پیکربندی زنجیره تامین برای انتشار محصولات جدید: روش بهینه سازی یکپارچه

We develop an integrated/hybrid optimization model for configuring new products’ supply chains while explicitly considering the impact of demand dynamics during new products’ diffusion. The hybrid model simultaneously determines optimal production/sales plan and supply chain configuration. The production and sales plan provides decisions on the optimal timing to launch a new product, as well as the production and sales quantity in each planning period. The supply chain configuration provides optimal selection of options and safety stock level kept at each supply chain function. Extensive computational experiments on randomly generated testbed problems indicate that the hybrid modeling and solution approach significantly outperforms non-hybrid alternative modeling and solution approaches under various diffusion and supply chain topologies. We provide insights on optimal production/sales plan and supply chain configuration for new products during their diffusion process. Also, managerial implications relevant to effectiveness of the hybrid approach are discussed.

In this paper, we consider the scenario where a firm needs to configure its supply chain before launching a new product. To respond to customer demand efficiently, the firm’s supply chain configuration encompasses decisions including selection of suppliers; manufacturing and transportation modes; as well as locations in supply chain network to place appropriate levels of safety stocks. Under a fixed production and supply capacity in the intermediate term, the firm might be overwhelmed by potentially rapid growth of demand for the new product due to marketing activities and positive word-of-mouth. Examples include Apple’s iPhone [1] and Nintendo’s Wii [2], where both manufacturers were hit by the rapid growth of demand for these innovative products. Often, such an impact affects not only the manufacturer itself, but also its vendors and suppliers through the supply chain. Example includes Apple’s PowerMac G4 [3], where Motorola, as the supplier of G4 chips, was not able to catch up with the rapid growth of demand for the popular computer. Another potential scenario for the firm is to experience a slow growth in demand for the new product and hence resulting in major financial risks. For example, initial sales of Sony’s Playstation 2 (PS2) were more than ten times that of the original PS’s introduction five years earlier [4]. However the launch of Playstation 3 (PS3) was not successful for Sony and hence resulted in $1.8B annual loss in its game division and layoff of 3% of its workforce [5]. Manufacturers are often able to save inventory cost by not keeping any initial stock before launching the new product, but they (and related players in the supply chain) may suffer later when supplies of the new product are outpaced by the fast growth of demand. Often, the saving on inventory cost may not compensate the cost due to lost demand. On the other hand, when a firm experiences demand below expectation, the inventory cost of safety stocks located at different tiers of supply chain network has a negative effect on efficiency. Thus when launching a new product, efficiency in terms of cost and speed is not the only quality a successful supply chain can own. As noted by Lee [6], supply chains that fail to adapt to changes in market structure will not gain sustainable competitive advantage. These have motivated us to model marketing-supply chain interactions, and in particular, the interaction between new product’s diffusion and the corresponding supply chain’s configuration. The dynamics of customer demand during diffusion of new products are well-captured by the classical Bass model [7]. Kumar and Swaminathan [8] and Ho et al. [9] have shown that the customer demand pattern during new product diffusion will affect the manufacturer’s production planning decisions during the new product’s lifetime. They extend the classical Bass model by considering production capacity of the firm, so that the demand of a new product may not be completely met due to the production capacity limit. Their model is used to find optimal production and sales plans that maximize profit during the new product’s lifetime, spanning from one to two years. They find that when supply constraint is present, the rapid growth of customer demand during diffusion may motivate manufacturer to buildup initial inventory and delay launching of the new product. We call their model the new product diffusion (NPD) model in the sequel. The NPD model focuses on the interactions between manufacturing and marketing/sales decisions within a firm by assuming a fixed per-unit product cost, but ignores other functions of the firm’s supply chain like procurement, sourcing, assembly and distribution. Graves and Willems [10] proposed a model optimizing the supply chain configuration for a new product, which we call the supply chain configuration (SCC) model. In this model, a firm selects options for each function (components, parts, or processes required) in the supply chain to minimize the system-wide total supply chain cost. Available options often differ in lead time and direct cost added. For instance, parts and raw materials can be purchased from different suppliers. Goods can be shipped via regular ground shipping or next day delivery. The SCC model also allows coordination among supply chain players by optimally determining their inbound and outbound service times, thus the inventory positioning through the supply chain. New product demand is assumed to be known in the form of mean and standard deviation for the entire planning horizon (usually 9 months–1 year). Because demand is exogenously given, the question of how the demand trajectory during new product diffusion will impact supply chain configuration is not addressed by the SCC model. The problems addressed by the NPD and SCC models are closely related. Both problems are tactical in nature. During the new product’s life cycle, the firm’s production and sales plan is only part of the big picture. Given the expanded complexity and scope of modern supply chains, it is rare to have a single firm being involved through all stages of sourcing, manufacturing, assembly, transportation, warehousing and delivery. Thus the firm is facing more important and wider scope of decisions on how to configure its entire supply chain to allow products as well as the required parts and components (described by the new product’s bill-of-materials or BOM) to be sourced, manufactured and delivered in an efficient and responsive manner. Therefore, there is merit in developing an integrated optimization model to study the optimal supply chain configuration decision in concert with dynamic process of new product diffusion. On one hand, the demand pattern (in terms of mean and variation) during new product diffusion has an explicit impact on supply chain configuration. Specifically, the mean customer demand serves as the external demand to be satisfied by the supply chain network, and the variation of demand directly impacts the amount of safety stock to carry (or inventory positioning) through the supply chain. On the other hand, the configuration of supply chain may in turn affect the optimal diffusion pattern of the new product. This is due to the fact that in the general supply chain settings, the per-unit cost of product should be calculated as an accumulative cost due to selection of suppliers/vendors and manufacturing/transportation modes through different supply chain stages such as sourcing, assembly, transportation, etc. This implies that the per-unit cost assumed to be constant in manufacturing planning models during diffusion, as in [8], can be extended and generalized to the so-called unit manufacturing cost (UMC) determined through configuring the corresponding supply chain [10]. In this study, we present a hybrid model to configure a new product’s supply chain by considering the dynamics of diffusion process through the product life cycle. Both the demand/supply pattern and unit-product cost are endogenously determined in one model, as opposed to being exogenous as assumed in the separate models. Our model offers a decision support tool for simultaneously optimizing an innovator’s production planning in a multi-period setting and supply chain configuration. It also provides a modeling framework to design a supply chain which is not only cost efficient, but also adaptive to the changing market demand during new products’ lifetime. As we will show, solutions that optimize supply chain performance from either the NPD aspect or the SCC point of view alone would not obtain optimal solutions. Lower supply chain configuration cost is often achieved by a myopic policy, i.e. selling as much as possible in each time period. This leads to less variation of the realized demand, thus less safety stock. However, such myopic policy may perform poorly from the diffusion perspective due to loss of demand. On the other hand, a buildup policy, i.e. delaying the launch of the new product and building some initial inventory, may generate higher sales revenue from the diffusion perspective; but too many buildup periods may lead to increased amount of safety stocks and hence increase supply chain configuration cost. Determining the optimal number of buildup periods will not be an intuitive task without the aid of an integrative optimization model in which both supply chain configuration decisions and new product diffusion outcomes are simultaneously considered. The remainder of the paper is organized as follows. Section 2 reviews the relevant literature. Section 3 presents the integrated model. The computational experimental study and results are presented in Section 4. Section 5 draws conclusions and suggests future research directions.

In this paper, we develop an integrated optimization model for configuring new products’ supply chains while explicitly considering the impact of demand dynamics during new products’ diffusion. It simultaneously determines optimal production/sales plan and supply chain configuration. The production and sales plan provides decisions on the optimal timing to launch a new product, as well as the production and sales quantity in each planning period. The supply chain configuration provides optimal selection of options and safety stock level kept at each supply chain function. The integrated model minimizes the total life-cycle profit during a new product’s entire life cycle. An in-depth computational experiment, including 2187 randomly generated testbed problem instances, was conducted to examine the performance characteristics of our integrated optimization model versus seven alternative heuristic policies under various diffusion and supply chain topologies. We obtain a number of managerial insights regarding production/sales planning and supply chain configuration for new products. (1) An optimal production/sales plan or supply chain configuration for a new product does not necessarily lead to maximum amount of sales revenue from the marketing perspective, or minimum amount of costs from the supply chain configuration perspective. An optimal production/sales plan and supply chain configuration during new products’ life cycle balances the tradeoffs among various cost and revenue components, and can only be achieved through an integrated optimization approach. (2) A smooth and even production/sales plan results in less inventory holding costs for the supply chain, but may suffer significant loss of sales by neglecting the dynamics of market demand. On the other hand, a production plan with certain number of buildup periods may increase sales revenue, but also incurs more supply chain configuration costs due to increased variation of sales/demand. Due to (1) and (2), our hybrid approach is more advantageous than models that consider production diffusion or supply chain configuration independently. The hybrid optimization approach is also robust for supply chain networks with different topologies characterized by the number of functions in the network and the lead time/direct cost profile of the new product. We find that as the inventory holding cost rate increases, it becomes more costly for the buildup policy to avoid sales loss, thus the benefit of hybrid model over buildup policy increases. When the percentage of backlogged demand (ξ) deviates from two extream cases, i.e. when ξ is neither high nor low as often the case in the real world, the benefit of hybrid model also increases. This research opens a number of extension opportunities for analyzing the integrated new product diffusion and supply chain configuration problem. Here, we mention a few of these potential research extensions. First, our current study assumes a generic product with an average coefficient of innovation (p) and coefficient of imitation (q). It will be interesting to study the impact of different product diffusion characteristics captured by paired p and q for specific products/industries. Some industry-specific decisions and/or constraints may also be modeled. For instance, designing supply chains in the fashion industry may emphasize responsiveness in terms of the supply chain cycle time, thus a deadline on the cycle time may be necessary to be included as a constraint. Products in the high-tech industry may have a different cost and lead time accrual profile than that of consumer durable goods. Second, some assumptions in the supply chain configuration can be relaxed. For example, one could relax the single-sourcing assumption and allow multiple options to be assigned to a function. Third, the current model considers only single product, it will be interesting to extend the model to consider a family of new products sharing the same BOM. This would require that we consider the diffusion process of multiple correlated products simultaneously. Fourth, in real-world settings competitors might offer simultaneously similar products in the marketplace with the intention of capturing as much market share as possible. Obviously, the competitive environment impacts the supply chain configuration decisions and the dynamic demand pattern emerges from the new product diffusion process. Then from the computational perspective, our current work relies on a commercial optimization solver to solve the integrated model. When problem size becomes large, solution times by solvers are not tractable. Thus a promising future research will be developing more advanced solution methods, e.g. various metaheuristics [36], to solve large size problems both effectively and efficiently.