts. This capability allows NMS to quickly respond back to customers with choices.

The Manufacturing Challenges

The most technically challenging changes involved in moving to a build-to-order system were in manufacturing. These changes revolved around how manufacturing itself was done, how production scheduling was done, how prototypes were built, and how raw materials were managed. SMTC needed to shift from a low-mix/high-volume capability to a high-mix/low-volume capability. Put another way, it needed to economically produce runs of five to 10 units, instead of the 500 to 1,000 units the lines were designed for.

“How long are your changeover times?” That is the first question that needs to be asked when thinking about adopting a demand-driven supply chain. In the case of NMS, the six-hour industry standard changeover time from making one product to making the next was woefully inadequate. NMS customers order anywhere from 10 to 20 different products each day. If it takes six hours to change the line over, then you can build at most two to three different products each day. Our simulation models showed that the contract manufacturer needed to reduce the changeover time to under one hour to keep up with actual customer demand.

By gathering the best manufacturing ideas from a tour of small job shops and large-volume producers and reading literature on the subject, the project team designed a new manufacturing model. Though that model had production rates somewhat lower than existing rates, it could achieve more rapid turnovers. SMTC’s own manufacturing engineers then took this model and adapted it to their high-speed lines. Now, by shifting many steps off-line, by using dedicated feeders, and by using numbered, pre-staged feeder racks, SMTC has reduced the longest changeover times to 30 minutes. And some are as short as five minutes!

Two huge benefits flowed from this dramatic reduction in changeover time. First, the number of different products that could be produced in a given day increased significantly, while the number of surface-mount lines required to support NMS’ volume was actually reduced by two-thirds. Second, faster changeovers, coupled with improved component sourcing strategies, resulted in manufacturing cost reductions of approximately 15 percent. In short, going to a fast-response build-to-order manufacturing system saved cost as well as cut cycle times and eliminated inventories.

In the past, NMS would provide SMTC with a one-year expected usage estimate for key component parts based on demand forecast. This would be accompanied by a purchase order (PO) for assemblies to be built for the first six months. SMTC would then convert the six-month PO into a series of two-week production schedules and issue purchase orders to its suppliers to acquire the parts. As time went by and the actual demand differed from the forecast, NMS would request changes in SMTC’s production schedules. The contract manufacturer, in turn, would ask their suppliers to speed up or slow down deliveries of parts accordingly.

Because the forecast was always wrong, schedules were constantly being changed and parts constantly expedited. All too often, the critical raw components in short supply were already sitting in finished-goods inventory—but soldered onto the wrong product. The truth was that having no set two-week production schedule at all was preferable to having created a “false” two-week production schedule and then changing it constantly. Customer orders now flow into SMTC daily and production scheduling is now done every evening for the next day’s production. No longer is anything built based on a forecast. All production is in response to actual customer orders.

Another manufacturing problem was that new product prototypes were built on the same surface-mount line as production runs of revenue-generating products. Frequently, these prototypes competed for constrained line capacity and limited raw materials such that one activity had to give way to the other. This presented NMS with an unacceptable choice: either delay developing a new product or delay building an existing product for sale. Now, SMTC has designated a specific production line for prototype builds only. The raw components needed for prototype builds are also physically segregated from the components used for volume runs.

Raw component availability becomes the critical issue once the buffer stocks of finished and work-in-process inventory are removed. BTO only works if the raw materials are there when you need them. But nobody wants to own a large supply of this inventory. SMTC’s solution incorporated a blend of strategies: Two weeks’ worth of component inventory is kept on the factory floor in storage carousels; another two to four weeks is kept in an adjacent third-party warehouse. Agreements with distributors call for them to keep 60 days of supply on hand. Finally, an active broker market for electronic components can be used as a last resort to source common parts.

New Information Systems Needed

NMS Communications’ existing information systems were typical of many companies of its size. It had a mid-tier enterprise resource planning (ERP) system, a sales quoting system, and several product design (computer-aided design or CAD) systems. The systems generally supported a traditional forecast-driven supply chain complete with monthly purchase orders, lots of inventory, and monthly planning cycles.

NMS wanted to convert its IT systems to support an on-demand supply chain while at the same time getting ahead of the integration curve. To accomplish these objectives, NMS needed new software that would:

• Perform demand planning and supply chain planning/simulation.
• Connect its internal systems.
• Integrate with its external trading partners.

NMS began by implementing a product data management system. This system, which allowed the company to maintain its own product-level and component-level data, was crucial in helping engineering increase the use of standardized parts.

The project team then initiated a rigorous evaluation and selection process to examine software for demand planning and supply chain planning and simulation. NMS envisioned receiving forecasts from three main sources: (1) customers, in a wide variety of formats, including some automated data feeds; (2) the sales force, through the sales quoting system; and (3) statistical forecasts, generated by the demand-planning software based on history. The demand-planning software needed to be able to combine these forecasts, which came in different formats with different part numbers, into a single useable demand plan. It also had to support the new demand planning process, which now included frequent collaborative discussions between sales, operations, SMTC, key customers, and the NMS executive staff.

The supply chain simulation software also needed to support both the order-management and the demand-planning processes. It needed to monitor the order stream and calculate a capable-to-promise date for each order. This date would then be compared with the rules-based order due date, with users alerted to any conflicts. The software also needed to quickly perform rigorous simulations to determine if the materials and capacity existed to accept large unplanned orders. Finally, it needed to calculate the feasibility of alternative forecasts. Based on the comprehensive evaluation process, Manugistics NetWORKS Demand Planner, NetWORKS Supply, and NetWORKS Collaborate modules were selected and implemented.

NMS created a new “hub-and-portal” IT architecture to integrate its internal systems and to integrate electronically with key customers and suppliers. (This architecture is depicted and described in the sidebar at the end of this article.) As its name implies, the new architecture is built around an EAI (enterprise application integration) hub, which is a centralized point of access for information from business applications within the enterprise. The architecture incorporates the following capabilities: communication and messaging, data transformation and routing, message abstraction through common business data objects, business process automation/workflow, and application connectors/adapters. IBM’s MQ Series platform was selected for the hub.

The portal connects NMS with external trading partners. It provides multimode (extensible markup language or XML), electronic data interchange (EDI), the worldwide Web, and extranet access to trading partner resources. The portal also enables similar transactions to be performed multiple ways, for example, order entry via Internet, EDI, XML, or extranet connection. The portal also facilitates process integration, visibility, and collaboration with trading partners. For example, it can accept orders, post status updates, or collaborate on forecasts. The portal leverages common business data objects to simplify integration with back-end systems via the hub. And importantly, the portal can accommodate a wide variety of trading partners’ capabilities. The project team selected IBM’s WebSphere and Partner Agreement Manager products for the portal.

Positive Customer Reaction

Customers were involved throughout this initiative. NMS wanted to know how its customers wanted the supply chain to perform, how they wanted to collaborate, and how the two parties could share information electronically. The company made numerous visits and presentations to customers describing the project and soliciting input. Overall, there was tremendous enthusiasm. Customers overwhelmingly welcomed a supplier that could react quickly and predictably, thereby allowing them to meet the service levels of their customers while at the same time reducing or eliminating their raw materials inventories. Because of historically long order-cycle times and unpredictable fill rates, some customers had maintained their own safety stock of NMS products. The BTO initiative will help relieve them of this financial and physical burden.

As the project progressed, it became apparent that many customers were themselves striving to achieve a demand-driven supply chain and enthusiastically embraced the NMS effort. Nearly all customers wanted to begin collaborating, although their respective capabilities varied widely. For example, some were ready to exchange XML files; others were still struggling to send a spreadsheet attached to an e-mail.

NMS was anxious to accelerate the integration because some customers, absent a strategy from NMS, were already inventing their own communication methods. Typically, these would require NMS to perform redundant work, type data into multiple extranet sites, and even manually interact with some customers’ ERP systems.

NMS Communications’ new hub-and-portal architecture addresses these issues. It allows the company to send and receive data by a variety of modes to accommodate the needs and abilities of customers and suppliers. Information moved from NMS to the customers includes: order status, shipment notification, available-to-promise/capable-to-promise information, inventory positions, and quality data. Information moved from customers to NMS includes: forecast information, customer inventory of NMS products, potential upside orders/opportunities, build schedules, and order stream/usage reports.

Reflecting on the Lessons Learned

Reflecting on its experiences in creating a demand-driven, build-to-order supply chain, NMS believes it has learned some important lessons:

• Converting to a build-to-order process involves changing “everything”—processes, software, skills, job descriptions, relationships, and most of all, thinking.
• A thorough quantitative analysis of the business environment and simulation of the BTO model is crucial to establishing the project’s vision and rallying momentum. Other change programs failed in the past because they were not perceived as being based on “facts.”
• Once the enthusiasm catches on, people will find and capture the quick wins without having to wait for the new processes or IT systems to kick in.
• Contrary to conventional thinking, moving to a BTO model does not make the forecast less important. In fact, with the removal of buffer stocks of inventory, the forecast is even more crucial for pre-positioning raw materials.
• Software selection can be an emotional, even contentious, experience. The project team used a meticulous, fact-based process to neutralize the various biases that might exist. Throughout the software evaluation process, you need close cooperation with the information systems support function.
• Creation of a simple project plan early on, complete with cost estimates for each subproject, helps sustain corporate support throughout the project. This is especially true when the economy heads south. Bringing the sub-projects in on time and on budget is of paramount importance to maintaining corporate sponsorship.
• Customers have a huge range of e-business desires, strategies, and capabilities. Some are much better positioned for supply chain integration than others.
• For a project such as this, it’s important to find the right contract manufacturer and consulting partners. Nearly all of the contract manufacturers visited by the project team claimed that they could do BTO but in reality, none could. Achieving BTO with such a high-mix, low-volume order stream was a significant accomplishment. NMS also needed a consulting partner that had the right skill mix and the passion to inspire the company to change its business.

Looking ahead, NMS is well positioned to take off as soon as the economy picks up. This project to reinvent the company around a demand-driven supply chain and to get ahead of the electronic integration curve was one of only two major NMS projects to survive the business-related budget cuts. NMS boldly moved ahead with the project at a time when most others were hunkering down. The company believes that it will emerge with a sleek, flexible supply chain that can easily integrate electronically with customers’ and suppliers’ systems.

The Hub and Portal Architecture

NMS Communications and Avicon are nearing completion of a hub-and-portal information systems architecture both to integrate the NMS internal systems and to exchange information with trading partners. The hub manages the exchange of data between the NMS internal business applications and provides the appropriate data transformations in support of the company’s supply chain execution processes. Applications connected by the hub include an ERP system (Syteline), a sales forecasting system (Overquota), and a supply chain planning system (Manugistics). A direct connection is already in place between the product data management system (Agile) and Syteline for the exchange of bill-of-material (BOM) data. This interface will be moved to the hub in a future phase.

The building blocks used to construct the hub include communications middleware infrastructure provided by IBM’s MQSeries, translation and transformation capability provided by MQSeries Integrator, and business process automation capability provided by MQ Workflow. The adapters are the connection between the business applications and the hub. They work through the application’s interface to send data to and receive data from the hub.

The portal enables NMS to communicate with its suppliers, contract manufacturers, and customers. RosettaNet standards are used for data and process interchange. IBM’s Partner Agreement Manager (PAM) supports and manages these communications. PAM also supports NMS e-mail interactions with customers, who e-mail spreadsheets on forecasts, build schedules, and so forth. The portal’s browser feature, provided by IBM’s Webserver and Websphere Application server, facilitates online order entry and inquiry through the Internet. Customers can log in, enter a new order, and inquire about the status of an existing order. Security services for the portal are provided by IBM’s Domino Directory.

 

Bruce C. Arntzen is vice president of supply chain consulting at Avicon. Herbert M. Shumway is vice president of supply chain at NMS Communications.