June 2000 Big Benefits for Big Blue A variety of projects demonstrate how operations research delivers business value to IBM By Brenda Dietrich, Nick Donofrio, Grace Lin and Jane Snowdon After many years of hard work in the field, operations research has become an integral part of IBM's core. World-class research in OR and the ability to apply the results to solve business problems have increased IBM's competitive edge — IBM's operations research team helped save hundreds of millions of dollars, while improving operations and competitive strategies. Three prestigious awards over the past year — the Franz Edelman Award for Achievement in Operations Research and Management Science, the Daniel H. Wagner Prize for Excellence in Operations Research Practice, and most recently, the INFORMS Prize — have also enhanced IBM's operations research reputation. "As we set out to reengineer our company," said Nick Donofrio, senior vice president of Technology and Manufacturing at IBM, in accepting the INFORMS Prize, "we knew that having world-class information technology simply wasn't enough. Since speed and efficiency were fundamental elements of our strategy, we needed entirely new approaches to our operations management." As the world's largest computer hardware, software and services company, with total gross revenues of more than $87 billion in 1999 and operations in 164 countries, IBM has no shortage of operations to manage. Headquartered in Armonk, N.Y., it is organized around six business groups that provide technology, software, personal computers, servers, solutions and services to its customers. The Technology and Manufacturing Group develops and manufactures storage, semiconductor and communication technology products for use in IBM systems and for sale in the Original Equipment Manufacturer market. The Software Group provides systems and application software to support stand-alone, distributed and networked computing. The Personal Systems Group develops and manufactures client and mobile computers for small business and personal use. The Enterprise Systems Group is responsible for the S/390, AS/400 and RS/6000 products. Solutions are developed and sold by the Sales and Distribution Group, and a wide range of information technology related services are provided by the Global Services Group. IBM manufacturing and service organizations support more than 3,000 hardware and 20,000 software products. IBM's Research Division consistently produces leading scientific and technical contributions to IBM's product portfolio, allowing IBM to lead the world in U.S. patents for the last seven years, with 2,756 patents awarded in 1999. Its researchers include five IBM Nobel Laureates, and have garnered four National Medals of Technology and three National Medals of Science. During the last decade, IBM Research has extended the traditional model of scientific and technological research to include work with IBM's customers. Bringing IBM scientists and customers together to tackle real-world business problems has advanced the application of information technology as well as the underlying science and mathematics. To conduct this "research in the marketplace," interdisciplinary teams coalesce to solve problems in such areas as finance, insurance, healthcare, travel and manufacturing. The application of operations research to IBM's own complex design, supply chain and services processes represents yet another opportunity to bring research scientists together to solve complex real-world problems. Operations Research in IBM Research Over the past three decades, operations research groups within the IBM Research Division have conducted numerous analytic studies and developed many applications across the global IBM enterprise. Analytic studies have addressed strategic decisions such as global manufacturing network designs and supply chain inventory policies. Software applications have been deployed for use in facility design and for weekly or daily manufacturing planning. A review of four recent projects illustrates the increased scope of operations research activities within IBM, as well as the integral nature of the collaboration between the business units and the Research division. These four OR projects alone are credited with saving IBM hundreds of millions of dollars. Service Call Management IBM Global Services employs more than 7,000 customer service representatives (CSRs) who service approximately 22,000 calls each day. These calls range from performing routine service, to installing hardware upgrades, to repairing high end machines that run mission critical applications, to repairing PCs, printers and copiers. Managing this workforce efficiently is a major challenge. As part of an effort to improve customer services and reduce costs, IBM Research developed an assignment engine that assigns CSRs to service calls, considering such factors as customer service entitlement, location, available parts, skills and labor costs. This engine also enables IBM Global Services to respond dynamically to changing conditions. Based on pilot results, IBM Global Services selected this engine as the strategic tool for scheduling all maintenance business. It is integrated with a call management system being deployed by IBM Global Services. The current business case projects $68 million savings over five years in North America and $76 million over five years for rest of the world when fully implemented. Mark Carey, Worldwide Solution Integration manager, first approached IBM Research for assistance in developing a call center scheduling system after hearing about the algorithms IBM Research developed for the Kasparov vs. Deep Blue chess match. Carey was delighted to find colleagues at IBM Research who could easily relate to his problem. "We went from a glimmer in the eye to something in production in just five months, which was remarkable for such a complex system," Carey says. The key ingredients that differentiate this CSR scheduling engine from other workforce scheduling tools are: 1. the ability to optimize a large problem quickly; 2. the ability to respond to changing conditions; 3. the simultaneous consideration of a spectrum of cost factors and assignment possibilities in assigning calls to CSRs; and 4. a novel way of considering part requirements and availability. The engine has two main stages. In the first stage an initial solution is computed with no backtracking. The CSR availability data is analyzed to determine availability intervals, and the jobs are ordered according to several criteria such as customer service entitlement and cost. Then the calls are considered one by one and each is assigned to the "best" slot that has the required skills and parts. The second stage of the optimization consists of local optimization within the calls assigned to each CSR and then global optimization in which calls are swapped between CSRs. This stage also groups calls to save CSR travel time. The main challenge in delivering the scheduling function was engine speed: IBM Global Services required that the schedule be recomputed every 10 minutes in order to react to changing conditions, such as new calls or CSR delays. The scheduling engine has been piloted for over a year on five work groups having different CSR availability and call volume characteristics. Using this engine to schedule the calls, a significant reduction in the total time spent on a call (between 10 percent to 35 percent) was achieved in four of the five work groups. An additional benefit was that the idle time of the CSRs was substantially reduced. This idle time decreased by a factor ranging from 5 to 1.75 across the work groups. Semiconductor Manufacturing Capacity Semiconductor manufacturing requires many hundreds of operations per product, a few hundred tool groups and weeks of manufacturing time. Tools cost up to $10 million, and the lead time for purchasing a tool may be up to a year. Capacity planners use long-term demand forecasts to determine the tools to purchase, but capacity planning is complicated by the fact that there are typically several different tools capable of each process step, and the time required for a process step depends on both the product and the tool. Thus, the capacity of a semiconductor line depends on both the mix of products and on the assignment of products to tools. Until 1995, capacity planners responded to a change in demand by running multiple product-mix scenarios through a spreadsheet, guessing how to change the volume or tool assignment of products to maximize profit and minimize additional tool purchases. Only a few scenarios could be analyzed, so there was no way to know whether the best solution found was anywhere near optimal. Managers needed an improved method for capacity planning — specifically a system that would identify the optimal product mix, optimally allocate products to tools, identify gating tools, and determine a minimum set of tools required to produce a given demand. They also needed a system that accurately modeled allocation rules and produced answers quickly. IBM Research developed a linear programming model in which the decision variables represent the production volume of each product and the allocation of each product to each tool. The constraints reflect production limits, the required process steps and the available tool capacity, while the primary objective is to maximize profit. Additional modeling constructs capture the preferred order in which tools are used and aid in identifying bottlenecks. The Capacity Optimization Planning System (CAPS) uses a graphical user interface on a PC client that presents the wafer-start and tool-set data files to the user in a convenient spreadsheet format for easy editing. Thus, the user can remain focussed on the familiar process of capacity planning. IBM's linear program solver Optimization Solutions and Library (OSL), running on a UNIX server, has been "institutionalized" for the non-expert, with most input data automatically formatted and downloaded from a tool planning database residing on a mainframe. CAPS has been in use at IBM's largest semiconductor manufacturing site in Burlington, Vt., since mid-1995. "CAPS is used for all tool planning and is credited with saving millions of dollars in capital expenditures," said John Waite, IBM Burlington's manager of Site Operations. Capacity planners also use CAPS to reconcile demand forecasts with manufacturing capacity. For example, when one customer requested an increase of several hundred wafers per day, planners used CAPS to rapidly respond, ultimately capturing additional revenue. CAPS has also been used to assess the feasibility of technology conversions at other IBM fabricator sites, contributing to IBM's decisions regarding the production of new products in existing manufacturing sites. "CAPS has been used not only to balance our product portfolio offerings given the uncertainty of the marketplace," Waite continued, "but also to analyze the impact of cycle time commitments on fabricator output. We have no other system that can be utilized to quickly analyze these capacity tradeoffs." Supply Chain In 1994, IBM embarked on a global supply chain reengineering initiative to reduce inventory levels and write-offs while keeping enough inventory throughout the supply chain in order to respond to customer requirements. To help with this, IBM Research developed an advanced supply chain optimization and simulation tool, called the asset management tool (AMT), for deployment throughout IBM. This tool integrates graphical process modeling, analytical optimization, simulation, animation and activity-based costing, allowing quantitative analysis of extended supply chains. The asset management tool analyzes supply chains and identifies optimal business policies. AMT addresses a range of logistics topics, including inventory and customer service level target, supply network configuration, product structure, channel assembly, parts commonality, postponement, supplier terms and conditions, vendor-managed inventory, lead-time reduction, fulfillment policies and forecast accuracy. AMT makes it possible to evaluate supply chains on the basis of financial tradeoffs associated with various configurations and operational policies using activity-based costing and financial analysis. AMT has been used in a several IBM business units as well as in external reseller organizations. Bob Moffat, general manager of Manufacturing, Distribution, Procurement and Reengineering for the IBM Personal Systems Group (PSG), recognized a critical business need to develop optimized supply chain strategies to cope with the large volumes, slim margins and dropping prices of the personal computer market. "We needed to better utilize our distributors and value-added resellers without unnecessary inventory that crippled cost competitiveness," Moffat said. Using the asset management tool, IBM PSG worked with IBM Research to reduce the overall pipeline inventory by over 50 percent from 1997 to 1998. Lookback expenses (payments made to distributors and resellers due to product price reductions) were reduced by more than $750 million in 1998. In addition, the cash-to-cash cycle time from component procurement to product sale was reduced by four to six weeks, resulting in a reduction of 5 percent to 7 percent of overall product cost. "We believe AMT has served as an invaluable asset to IBM in developing world-class supply chain management policies" said Moffat. Jim Manton, president and COO of Pinacor, one of IBM's major distributors, echoed the sentiment. "The results that the IBM team delivered on the supply chain analysis helped Pinacor identify opportunities for optimizing the product flow between our companies," he said. Other IBM units have also benefited from the use of AMT. The AS/400 Division used the tool to analyze and quantify the impact of product complexity. The analysis contributed to feature reduction, substitution of alternate parts and delayed customization. AMT was also used to analyze the tradeoff between serviceability and inventory by the IBM QuickShip Program. The results helped maximize the efficiency of the program, reducing the operational cost by as much as 50 percent. "The AMT tool has found application in almost every supply chain within IBM," said Jean-Pierre Briant, former IBM vice president for Integrated Supply Chain. "It helps us understand our extended supply chain — from our suppliers' suppliers to our customers' customers — and we have been able to deploy AMT to assist external companies in managing their supply chains with very effective results." Following the success of the asset management tool, IBM launched several new projects in 1999: supply chain strategies for new routes to market, improved forecasting techniques for flexible product structure and risk assessment methodology to analyze various demand scenarios on profitability. These projects will help IBM improve its extended supply chains with e-business technology to maintain a competitive edge. Supply/Demand Reconciliation The Supply Capability Engine (SCE) is a software system for solving resource allocation problems in IBM's Supply Chain. SCE addresses the allocation of scarce raw materials and capacity through a multi-echelon bill of materials to satisfy demand. Prior to the deployment of SCE, IBM's supply/demand reconciliation process — from the generation of a customer demand outlook, through the propagation of internal demands throughout the network of IBM plants, through the assessment of manufacturing capacity and external component availability, to the commitment of resources — took more than two months to complete. The use of SCE reduced this planning cycle to just 20 days. The core resource allocation problem, known as the "implosion" problem, is solved using linear programming and greedy heuristics. SCE includes modeling constructs to address multi-plant enterprises, semiconductor planning and component aggregation, as well as a suite of algorithms that provide sensitivity analysis beyond the standard shadow pricing of linear programming. A key indicator of SCE's value is its ability to identify the production bottlenecks and to indicate how to fix them. SCE has been used in production in IBM since 1996, primarily to provide an end-product commitment to marketing organizations on a weekly or monthly basis. The use of SCE in the AS/400 division led to a significant improvement in customer responsiveness. In a Personal Systems Group plant, SCE replaced an extensive manual process with a quick, consistent and reliable process, and enabled "what if" analysis that was not possible with the previously available tools. IBM also used SCE as part of an enterprise-wide supply/demand planning process. In that deployment, the scope of the implosion problem extended from the lowest level components and procurement though the entire IBM supply chain to "box" plants that produce mainframes, workstations, business systems, PCs and storage systems. This process also addressed independent demands for parts. With over 300,000 part numbers, a one-year planning horizon and weekly planning periods, these implosion problems were huge. The enterprise planning process was used monthly for over a year, during which SCE consistently solved these massive implosion problems. The enterprise process was eventually replaced with a distributed planning process that requires each division to run its own supply/demand planning cycle on a weekly basis. The Personal Systems Group runs SCE at each of its four plants. The Microelectronics Division runs SCE as a central process to generate a weekly "available-to-promise" schedule in a multi-plant application. Finally, the corporate supply chain group runs SCE for several plants, including both box and component plants. Stuart Reed, IBM's vice president of Integrated Supply Chain Development and Deployment, describes business transformation as having two phases. "First you must fix the processes," Reed explains, "and then you apply information technology to cement the transformation. In the Microelectronics Division, SCE became the cement that allowed us to achieve better responsiveness and significant reductions in cycle time. Today we are able to commit instantaneously to the vast majority of our customers; in the past, it took on the order of days." In summary, IBM Research has developed and deployed many advanced analytic techniques and management tools to help the IBM Corporation better manage operations, deliver better products and develop more competitive strategies. In today's competitive business environment, the application of operations research has enabled IBM to quickly respond to customer's needs, cut operating costs and increase market penetration. Acknowledgments The authors acknowledge the members of the IBM T. J. Watson Research Center who contributed to these projects and reviewed earlier drafts of this paper. Nick Donofrio is IBM's senior vice president of Technology and Manufacturing. Brenda Dietrich is senior manager of the Optimization Center, Grace Lin is manager of Extended Enterprise System Research, and Jane Snowdon is a research staff member in the Mathematical Sciences Department, all at the IBM T. J. Watson Research Center. Table of Contents OR/MS Today Home Page OR/MS Today copyright © 2000 by the Institute for Operations Research and the Management Sciences. All rights reserved. Lionheart Publishing, Inc. 506 Roswell Street, Suite 220, Marietta, GA 30060, USA Phone: 770-431-0867 | Fax: 770-432-6969 E-mail: lpi@lionhrtpub.com URL: http://www.lionhrtpub.com Web Site © Copyright 1999, 2000 by Lionheart Publishing, Inc. All rights reserved.