...production factors such labor costs, logistics ease and efficiency, and market presence. Multiple plants generally offer the benefits of an increased responsiveness, ease in local customization and logistical gains afforded by simplification of the distribution network. However, accompanying these benefits is the increased managerial complexity associated with planning and operating multiple units.

In this paper we address the problem of deciding the configuration of multiple plants in a manufacturing network when products have component commonality. We compare two broad choices: (i) a product network configuration; and (ii) a process network configuration. In a process network configuration, a firm chooses to consolidate common component production into a single plant. This allows the firms to risk pool the uncertain demand for the component from all sources (products and markets). Hayes and Wheelwright (1984) suggest that the process network configuration would have less duplication of equipment for producing common parts. However, the firm incurs the incremental logistics cost now associated with shipping common components to other "final" assembly plants. A firm may alternatively choose a product network configuration where each plant within the network is assigned the responsibility of manufacturing and assembling the final product under one roof, thereby spreading common component production over multiple plants.

This paper analyzes the trade-offs incurred in making such choices. How would a company decide on configuring its multi-plant network when there are two or three, key but disparate technologies/operations involved? How does a company decide on the configuration of its subassembly and assembly plants? What are the operational trade-offs? How do demand uncertainty and the presence of a common component affect the configuration choice? Are there specific advantages to risk pooling in the presence of demand uncertainty and component/subassembly commonality in the bill-of-materials structure? We aim to answer these questions in this paper. In the remainder of the manuscript, the terms "process-focused network," "process plant network," "process configuration," "process plant" and "process network" are used interchangeably. Similarly the terms "product-focused network," "product plant network," "product configuration," "product plant" and "product network" are used interchangeably.

2. Literature review

The importance of the multi-plant configuration problem has been long recognized in operations strategy. Schmenner (1982) identifies four distinct multi-plant configuration strategies prevailing among Fortune 500 companies in the United States. These are:

* Product plant strategy: each plant produces a product in its entirety for distribution anywhere.

* Market area plant strategy: plants are focused according to geographic regions and each plant produces almost the entire line of products for distribution to markets within a geographic zone.

* Process plant strategy: plants are focused according to a specific process and take advantage of scale economies.

* General-purpose plant strategy: in this case, plants are flexible and are capable of being assigned any charter from within the firm's existing product line and structure.

Cohen and Lee (1989) have extended Schmenner's work and formulated several plant charter strategies in global manufacturing networks. One of the focal points of these strategies is on specialization by geographic region which may involve among other things, agglomeration economies. Although we do not attempt to address such issues in this paper the overall framework outlined in Cohen and Lee can serve as a motivator for designing more complex network configuration models.

Specific attempts have been made to resolve decisions such as, which plant makes what and how much capacity to acquire at each plant. For example, Moon (1989) and Cohen and Moon (1991) develop single-period, deterministic mixed-integer programming models for plant product mix loading and develop policies for the optimal configuration of a set of capacitated plants. Cohen and Moon (1990) investigate the impact of scale economies, scope economies and supply chain transportation costs on network configuration and distribution policy with a mixed-integer nonlinear programming model and develop insights on the impact of changes in a firm's costs on its supply chain structure. Li and Tirupati (1995) and Li and Qui (1996) examine investment-planning problems, and determine the choice of technology and capacity additions to satisfy prescribed service levels in the presence of product families with dynamic demands. An insightful analysis of the multi-plant capacity choice problem can be found in Jordan and Graves (1995); they develop principles for configuring a pre-existing network of plants for near-optimal process flexibility at a fraction of the cost for total flexibility. Benjaafar and Gupta (1998) develop explicit expressions for the number of facilities (work centers), the number of products assigned to each facility and their corresponding capacities, given a product mix for a plant.

Kulkarni et al. (2004), compare the choice of product-versus process-focused networks for a single-product two-market scenario. Conventional thinking about process plants suggests that they are very attractive when a firm can achieve economies of scale in component production. In Kulkarni et al. (2004) it is shown that a process plant configuration is also advantageous to firms with sufficient demand uncertainty to warrant risk pooling in component manufacturing. These firms would also typically have low shipping costs per unit and desire high service levels (high penalty costs). They consider the impact of pure risk pooling without scale/scope economies. It is also shown that firms that have an agenda to provide intermediate service levels (when compared to standard industry benchmarks) should also consider the process plant network strategy and that at very high and very low shortage costs per unit, product plant network strategies tend to dominate.

Recently, Wang and Gerchak (2003) model an assembly system with uncertain demand. They consider the case where the firm needs to construct individual production capacities before observing actual demand. They show that if the system is centralized then the optimal capacity acquisition decisions depend heavily on the system cost parameters. They also consider a decentralized setting wherein component manufacturing and assembly stages are owned by different firms. Although the work of Wang and Gerchak, addresses issues of power play between assemblers and suppliers in a game theoretic setting, their work is one of the first to highlight the "strategic complementarity" of decisions (i.e., more or less capacity of a component implies more or less capacity respectively of other components) in an assembly setting. We will show that the process plant network developed in this paper exhibits this type of strategic complementarity while also displaying classic strategic substitution (i.e., less capacity of one product can lead to more of another).

Harrison and Van Mieghem (1999) (henceforth HVM) develop a multi-stage model to invest resources within a single plant. Their model extends the classic newsvendor formulation to multiple products and multiple processing points. This "multi-dimensional newsvendor model" by HVM allows the investigation of much more complex problems in stochastic planning than would be possible with the traditional newsvendor model. Van Mieghem and Rudi (2002) (henceforth VMR) introduce a new class of models called "Newsvendor Networks" (NNs), which extend the multi-dimensional newsvendor model of HVM to include inventory holding points. This is a direct extension of the multi-dimensional newsvendor model to the multi-period or dynamic setting. NNs can be utilized to study capacity investment under uncertainty while capturing features such as commonality, flexibility, substitution and transshipment.

The single-period models developed and analyzed in this paper are related to both HVM and VMR. We use the framework developed in HVM and extended in VMR in order to carry out a detailed analysis of alternate network configurations. Specifically, we have three inputs (dedicated subassembly 1, dedicated subassembly 2 and common component) which are transformed into two outputs (product 1 and product 2) with a linear production technology structure and linear financial structure. The ex-ante activities include exogenous inventory procurement and capacity acquisition decisions and the ex-post activities include resource utilization to convert inputs to outputs. The ex-ante activities are carried out when demand is uncertain and the ex-post activities occur after demand realization; the classic "recourse" in a stochastic sense. VMR mention that NNs are about three decisions: (i) ex-ante capacity investment; (ii) ex-ante inventory procurement; and (iii) ex-post input-output transformation activities constrained by capacity. HVM's multi-dimensional newsvendor model on the other hand is only about two decisions: (i) ex-ante capacity investment; and (ii) ex-post input-output transformation activities constrained by capacity. An additional aspect of NNs is that the ex-post transformation activities are classified as "basic" or "discretionary." The basic activities represent the most profitable (alternatively, lowest cost) fulfillment of deterministic demand if the system is not capacity constrained. The non-basic or discretionary activities reflect flexibility to meet random deviation from the "expected" demand operating point as decided by the ex-ante capacity acquisitions. The identification and consideration of the value under uncertainty of this ex-post discretionary flexibility available in processing inputs is another point which differentiates NNs from the multi-dimensional newsvendor model. NNs can be used for network design and the power of NNs emerges when optimization of the general network yields the most effective network structure for the given set of cost and demand vectors. For a recent and insightful application of NNs which captures the equivalence between commonality strategies and flexible capacity/substitution see Van Mieghem (2004).

The specific scenario that we are concerned with in this paper is that of a firm assembling two products in two plants. Each of the products is assembled with a dedicated component and a common component. The common component may be produced in both plants (product-focused network) or in only one of the plants (process-focused network). Clearly, in the latter situation, the common component will have to be shipped to the plant which does not manufacture it. Thus,...

NOTE: All illustrations and photos have been removed from this article.



More articles from IIE Transactions
Design of Logic-Based Intelligent Systems.(Book Review), March 01, 2005

Looking for additional articles?
Search our database of over 3 million articles.

Looking for more in-depth information on this industry?
Search our complete database of Industry & Market reports by text, subject, publication name or publication date.

About Goliath
Whether you're looking for sales prospects, competitive information, company analysis or best practices in managing your organization, Goliath can help you meet your business needs.

Our extensive business information databases empower business professionals with both the breadth and depth of credible, authoritative information they need to support their business goals. Whether it be strategic planning, sales prospecting, company research or defining management best practices - Goliath is your leading source for accurate information.