m6a2sc
The Evolution of Supply-Chain-Management Models and Practice at Hewlett-Packard
H A U L. LEE Departrneut of industrial Engineering and Engineering Management
Stanford University Stanford. California 94305
COREY BILLINGTON strategic Planmng and Modeling Hewlett-Packard Company 3000 Hanover Street Palo Alto, California 94304
Late in the 1980s, Hewlett-Packard (HP) faced inventories mounting into the billions of dollars and alarming customer dissatisfaction with its order fulfillment process. HP produces computation and measurement products whose supply chains include manufacturing integrated circuits, board assembly, final assembly, and delivery to customers. To reduce inventory and improve order fulfillment, HP called on an internal team of in- dustrial engineers and management scientists augmented by academic collaboration. The team used an iterative process, enriched by the interaction of model development and applica- tion. HP reaped benefits well beyond its manufacturing operations, extending to diverse functions throughout the orga- nization. Similarly, the academic partners have infused their re- search with real-life experience. The supply-chain methodology is now mature, and HP is transferring the technology into the product divisions.
The Hewlett-Packard Company (HP) of knowledge and fundamentally improve was founded in 1939 by William the effectiveness of people and organiza-
Hewlett and David Packard, Today, the tions. In 1993, the company employed company's mission is to create information 96,200 people, 37,300 of them outside products that accelerate the advancement the US. Ciipynghl P 1995, Instilute for Operations Research INVENTORY/l'RODUCTfON—APPLICATIONS and the Manafiement Sciences INDUSTRIES—COMPUTtRS 1)1192-2102/q5/25O5/OO42SOI 2S This piifUT was ri'tereed
INTERFACES 25: 5 September-October 1995 (pp. 42-63)
HEWLETT-PACKARD
In the past six years, HP has faced spi- raling inventory and plummeting customer satisfaction. Solving the latter crucial prob- lem was flagged in 1990 by then CEO John Young as a key objective for the entire company, and one way of accomplishing the objective was through better supply- chain management.
A supply chain is a network of facilities that procure raw materials, transform them into intermediate goods and then final products, and deliver the products to cus- tomers through a distribution system. It spans procurement, manufacturing, and distribution. The importance of integrated management of the supply chain cannot be overstated. To fill orders efficiently, one must understand the linkages and inter- relationships of all the key elements of the supply chain. The supply chain for HP's products contains manufacturing, research and development (R&D) sites in 16 coun- tries, and sales and service offices in 110 countries. The total number of catalog products exceeds 22,000.
HP management has recognized that its performance filling orders will cause it to win or lose the competitive battle. In high technology industries, order fulfillment has become a major battlefield in the 1990s [Fuller, O'Connor, and Rawlinson 1993]. Current president and CEO Lew Platt identified successful order fulfillment as one of his top goals for fiscal 1993 and ap- pointed a vice-president to work full time toward that goal.
Grappling with supply-chain issues is a complex and multi-faceted task, especially in a company like HP whose products are diverse and complex. Variabilities and un- certainties can occur at any point along tbe
chain. Suppliers can be late in their ship- ments, or the incoming materials may be flawed. Tbe production process may break down, or the production yield may be im- perfect. Finally, product demands are also highly uncertain. It is imperative that man- ufacturing managers understand the effects
We discovered data-recording problems.
of uncertainties. Tbese effects are best un- derstood through analytical modeling. By using analytical techniques, managers can arm themselves with data that enable them to design manufacturing and distribution processes that provide efficient and cost- effective order fulfillment.
HP produces many products whose sup- ply chains include manufacturing and test- ing integrated circuits, board assemblies, and final assemblies, and distributing the products to customers through a global distribution network (Figure 1). Of HP s product groupings, computers and periph- erals provide the largest percentage of net revenue {77 percent) (Figure 2).
The approach HP used to address its supply-chain problems contained an ele- ment of serendipity and yielded a host of rewards. Recognizing the need for quanti- tative-based models to support top man- agement decision making, HP formed a group known as Strategic Planning and Modeling (SPaM) in 1988 and staffed it with industrial and computer systems engi- neers. HP charged the group with develop- ing and introducing innovations in man- agement science and industrial engineer- ing. By 1989, it had implemented a quantitative-based methodology, the
September-October 1995 43
LEE, BILLINGTON
Inventory Inventory
Figure 1: A supply chain includes all the interactions between suppliers of materials, manufac- turers, distributors, transportation links, and customers.
bubble model, to assess plant charters by capturing the costs at a site as a combina- tion of fixed and variable costs [Billington and Davis 1992].
While the bubble model was effective in capturing such costs as fixed overhead, equipment, transportation, and other vari-
computers & peripherals
77.0%
ables, it did not provide an accurate picture of the effect of uncertainties on inventory investment, customer service, and response time in filling orders. Given the importance of order fulfillment and the rising costs of inventory and logistics, SPaM set out to develop a new methodology that would
test & measurement 11.0%
60% medical
30% components
3.0% analytical
Figure 2: Of Hewlett-Packard's product groupings, computers and peripherals bring in the largest percentage of new revenue (1993).
INTERFACES 25:5 44
HEWLETT-PACKARD
address the aspect of uncertainty in the supply-chain problem. The quantification of the impact of uncertainties could then be integrated with the bubble model meth- odology to support strategic decision mak- ing. In 1989, we started working together tt) bring supply-chain-management con- cepts into HP.
Our collaborative efforts and those of our institutions resulted in an enduring co- operative relationship. Between 1989 and 1994, our partnership has grown, and it is now at the core of a series of joint research and application projects. Its evolution has been iterative, characterized by the interac- tion of model development and application experiences. We refined our methodology continuously as we accumulated experi- ence with business applications. The work, which began as a manufacturing initiative in 1989, has evolved to span several func-
tional areas: design and engineering, distri- bution, finance, and marketing (Figure 3). The evolution of supply-chain manage- ment at HP follows a dual path of meth- odology expansion and functional bound- ary expansion. Background and the Identification Phase
In the spring of 1989, SPaM began mod- eling the supply chain. The initial project was to investigate the supply-chain prob- lems of the persona! computer (PC) and deskjet printer divisions. The project was expected to take one year. We devoted the first six months of that project to identify- ing and characterizing the problem. We contacted primarily participants from ma- terials and manufacturing groups within the PC supply chain and spent hours talk- ing to production planners, members of the purchasing staff, production control engineers, and managers. We made many
Identification & Problem Definition
WINO Development
Inventory Reduction Projects
Supply-Chain Design
Design for Localization
Distribution Strategy Design
Design for Supply- Chain Management
Dealer Linkage
Forecasting & Planning
Industrial Forums
PC, desk|et
deskjet
deskjet, laserjet. Integrated circuit, InkJet component, workstation, computer, component, procurement, personai computer
MAXIM, network printer deskjet iow end computer
deskjet
European computer products
desklet, network printer, workstation
iaserjet, PC, network printer
iaserjet, InkJet component
HP Appie Stanford Stanford
89 90 91 92 93 94
Figure 3: Supply-chain model development parallels project engagements from 1989 lo 1994.
September-October 1995 45
LEE, BILLINGTON
site visits. At the sites, the project team came to a clear appreciation of the inter- relationships of all of the processes fron:i integrated circuit (IC) production, printed circuit board (PCB) assembly, PC final as- sembly, to distribution. The project team also collected massive amounts of data pertaining to manufacturing performance, demand characteristics, and order fulfill- ment performance. The PC supply chain provided the setting where HP developed
HP's peripheral products have been growing at a record pace.
an understanding of the interactions of dif- ferent entities in a supply chain, the com- mon pitfalls that operations managers of these entities faced, and the effects of performance measures on supply chains.
Our initial focus was to develop a clear understanding of the business problems faced by supply-chain managers. Having no preconceptions about what modeling work would follow, we were open to any and all ideas as to what we needed to model and what assumptions would even- tually be relevant. This explorative ap- proach served us so well on this first proj- ect that it has become one of the building blocks of SPaM's philosophy. Now in each project, models are developed only after gaining a thorough understanding of the problem. If one starts with a preconceived model structure before diving into the problem, one risks force-feeding an inap- propriate model to the problem or over- looking important elements of the problem that do not fit the model approach.
Supply-chain problems are so complex that even defining and shaping the investi-
gation is a challenge. The initial task within the PC division was to identify and understand the problem. The danger at this stage of a project is spending a great deal of time on something that turns out to be inconsequential. We obtained and ana- lyzed data that we eventually found to be of marginal significance. For example, we collected data related to manufacturing ef- ficiencies, machine breakdowns, change- over times, and so forth at a very micro- scopic, detailed level, and we collected characteristics of stock-keeping units with- out any grouping to form summary statis- tics. Nevertheless, these investigations are important, and, looking back, we can see that this initial phase set the foundation for SPaM's understanding of supply-chain problems. The relevance of the models we subsequently developed was enriched by our efforts to develop a deep understand- ing of the decision makers' environment.
The project's first data analysis was also a useful exercise. We learned what was available in HP's databases and how reli- able those data were. As we analyzed this data, we discovered data recording-prob- lems, for example, order data that were missing or obviously wrong. While this discovery was painful, the exercise itself was fruitful. It illustrated the importance of sound data recording and management and the value of consistent data definitions across different supply-chain models. We learned early on that reliable data is im- portant in formulating models.
We concluded the project with the PC division by identifying the common pitfalls operations managers fall into in the PC supply chain. Drawing on our knowledge from prior research with other companies
INTERFACES 25:5 46
HEWLETT-PACKARD
in a variety of industries, we summarized the pitfalls and outlined the opportunities they afford [Lee and Billington 1992]. An- other member of SPaM [Davis 1993] wrote general discussions of how supply chains can be more effectively managed.
Our effort to identify the problem was augmented by the results of an internal survey of HP's manufacturing managers. This survey, conducted by the manufactur- ing manager of the disk-memory division, identified the challenges and problem areas manufacturing managers faced. The survey results confirmed our findings that unreliable deliveries of materials by inter- nal suppliers greatly impeded efficient management of the supply chain.
As the PC project neared conclusion, SPaM began work with the Vancouver di- vision. That project called for modeling its supply chain to provide analytical support for manufacturing decision making. In our first meeting with the Vancouver manufac- turing management team, the manufactur- ing manager described four areas in which he wanted analytical modeling support: (1) Benchmarking inventory and service trade-offs, that is, identifying the efficient frontier of inventory-service trade-offs; (2) Assessing the impact of uncertainties on operational performance; (3) Analyzing what-if questions for differ- ent scenarios and operating characteristics; and (4) Evaluating product design impacts on his supply chain, that is, predicting how his supply chain would perform under dif- ferent product and process design alterna- tives.
This request led us to the next phase of developing a supply-chain model for HP
using Vancouver as the test site. The Worldwide Inventory Network Optimizer (WINO)
Early in 1990, we began to develop a model that would capture material fiows and the associated uncertainties of the Vancouver supply chain. In parallel with this modeling effort, SPaM started collect- ing detailed data at the Vancouver division concerning the deskjet product line. The Vancouver division manufactures printers based on the inkjet technology, sold under the deskjet label. First introduced in 1988, these printers won the 1988 Datek Printer of the Year Award, and sales grew to over 600,000 units in 1990 ($400 million). The division also manufactures other printer products, but their primary volume is desk- jet printers. Since its introduction, the desk- jet has been one of the fastest growing product lines at HP. The division has consis- tently maintained accurate data on supplier lead times, delivery performance, manu- facturing lead times, process downtimes, demands, and forecast errors, which facih- tated our early investigations. Even with Vancouver's efficient data management system and effective support, the task of gathering the information, analyzing it, and computing the summary statistics needed for the supply-chain model was still tedious.
We used a modeling approach based on developing a single-site inventory model to represent each operation in the supply chain and then integrated all the individ- ual site models to cover the complete sup- ply chain. At each site, we modeled two operations: receiving and production. A re- ceiving operation can be viewed as a pro- duction operation in which the "produc-
September-October 1995 47
LEE, BILLINGTON
tion" activities are receiving materials, in- specting incoming materials, and putting them into storage or sending them to the manufacturing line. Thus, one can model a single site as a combination of two sepa- rate inventory models, one for receiving and one for production. The sites are all linked in two ways. First, the demand for the end items forms the demand for the intermediate products, which in turn forms the demand for the raw materials, through the bill of materials. Second, the material or production delays of an upstream site result in delays in material receipt at a downstream site. These linkages are the cornerstone of the complete supply-chain model [Lee and Billington 1993], known as the Worldwide Inventory Network Opti- mizer (WINO).
WINO is basically a network of nodes in which we assume each node operates like a periodic-review, order-up-to-level inven- tory system. At this time, HP measured the ultimate customer service for final products in terms of fill rates, that is, the percentage of demands filled immediately. The model is therefore applicable to high-volume products for which immediate availability is essential to the customers (make-to-stock products). A model of this type is appro- priate for such HP products as persona! computers, printers, and other peripherals.
WINO relies on the following inputs: (1) Operation characteristics, such as spec- ification of the network structure, opera- tions performed at each node in the net- work, the bill of materials, the review pe- riod for each stock-keeping-unit (SKU), transportation times between nodes (means and variances), and desired service target or inventory levels at a node for
each SKU; (2) Supply characteristics, such as supplier lead times (means and variances), supplier delivery performance, and supplier quality (percent acceptable); (3) Process characteristics, such as produc- tion cycle and flow times (mean and vari- ances), production capacity, downtime characteristics, and yield; and (4) Demand characteristics, such as the mean and variance of demand for each final product.
WINO produces a variety of outputs: (1) Inventory in different forms (raw ma- terials, intermediate products, and final products), for different functions (as safety stock, cycle stock, and in transit), and in different locations; and (2) Customer service in terms of immedi- ate fill rate and response times.
The mathematical specification of WINO was only the beginning of the modeling ef- fort. It took SPaM much longer to program.
Management wanted to redesign the distribution strategy in Europe.
debug, and test WINO for efficient usage. The original WINO was programmed in C+ -I- running on a UNIX workstation, with a spreadsheet interface. We found the spreadsheet interface necessary for data in- put and result output because it allowed for easy statistical analyses. From 1991 on- wards, HP has run revisions of WINO runs on a PC. We validated WINO using the Vancouver supply-chain data [Lee, Billington, and Carter 1993]. We continue to test the effectiveness of the approxima- tion formulae and the robustness of the as-
INTERFACES 25:5 48
HEWLETT-PACKARD
sumptions in WINO, and we develop new service measure expressions through exten- sive simulation experiments [Johnson et al. 1995] and empirical testing. Inventory Benchmarking and Uncertainties
SPaMs project with Vancouver enabled the manufacturing manager to understand the inventory-service trade-offs for the supply chain and how well the division was performing relative to the inventory- service efficient frontier. In SPaM's subse- quent work on supply-chain projects this has been the fundamental first step in sup- ply-chain analysis.
The efficient frontier plots the minimum amount of inventory needed to meet a par- ticular service target, given the current de- mand, process, and supply characteristics
(Figure 4). Another basic analysis dealt with understanding the impact of uncer- tainties upon the performance of the sup- ply chain. For Vancouver, successive scenar- ios were tested. First, supplier delivery lead time uncertainties wore reduced to zero, that is, suppliers were assumed to be 100 percent reliable in their delivery perform- ance. Second, the uncertainties of both supplier deliveries and production pro- cesses were eliminated. Third, all uncer- tainties, including demand, were elimi- nated. Inventories decreased in each of these three cases (Figure 5). This Pareto analysis of uncertainties allowed Vancou- ver's manufacturing manager to gain a general idea of which source of uncertainty resulted in the highest cost impact to the division. The exact magnitudes of inven-
78 80 82 84 86 88 90 92 94
Customer Fill Rate (%)
96 98 100
Figure 4: The amount of inventory and customer service are measured by the percentage of customer orders filled immediately for the deskjet products over time. They are quite close to the inventory-service trade-off curve.
September-October 1995
LEE, BILLINGTON
Inventory in Supply Chain (Normalized) 100
20
Base Case No Supply
Uncertainty
No Supply & Process
Uncertainties
No Uncertainties
(Supply, Process, Demand)
Figure 5: Inventory can be reduced by successively eliminating the uncertainties in a supply chain.
tory reduction would depend on the se- quence in which the uncertainties were ehminated. Nevertheless this analysis helped managers to prioritize the invest- ments needed to improve the performance of the supply chain. We found that de- mand uncertainties had the greatest impact on inventory. This prompted the division to consider redesigning products and pro- cesses to lessen the impact of demand un- certainties on its supply-chain perform- ance.
From 1991 to 1994, SPaM did similar analyses for the laserjet printer division based in Boise, Idaho; the inkjet compo- nent division based in Corvalis, Oregon; the integrated circuit division based in Fort Collins, Colorado; and the computer man- ufacturing division based in Cupertino, California. The supply chains of all these divisions contain multiple sites inside and
outside the United States. In all cases, the basic analysis was effective in identifying opportunities for operational improve- ments. For example, in one case, the proj- ect team found that shortening review pe- riods was effective and important. In an- other case, the team discovered that safety stocks were being held at two locations (one HP and the other non-HP). In a third situation, the team observed products routed through many sites before comple- tion; the resulting long cycle times necessi- tated high inventories and resulted in poor response time to the customers. For an- other group, SPaM explored alternative transportation options and frequencies of shipments as means to improve the effi- ciency of the supply chain.
It became apparent during the early ap- plications of WINO that we could improve supply-chain performance by employing a
INTERFACES 25:5 50
HEWLETT-PACKARD
variety of fine tuning techniques, including more frequent reviews of inventory; bal- ancing inventories in component, subas- sembly, and finished product forms prop- erly; using different means of transporta- tion; and reducing manufacturing cycle times. Inventory reductions by 10 to 30 percent were identified. More important, SPaM found that, by using a more radical approach, it could make genuine break- through improvements. Three avenues offer this type of high payoff:
(1) Realignment of manufacturing and dis- tribution strategies;
(2) Improvement in forecasting processes and methods; and
(3) Product and process redesign for supply-chain management.
From Inventory Modeling to Manufacturing and Distribution Strategy
In a logical progression, SPaM and its partnering divisions have moved from ex- amining manufacturing processes to evalu- ating the methods HP uses to deliver the resulting products to customers. HP dis- tributes most of its personal computer products and peripherals through its own distribution network. This network consists of two major distribution centers (DCs) in the US, several in Europe, and one in the Asian-Pacific region. Manufacturing sites are located all over the world. HP's pe- ripheral products—laserjet printers, deskjet printers, and inkjet components—have been growing at a record pace. For exam- ple, the deskjet printer volume has grown from 600,000 units per year in 1990 to over 400,000 units a month today, an 800 percent increase. The inkjet components division has grown from a $200 million business a few years ago to over $1.2 bil-
lion today. Because product mixes are di- verse and the distribution channel in Eu- rope is complex, the cost of distribution is a large part of the product cost. Conse- quently, management wanted to redesign the distribution strategy in Europe, inven- tories can be targeted to serve clusters of customers. Savings can be realized by opti- mizing freight and delivery mechanisms.
HP has considered several factors in de- veloping a strategy for distributing periph- eral products in Europe: (1) Many products are shipped from a sin- gle European DC to customers on the con- tinent. However, some products (both pe- ripherals and computer products) can be distributed from multiple DCs in France, Germany, and the Netherlands.
(2) HP distributed some products, such as laserjet printers, out of two different DCs. One location specialized in bulk packed, pallet movements for customers with large order volumes, while the other location served customers with small order vol- umes. While it gained some economies through this specialization, HP lost the benefits of pooling safety stocks and in- curred additional overhead expenses and other fixed costs.
In designing a new European distribu- tion network, SPaM used a combination of cost modeling (the bubble model) and sup- ply-chain-inventory model (WINO) meth- odologies. For a specific distribution net- work strategy, it used the bubble model to evaluate the costs of transportation, cus- toms, duties, fixed and variable processing, and implementing changes to the existing network. It used WINO, on the other hand, to analyze the trade-offs between in- ventory investment and customer service
September-October 1995 51
LEE, BILLINGTON
and to augment the bubble model with a more accurate estimate of inventory carry- ing costs.
HP's general manager of worldwide dis- tribution initiated the European distribu- tion network strategy project. The project team was led by a member of the world- wide distribution organization and con-
Supply-chain analysis is much more than inventory modeling.
sisted of representatives from the Euro- pean DCs and from the key product divi- sions that supplied them: deskjet, laserjet, and inkjet components. Ft)r two months the team worked intensively, collecting and validating data, and debating and evaluating alternative strategies. Team members traveled to all the sites, since they could obtain some information only by talking directly to people at the sites, in early 1992, the executive vice-president of the computer products operation used the data from this investigation to make the decisions that resulted in the realignment of the distribution network. This realign- ment reduced the total distribution cost in Europe by $18 million a year.
The European distribution netwt)rk strat- egy project was very successful and vali- dated many of our beliefs. First, it quanti- fied many cost elements. Historically, man- agement had often made such decisions subjectively or intuitively. Second, we vali- dated the importance of integrating the supply-chain-inventory (WINO) and fun- damental cost models. Third, this effort convinced us that the composition of a project team greatly affects management
receptivity and the implementation success of modeling projects. Because the project team comprised representatives from all the sites concerned, it obtained the appro- priate data and information, assessed and interpreted the analyses properly, and vali- dated the output. Finally, it is clear to us that supply-chain analysis is much more than inventory modeling. It can be ex- tended to distribution strategy analysis and to other types of supply-chain problems.
As the European market grew, I IP con- sidered expanding manufacturing capacity on the ctintinent. In particular, deskjet printers experienced explosive growth. In 1992, HP's deskjet orders doubled in Eu- rope in a single month! The need for a Eu- ropean plant became apparent. Yet HP needed to consider its manufacturing and distribution strategies jointly and coordi- nate them. Management asked SPaM to evaluate these strategies for the deskjet ex- pansion into Europe.
In reality, HP s manufacturing activities were often not completely decentralized. For example, boards were often assembled in one location to serve all of the compa- ny's requirements worldwide. Conse- quently, we could not isolate the European manufacturing strategy from that of the rest of the world. Therefore, when SPaM embarked on the European manufacturing and distribution strategy, it had to model the worldwide deskjet supply chain. Even though our focus was the European strate- gies, we analyzed the worldwide manufac- turing and distribution strategies of the deskjet printer division. With global data, HP could align the European strategy with the worldwide strategy and avoid less ef- fective strategies.
INTERFACES 25:5 52
HEWLETT-PACKARD
1 IP has relied on supply-chain modeling in introducing new products as well. In late 1990, SPaM worked with a product development team from Vancouver to ana- lyze alternative manufacturing strategies for J new product that involved a Japanese partner. The project team discovered that the expected cost-service efficient frontier was much worse than that for Vancouver's existing products. Declaring this unaccept- able, the team explored alternative supply- chain networks based on different manu- facturing locations fur both HP and the Japanese partner. As a result of the anah- sis, both companies decided to redeploy their existing facilities so that they could produce this new product with optimal ef- ficiency. Performing this kind of analysis before the new product was launched was extremely valuable [Lee and Billington ]993|.
SPaM made a similar evaluation in 1991 when HP introduced the network printer in Europe. We assessed alternative strate- gies that looked at relocating various HP facilities in Europe and the US. The supply chain uf the network printer involved an Asian partner that manufactured the key engine. The engine was integrated with HP's printed circuit board tu form the end product. The locations of this Asian part- ner's factory, HP's board factory, and the DCs cuuld affect the costs and inventory performance of the supply chain. HP's management used the results of our inves- tigation during its negotiations with its Asian partner to determine the manufac- turing and distribution plan fur the new product. Design for Supply-Chain Management
The concept uf supply-chain manage-
ment had begun tu gruw as projects prolif- erated throughout the company. In the early supply-chain projects that dealt with inventory and service trade-offs, we treated the bill tif materials and the pru- ductiun prucess as knuwn a priuri. We fu- cused on huw to improve the performance of the chain. Our involvement with a number uf divisions led to the emergence t)f a concept known at IIP as "design for supply-chain management." The key puint of this concept is that the design uf prod- ucts and the design uf the assuciated pro- ductiun processes can produce big payoffs.
The Vancouver division provided the first instance of such payoffs. The division used tu make all models of their printers in the factories and then ship them tu the DCs fur distribution. The uncertainties of different countries' demands had resulted in inventory imbalances: the DCs had ex- cess inventuries uf some printers, and shortages of others. Since printers are shipped by sea fruni the factories tu the uverseas DCs, the DCs had trouble re- sponding quickly to variations in customer demands. Despite a large investment in in- ventory, they could nut provide good cus- tomer service.
SPaM used WINO tu analyze what pay- offs were possible from redesigning the product. The project team quantified the savings that HP could realize if the remote DC performed the final localization step (customized the printer for the local cuun- try) instead of the factories. This concept is known as "design for localization. " A change as far reaching as this required the cullaburation of all of the division's major functions: engineering, manufacturing, and distribution. Engineering had tu redesign
September-October 1995 53
LEE, BILLINGTON
the product so that the power supply mod- ule could be plugged in externally instead of being internal to the product. Buy-in from manufacturing was imperative be- cause the tasks performed by the factories and DCs would change. Distribution also had to support the change because it would now be responsible fur procuring parts, power supply modules, and man- uals, performing final localization opera- tions, and ensuring quality, functions that had not been part uf its normal routine. In addition, we expected that these activities would be more expensive when performed at the DC instead of at the factories. How- ever, the Vancouver division was consider- ing the entire system, nut just isolated ac- tivities.
For this project, we used WINO to assess the risk-pooling benefits in the form of in- ventory savings that would result from Vancouver making one model of the printer, instead uf multiple country ver- sions. On the other hand, with DC-local- ization, the localization materials would need tu be stocked at all DCs, instead uf in one central location. Hence, in terms of in- ventories of localization materials, we have a 'reverse risk-pooling" phenomenon, and WINO was used to quantify those differ- ences as well.
This project gave SPaM its first taste of the impact of organizational issues in sup- ply-chain management. To convince all three groups to work together un this rede- sign, the project team had to quantify the costs and benefits of the change.
The Vancouver division implemented the DC-localization strategy successfully, improving both cost and service. It is now fully committed to the design-for-localiza-
tion concept and currently designs all its new products to be localized at the DCs [Lee, Billington, and Carter 1993].
This project convinced us that supply- chain performance can be greatly im- proved by concurrently redesigning a product and its pruductiun prucess. Such traditional design concepts as design for manufacturability are inadequate to cap- ture these logistics and distribution aspects. At the conclusion of this project, we began working on other projects where product designs can affect supply-chain perform- ance. Localization is only one of many op- tions within the design-for-supply-chain- management concept. Another is the use of common parts so as to delay the differ- entiation uf products during manufactur- ing. This strategy of delaying product dif- ferentiation has gained momentum at HP. We have interviewed individuals in a number of divisions who have changed designs to gain flexibility, and as a result, we have developed a mure formal struc- ture of the design-fur-supply-chain-man- agement concept [Lee 1993].
SPaM's work in designing for supply- chain management also included strategies that involved strategic partners, expanding the scope of its work over company boundaries to include key suppliers. We have discovered tremendous economic po- tential. In a 1993 project to introduce a pe- ripheral product, we found we could re- duce the unit cost of the product (based on the traditional supply-chain structure) by as much as $80 by appropriately aligning the supply-chain structures of the strategic partner. Peripherals have thinner margins than complex computers and instruments, and this reduction in unit cost goes directly
INTERFACES 25:5 54
HEWLETT-PACKARD
to the buttum line. Postponement
In general, the production of multiple versions of a product (for different geo- graphical markets or different user seg- ments) generally begins/can begin/could begin with a single engine. Throughout the production process, different subassemblies are added, leading to multiple versions uf the end product. By delaying product dif- ferentiation one delays as much as possible that moment when different product ver- sions assume their unique identities, thereby gaining the greatest possible flexi- bility tu changing customer demands. Flex- ibility can improve the cost-effectiveness of the supply chain, because inventories are stocked in the predifferentiation furms. At HP, the concept of delaying product differ- entiation is knuwn as "postponement" [Lee and Billington 1994].
The Vancouver division's printer family proliferated as the division tried to satisfy many different customer segments, each requiring different functionalities. We adapted the basic supply-chain model to analyze the potential benefits of using al- ternative process designs so that the divi- sion could delay product differentiation [Lee forthcoming. Lee and Tang 1994].
Increasingly SPaM is asked tu work on delayed-product-differentiation projects. Another printer division was a recent can- didate. Traditionally, printers destined for the European market use dedicated en- gines with 220V power supplies and those for the North American markets use HOV power supplies. Reworking the engines to switch the power supply is nut a trivial task and would incur some significant la- bor and material costs. Using such a design
for printers would nut provide HP with the flexibility to use the printer inventory to satisfy both markets. Imbalances in the inventuries uf the two types of printers could cause serious problems during two stages uf the product life cycle: early intro- ductiun and end uf life. In 1993, SPaM an- alyzed the cost implications of switching to a design that specifies engines with a uni- versal power supply. We examined such costs as materials, inventory, rework, shortages, and transshipments from one region to another tu correct imbalances.
The costs are quite dependent on the lead times frum the supplier for the en- gines with dedicated and universal power supplies. With dedicated puwer supply, product differentiation for the two markets effectively starts at the beginning. With universal power supply, differentiation dues not occur until the engines are trans- ported to the two continents. The value of this postponement, in the form of inven- tory reduction, can be quantified [Lee and Sasser 1995]. With product life cycles that are becoming shorter and shorter, the ben- efit of postponement with universal power supply outweighs the additional material cust. Expanding into Distribution Channel
SPaM's supply-chain projects at HP have progressed from considering material management to considering manufacturing and ultimately distribution partners. We have expanded their scope across divi- sional and functional boundaries and be- yond the corporate structure, We now in- clude suppliers un the upstream side and, on the downstream side, HP's salesforce and dealers. In expanding the distribution channels, we cross a frontier, because HP
September-October 1995 55
LEE, BILLINGTON
does not usually own these agents. They are, however, an integral part of the over- all supply chain of 111̂ products. HP's ex- panded view of its supply chain has led it to try to create win-win situations for HP and its dealers [Mullich 1993]. HP is now much more responsive to the requirements of its distribution channels and tnore flexi- ble in satisfying their changing needs.
The interdependences between HP dealers and }1P were the subject of a study early in 1991. We began with a survey of dealer needs. We sent out teams to a num- ber of major dealers in the US. The divi- sional marketing managers coordinated these visits. This was marketing's first in- volvement in the supply-chain efforts. During these visits, the teams tried to un- derstand the dealers' operational environ- ments, their crucial needs from HP, their past experience with HP, and the impr(.>ve- ments HP could make to enhance their competitiveness. HP manufacturing man- agers used this survey as a springboard for greater involvement with their down- stream external customers as supply-chain partners. The survey provided HP with factual information and also generated a spirit of collaboration with the dealers.
We found, first, that tbe dealers are much more concerned about the reliability of delivery than the length of delivery lead times. Second, dealers depend upon hav- ing accurate and timely information on de- livery dates, order status, revisions of de- livery dates, and product availabilities.
The make-to-order environment has im- portant implications. In make-to-stock sys- tems, customers expect products to be available on the shelf. In make-to-order systems, customers expect products to ar-
rive within a delivery time window, known as the promised lead time. When a customer requests a product and expects the order to be filled within a time win- dow, the order reliability is the probability that the order will be filled within this tar- get lead time. At HP, some computer prod- ucts—personal computers, computer pe- ripherals, and consumable units like toners and inkjet cartridges—operate like a make- to-stock system, because immediate avail-
Immersing themselves in an industrial setting is recommended to all academics.
ability is a critical competitive factor. How- ever, other products, such as minicomput- ers, workstations, medical instruments, and other test and measurement products, op- erate like make-to-order systems. Because of the needs of these other product divi- sions, HP eventually extended WINO to handle make-to-order situations.
In 1991, HP started a project to extend a division s supply chain to include the key dealers' inventory systems. The project re- vealed the existence of duplicate safety stocks throughout the supply chain. Fin- ished products were stocked at both HP DCs and the dealers. HP used WINO to quantify how much inventory could be re- duced at these two locations, while main- taining the same service target for the cus- tomers. Quantifying this duplication was useful to management.
HP and the dealers have worked closely to improve the performance of the overall supply chain [Mullich 1993]. In another
INTERFACES 25:5 56
HEWLETT-PACKARD
initiative, HP has collabtjrated with its dealers in applying the postponement strategy. Possible postponement strategies range from dealers performing only distri- bution functions, to performing distribu- tion plus some localization operations, to performing distribution, localization, and customization tasks. Postponement strate- gies involving dealers must be carried out in close collaboration with them. We can- not emphasize too strongly the importance of interactions and empirical studies. To assess the capabilities of dealers to perform some of these operations and to develop operational relationships, HP visited a number of US and European dealers. Be- fore modeling the situation in which deal- ers customized and configured end prod- ucts, HP had to find out if they were equipped to perform these operations, if they were willing to do so, and if they had adequate incentives. Managers considering postponement strategies should conduct detailed analyses before making such deci- sitins- Tbese analyses should include visit- ing and negotiating with dealers, quantify- ing the costs and benefits, assessing the marketing implications, and considering government regulations and local content laws, environmental requirements, and or- ganizational impacts. Engaging Marketing and Finance
As SPaM s supply-chain-modeling work progressed, a shift took place within HP. Some heretofore isolated and internally fo- cused groups began to move to a broader perspective and started to work together to improve the performance of tbeir supply chains. SPaM saw tbat beyond engineer- ing, manufacturing, and distribution, the next functional area it needed to involve
was marketing. A number of projects have revealed that demand uncertainty resulting from forecast errors was the key source o( inefficiency in a supply chain. Design changes, sucb as common parts, delayed product differentiation, and other post- ponement strategies, helped lessen the im- pact of forecast errors. Nevertheless, SPaM felt that it should address the fundamental problem of forecasting processes and methods.
In 1992, HP s Boise Printer Division formed a major team to tackle this prob- lem. The project team members came from SPaM and from tbe divisional marketing and material departments. At the outset of this project, SPaM was convinced that forecasting was much more than a statisti- cal tool for projecting demand into the fu- ture. It also embodied the analysis of the impact of forecast errors on the efficiency of the supply chains. We believed that a forecasting project should also include tbe process in which the firm used forecasts to generate production, inventory, capacity, and financial plans. As a result, the project team was expanded to include representa- tives from production planning, finance, product marketing, process engineering, and materials management. In fact, the project name FLAP (forecasting, logistics, and planning) indicates the expanded scope of this project.
The FLAP project unveiled interesting information, such as the importance of in- stalling a measurement system to track forecast accuracy, the importance of care- fully and completely documenting tbe forecast processes that translate basic cus- tomer needs into production requirements, and the need for forecasters and planners
September-October 1995 57
LEE, BILLINGTON
to communicate in a timely and accurate manner.
Another extension to our supply-chain modeling considers the financial implica- tions of a worldwide supply chain. As tbe supply chains for most of HP's products are global in nature, HP must evaluate cost and revenue implications in light of fluctu- ations in exchange rates. We need to ac- count for currency exchanges in the global supply chain, value-added activities in the global manufacturing and distribution net- work, and tbe revenues generated from sales in a worldwide market. Financial hedging strategies and tbeir associated costs can affect where HP sbould locate plants, how it should deploy its resources, and how it should add capacity. We must model exchange rate uncertainties explic- itly, or perform extensive sensitivity analy- ses. In recent projects wbich dealt witb supply-chain designs, the divisions in- volved dedicated a financial analyst to the project, along with tbe controller of tbe di- vision wbo oversees the development of tbe project, and members of tbe manufac- turing staff.
With marketing, finance, manufacturing, engineering, and distribution all involved in its projects and with recent supply-chain modeling developments, SPaM has entered a mature phase of supply-chain work at HP. SPaM and Academic Collaboration
It is worthwhile to describe the events that led to the working relationship of the two authors. In 1989, Hau Lee took a sab- batical leave from Stanford and expressed an interest in using that time to work on industrial projects. Prior to that, he had been working on supply-chain-related
problems with Booz-Allen-Hamilton, IBM, and Apple Computer [Cohen and Lee 1988, 1989]. Integrating an academic part- ner into a corporate researcb and develop- tnent effort at HP required both skillful po- sitioning and higb-level champions. To bring him into HP, Corey Biilington had to convince the executives of the value of sucb an undertaking. He approached two vice-presidents: Hal Edmonson, then vice- president of corporate manufacturing, and Lew Platt, then executive vice-president of computer peripherals, and tbey approved of the plan. In addition, the team secured the support of the manufacturing manag- ers of the divisions that produce personal computers and deskjet printers as pilot project sites. In tbis way, we launched a bighly successful, ongoing collaboration.
SPaM values its close ties with academic research and teaching. Hau Lee spent bis sabbatical year at HP in 1989. Initially, we expected to introduce rationalized supply- chain inventory models into HP during that year. Over the last five years, our rela- tionship bas continued, and its scope bas expanded. This relationship is collabora- tive, not consultative. Hau Lee is a full participant on such issues as organizational design and assists in technology develop- ment for the group. On a reciprocal basis, tbe members of SPaM have availed them- selves of his expertise. As a result, tbe amount of learning that bas taken place witbin the group has been tremendous.
Academia has benefited from this rela- tionship by virtue of the mandate for rele- vancy witbin a working corporate environ- ment. Priorities must be clearly defined, and that very ranking, in turn, helps to identify industrial trends tbat can yield
INTERFACES 25:5 58
HEWLETT-PACKARD
topics for future research. Immersing themselves in an industrial setting is highly recommended to all academics who can ar- range such an opportunity.
SPaM's work in supply-chain manage- ment has also contributed to the curricula at Stanford University. One of the projects witb the Vancouver division bas been de- veloped into a teaching case for classroom discussion [Kopczak and Lee 1994], and tbe case is now being used at supply-chain training workshops for HP employees.
SPaM bas continued to host academic researchers who fit SPaM's research needs. In 1992, when SPaM began its forecasting work, it enlisted a statistician from tbe University of Texas to spend half of her sabbatical year with SPaM to develop fore- casting methodologies. In the same year, SPaM brougbt in a professor from the Graduate School of Business at the Univer- sity of Chicago for six montbs to explore financial hedging problems. In the summer of 1993, SPaM hosted a professor from the Graduate School of Business at Vanderbilt University to help the group with the sim- ulation testing of some key assumptions and approximation formulae of WINO. In 1994, SPaM began working with a faculty member from the Graduate Scbool of Management at UCLA on the impact of product-line structuring on supply-chain performance.
SPaM contributes to the academic com- munity through its strong ties with specific research programs. For example, HP is an industrial partner in the Stanford Inte- grated Manufacturing Association (SIMA), a consortium of industrial corporations that are partners to the teaching and researcb programs on manufacturing at Stanford.
HP has suggested areas of work and of- fered direction to SIMA researchers. As a result, SIMA has initiated projects that are supply-chain related. These projects have received strong support within the com- pany in terms of advisory assistance from HP managers. Just as HP bas begun ex- panding the involvement of multi-func- tional teams in supply-chain modeling ef- forts, Stanford has expanded its involve- ment in supply-chain research. Professors from industrial engineering, the design di- vision of mechanical engineering, and the school of business who specialize in opera- tions, incentives, and information systems have been involved in the SIMA supply- chain projects.
In the last few years, other divisions within HP have also funded Stanford doc- toral students doing research on supply- chain management. These doctoral stu- dents worked as summer interns at HP and have used the problems they studied to motivate their thesis research. Industrial Interactions
We are aware that supply-chain prob- lems are not unique to HP. An increasing number of reports and articles describe how other companies are wrestling with the same issues [Interactive Information Services 1994]. To improve the methodolo- gies that SPaM has developed, we wanted a broader perspective than that available solely at HP. In March 1991, HP hosted an industrial forum on supply-chain manage- ment, not a public conference, but a gath- ering of SIMA industrial partners inter- ested in research on supply-chain manage- ment. The forum's objectives were to share common problems and best practices, and to generate ideas and directions for the
September-October 1995 59
LEE, BILLINGTON
Stanford research team. Hal Edmonson, former vice-president of corporate manu- facturing at HP, was very supportive. He presented the opening address in which he said that, \ilthough we are fierce competi- tors, the sharing and interactions at this forum will make us better competitors." The same idea was echoed by Bill Hansen, tlie former vice-president a( manufacturing and logistics at Digital Equipment Corpora- tion. The participants at the first forum in- cluded HP, Apple Computer, Digital, Ford Motor Company, and Stanford University. Each participating company was required to give a presentation on the challenges and problems it faced, the methods or ap- proaches it used, or the lessons learned from implementation experiences. The forum was a great success. Representatives from all the participating companies felt that the forum provided a very useful ex- change and agreed that there should be a follow-on forum.
For HP, the forum was a means to con- duct a first-level benchmarking on indus- trial practices in supply-chain manage- ment. It also provided a means to build up industrial relationships and networking. There are always some areas in which HP employees can learn from its peers.
Apple Computer hosted the next indus- trial forum in November 1991. In addition to the first year participants, the 1991 at- tendees included IBM, Xerox, AT&T Bell Labs, Raychem, and SUN Microsystems. Returning participants reported on their progress with the supply-chain issues they had identified in the preceding forum. For example, Apple Computer had gone through the problem definition phase, and it reported on the redesign of its distribu-
tion strategies. Digital Equipment Corpora- tion described how it had integrated sup- ply-chain management with its learning organization concept. IIP shared with the participants its first success in design for localization.
SIMA hosted the third industrial forum, in April of 1993, focusing on design for supply-chain management. New partici- pants included General Motors, General Electric, Seagate Technologies, Quantum, Silicon Graphics Inc., Lockheed, and Ad- vanced Micro Devices. We began to see more companies working on the design and organizational aspects of products and processes to improve their supply-chain performance. At two separate panel dis- cussions, participants reiterated the impor- tance t)f multi-disciplinary research and coordination. Moreover, they suggested that supply-chain research has not yet fo- cused on the accounting and financial as- pects of supply chains.
The 1994 forum was again hosted by Stanford. This time participants came from industries beyond the high technology area, such as pharmaceuticals {Eli Lilly), distribution (McKesson), fast food (Jack-in- the-Box), general {Siemens, Texas Instru- ments), and consulting {Andersen Consult- ing, PRTM), Companies began sharing information on how they carried out supply-chain education internally. Current and Next Steps
SPaM plans to ctmtinue its supply-chain efforts in the future. One major goal is to transfer the analytical supply-chain meth- ods to HP divisions, it has initiated several steps toward this end already. They in- clude enhancing the analytical tools and educating users.
INTERFACES 25:5 60
HEWLETT-PACKARD
We originally developed WINO as an in- ternal tool for SPaM analysts tt) use in their consulting work at HP. We designed the user interface for use by highly trained analysts. The divisions in Computer Prod- ucts Operations need a tool to aid material planners with inventory and service man- agement. SPaM engaged professional sys- tems programmers to rewrite WINO into a more powerful tool, named the Supply- Chain-Analysis Tool (SCAT). This tool en- hances WINO in three ways. It has a better graphical interface, supports more help functions, and is more user-friendly. SCAT was implemented on a PC. We have com- pleted pilot projects using SCAT, and so far, the divisions' reaction has been very positive. We hope that SCAT will make analytical supply-chain analysis more ac- cessible to HP divisions.
In 1994, SPaM launched a major initia- tive to develop an education program for HP managers and engineers on supply- chain management. It developed the mate- rial for the program in collaboration with the corporate education group. We intend to design and offer a workshop that high- lights the importance of supply-chain management, teaches the basic methods involved, and illustrates some case exam- ples. Formalized education should hasten the diffusion of supply-chain competence throughout the company. In 1994, SPaM conducted close to two dozen workshops. The demand for such management educa- tion programs has continued to grow, and we have plans for many more workshops in the future.
Recognizing the importance of supply- chain management, several key divisions at HP have now formalized such positions
as supply-chain project managers, supply- chain analysts, or supply-chain coordina- tors. These individuals work closely with SPaM to ensure that supply-chain models are used effectively at the divisions and to identify new problem domains that feed SPaM's research and development efforts. Reflections
Supply-chain modeling has entered its sixth year at SPaM. Over the years, the SPaM projects have produced tremendous value for HP. SPaM has often identifieci cost savings of 10 to 40 million dollars per year from a single project. SPaM's role is in helping the divisions to identify the problems, to structure the decision process, and to provide support in management de- cision making related to supply chains. The accountability, decisions, and the final implementation of the decisions still rest with the divisions. While SPaM's projects have resulted in great dollar savings, an- other measure of SPaM's success is the in- creasing demands for SPaM's service. In 1988, the group size was only two. In 1994, the group, including part-time con- tractors, had risen to about a dozen. Even with this growth, the demand for SPaM's service continues to exceed what the group can handle.
We've learned a great deal over the past six years: {]) Supply-chain management is a busi- ness fundamental. As SPaM examined many of its customers, it often found that very basic changes in operations (for ex- ample, installing a performance measure- ment system) woulci improve supply-chain performance. Sophisticated inventory models were not always required. Identifi- cation of such problems and developing
September-October 1995 61
LEE, BILLINGTON
Supply-Chain Problems
Material Manu- Distri- facturing bution
R&D Marketing Finance
Inventory/Service Improvement
Operational Efficiency
Chain Structure, Distr. Strategy
Distribution Channel
Global Financial Hedging
Forecasting & Planning
Design for Supply Chain
X
X
X
X
X
X
X
X
X
X
X
X
X
X X
X
X
X
X
X
Figure 6: Supply-chain development has been tied lo functional boundary expansion, with increasing involvement and collaboration of cross-functional areas.
plans to address them are the first steps to- wards effective supply-chain management. (2) Interactions with the academic and in- dustrial communities are of tremendous benefit. In particular, having a professor spend a sabbatical or internship with a company is a valuable means of technol- ogy transfer and mutual education. (3) It is very important that research and model developments do not happen in a vacuum. All our model developments hap- pened while SPaM was working with HP divisions on real projects. Real projects and interactions with managers and relevant decision makers are the motivating forces that guide SPaM's development work. (4) Supply chain is not just a problem for manufacturing or for distribution. All func- tional areas of the company must even-
tually be included, and these areas must participate in improvement projects (Figure 6). In a parallel fashion, research at univer- sities on supply-chain management cannot continue as a problem just for operations management groups but instead must be tackled by multi-functional teams. (5) Without sound analysis and quantita- tive data support, management may settle for the status quo, and engineers may avoid making substantive engineering changes. Our analytical work has provided different functional teams with bases for their negotiating and deliberating on strat- egies. They have helped management see the costs and benefits of implementing changes in their policies. (6) Great payoffs can be achieved by tak- ing an expanded view of the supply chain
INTERFACES 25:5 62
HEWLETT-PACKARD
and, in particular, including the key sup-
pliers and customers in supply-chain
decisions.
Acknowledgments
Many people contributed to the develop-
ment and application of supply-chain
management models at Hewlett-Packard.
We especially acknowledge the contribu-
tions of David Archambault, Everett Baily,
Brent Carter, Tom Davis, Craig Fix, Debbie
Flandro, Allan Gross, Robert Hall, Eric
Johnson, Jeff Larsen, Steve Rockhold,
Marguerita Sasser, Don Schmickrath, and
Chuck Walter. We also thank the editors of
the special issue for their helpful com-
ments, which greatly improved the
exposition of the paper.
References Billington, C. A. and Davis, T. C. 1992, "Manu-
facturing strategy analysis: Models and prac- tice," OMEGA, Vol. 20, No. 5/6, pp. 587- 595.
Cohen, M. A. and Lee, H. L. 1988, "Strategic analysis uf integrated production-distributiun systems: Models and methods," Operations Ren-arch, Vol. 36, No- 2, pp. 216-228.
Cohen, M. A. and Lee, H. L. 1989, "Resource deployment analysis of global manufacturing and distribution network," journal of Manu- facturing ami Operations Management, Vol. 2, No 2, pp. 81-104.
Davis, T. 1993, "Effective supply chain man- agement," Sloan Management Review, Vol. 34, No 4 (Summer), pp. 35-^46.
Fuller. J, B,; O'Connor, J.; and Rawlinson, R. 1993, "Tailored logistics: The next advan- tage," Harvard Bnsinesf Review, Vol. 71, No. 1 (May-June), pp. 87-98.
Interactive Information Services 1994, Research Report for Logistics ivu1 Manufacturing: Appli- cations for Managing the Snpphj Chain, IN- FACT Research Limited, London.
Johnson, M, E.; Lee, H. L.; Davis, T.; and Hall, R. 1995, "Expressions for item fill rates in pe- riodic inventory systems," Naval Research Lo- gistics, Vol. 42, No. 1, pp. 57-80.
Kopczak, L. and Lee, H. L 1994, "Hewlett-
Packard's deskjet supply chain," Teaching Case (A) and (B), Stanford University, Stan- ford, California.
Lee. H. L. 1993, "Design for supply chain man- agement: Concepts and examples," in Per- spectives in Operations Management: Essay/s in Honor of Elwood S. Buffa. ed. R. Sarin, Kluwer Academic Publishers, Boston, Massachusetts, pp. 45-65.
Lee, H. L. forthcoming, 'Effective management of inventory and service through product and process redesign," Operations Research.
Lee, H. L. and Billington, C. 1992, "Managing supply chain inventory: Pitfalls and opportu- nities," SlocDi Management Review, Vol. 33, No. 3, pp. 65-73.
Lee, H. L. and Billington, C. 1993, "Material management in decentralized supply chains," Operations Research. Vol, 41, No. 5, pp. 835- 847.
Lee, H, L. and Billington, C. 1994, "Designing products and processes for postponement," in Majnigenicnt of Design: E)igi)icering ami Man- agement Perspectives, eds. S. Dasu and C. Eastman, Kluwer Academic Publishers, Bos- ton, Massachusetts, pp. 105-122.
Lee, H. L; Billington, C ; and Carter, B. 1993, "Hewlett-Packard gains control of inventory and service through design for localization," Interfaces, Vol. 23, No. 4, pp. 1-11.
Lee, H. L. and Sasser. M, 1995, "Product uni- versality and design for supply chain man- agement," Special issue on supply chain management, Interiiational jonrnal of Produc- tion Plan}ii)ig and Control, Vol. 6, No. 3, pp. 270-277.
Lee, H. L. and Tang, C, S. 1994, "Models to evaluate the costs and benefits of delayed product differentiation," working paper. De- partment of Industrial Engineering and Engi- neering Management, Stanford University, Stanford, California.
Mullich, J. 1993, "HP conquers channels with speed," Business Marketing, Vol. 78, No. 7 (July), pp, 44.
September-October 1995 63