Operations planning and exibility in a supply chain DOMENICO APRILE*, A. CLAUDIO GARAVELLI and ILARIA GIANNOCCARO Dimeg Politecnico di Bari, Viale Japigia 182, 70126 Bari, Italy In dynamic competitive markets, the exibility of manufacturing system networks such as supply chains (SCs) is particularly interesting. The SC exibility considered in this paper takes into account two main aspects: the process exibility of each SC rm and the logistics exibility concerning the possible connections between suppliers, assemblers and markets. Dierent congurations of an SC are proposed, in correspondence to dierent degrees of the process and logistics exibility. The eects of SC exibility are then investigated on the operations planning performance of an SC subject to production capacity uncertainty and coping with demand volume and mix variability. In particular, an optimization model is dened to analyse the SC performance in every SC conguration. Managerial guidelines, supporting the management of selecting the appropriate degrees of exibility and the corresponding SC conguration to be adopted, are nally obtained. Keywords: Supply chain; Process and logistics exibility; Operations planning; Performance 1. Introduction A supply chain (SC) is a network of organizations involved in dierent processes and activities producing value in the form of products and services in the hands of the ultimate customer (Christopher 1992). Supply- chain management (SCM) concerns the integrated and process-oriented approach to the design, management and control of the SC, with the aim of producing value for the end customer, by both improving customer service and lowering cost (Bowersox and Closs 1996, Giannoccaro and Pontrandolfo 2002). SCM mainly involves two classes of managerial problems at dierent levels: conguration, dealing with the design of the SC at a strategic level, and coordina- tion, concerning the management of the SC network prevalently under tactical and operative levels (Giannoccaro and Pontrandolfo 2001). Within SCM, a critical role is played by exibility, which is considered a crucial weapon to increase competitiveness in a turbulent marketplace (Upton 1994). In fact, SCs have to deal with many sources of uncertainty, such as customer demand, supply quality and lead-time, and information delay (Giannoccaro et al. 2002). In the literature, there are many studies on exibility of manufacturing systems, but most of them are limited to the single-rm perspective, and focused on the main factors inuencing the manufacturing exibility, in particular pointing out the dependence of exibility on the nature of the rms internal operations and exter- nal environment. On the contrary, as pointed out by Duclos et al. (2001), there are not many specic works that investigate exibility in the SC. SC exibility extends the concept of exibility of manufacturing systems to the entire SC. It encompasses not only the manufacturing exibility, but also the exibility of the dierent SC functions and processes, e.g. market- ing, supply, distribution and transportation (Vickery et al. 1999). *Corresponding author. Emails: daprile@poliba.it Production Planning & Control ISSN 09537287 print/ISSN 13665871 online # 2005 Taylor & Francis Ltd http://www.tandf.co.uk/journals DOI: 10.1080/09537280412331313348 Given the critical role played by exibility in SCM and due to the limited literature on the topic, more studies are needed to investigate the eects of the dier- ent types and degrees of SC exibility required to cope with several sources of SC uncertainty. This paper intends to oer a contribution in this research direction. In particular, two dierent types of SC exibility (process and logistics) are considered here. Generally, in the literature such dimensions have been either sepa- rately analysed, or studied with respect to a single rm more than to the SC as a whole. Our aim is to analyse the benets on SC performance due to dierent degrees of process and logistics exibility and to study the syner- gic eects coming from the simultaneous design of two logistics networks, namely supply and distribution. While this latter aspect has been studied by many scholars with respect to coordination problems (Bhatnagar et al. 1993, Whang 1995, Thomas and Grin 1996, Sarmiento and Nagi 1999, Ballou et al. 2000), it has been seldom emphasized in SC congura- tion problems. To pursue this aim, a model is proposed for the evaluation of the eects of dierent degrees of SC ex- ibility on the operations planning performance of a multi-product SC, subject to variable market demand and production uncertainty, and made up of several customers, assembly facilities and suppliers sites. Dierently from other simulation studies aimed at assessing SC performance such as lead-time and work- in-process (e.g. Garavelli 2003), a demand assignment problem is investigated in order to maximize the produc- tion planning performance of the SC. This is consistent with other studies focused on the exibility of a single- stage manufacturing system (e.g. Albino and Garavelli 1999). By this model we intend to derive some guidelines for the design and, in particular, for the selection of exibility degrees of SCs which t our model hypotheses. The paper is organized as follows. In section 2, the SC exibility is pointed out. In section 3, dierent SC ex- ibility congurations are described. In section 4, a linear programming model is developed for a multi-product SC facing a stochastic market demand. In section 5, the performance analysis of a sample case of SC is carried out. Final considerations are reported in section 6. 2. Supply-chain exibility Flexibility is a complex and multi-dimensional concept, dicult to summarize (De Groote 1994, Sarker et al. 1994, Upton 1994, Gupta and Buzacott 1996). According to a common denition, exibility reects the capacity of a manufacturing system to adapt successfully to changing conditions, coming from inside the system as well as from the environment. From an operational perspective, one of the most interesting aspects of the exibility concerns the content of change, i.e. exibility of product, mix, volume, etc. (e.g. Browne et al. 1984, Sethi and Sethi 1990, Hyun and Ahn 1992, Gerwin 1993, DSouza and Williams 2000, Vokurka and OLeary-Kelly 2000). Although the studies available on the specic subject of SC exibility are not very numerous yet, it is possible to nd a few denitions of SC exibility, which usually include the dierent types of manufacturing exibility, the exibility of the SC functions and the SC processes. For instance, Vickery et al. (1999) propose the following types of SC exibility: product, volume, launch, access and target market. Viswanadham and Raghavan (1997) consider volume, mix, routeing, deliv- ery time and new product exibility. Duclos et al. (2001) identify six components, i.e. product system, market, logistics, supply, organizational and information system exibility. The SC exibility addressed in this paper takes into account two main dimensions: (i) the process exibility of each SC rm, concerning the number of product types that can be manufactured in each production site (supplier or assembler), and (ii) the logistics exibility, related to the dierent logistics strategies, which can be adopted either to release a product to a market (downstream or distribution exibility) or to procure a component from a supplier (upstream or procurement exibility). While the process exibility is a well-known type of manufacturing system exibility (Jordan and Graves 1995, Garavelli 2001), the logistics exibility can be properly referred to the routeing exibility at the shop- oor level, that is the ability of using alternative routes to move the work-in-process through dierent resources oering the same processes (Ho and Moodie 1996, Das and Nagendra 1997). Logistics exibility is then here intended as the possibility of shifting the production assignment of an item (component or nal product) to dierent sites of a given stage of the SC, allowing the negative impact of demand and process variability on the SC performance to be reduced. As far as the process exibility is concerned, the costs associated with the plant capability of producing dier- ent items usually grow with the number of items. In most cases then, due to the high exibility investments and management costs required, it is not economic to produce all the items required by the global market in all the plants of a company (total exibility), which would be the case of the multi-domestic conguration, where local plants have to deal with a certain product dierentiation from one local market to another. 22 D. Aprile et al. From the opposite side, the focused factory congura- tion of a company, i.e. the specialization of each plant in a given production phase or in a few products (such as a product family) not released by other plants (no exibi- lity conguration), can also prove neither ecient nor eective, due for instance to the higher logistics costs and lower system exibility, respectively. An intermediate level of process exibility has been introduced by Jordan and Graves (1995). In fact, refer- ring to the production planning problem of a multi- plant company, they have shown the advantages of limited exibility, i.e. a conguration of plants and products that, chained together in forming the longest close loop, provides most of the benets of the total exibility conguration without requiring the corre- sponding investments and management costs. The concept of limited exibility is here considered in order to analyse three dierent situations for any production site (supplier or assembler): each site produces (a) one, (b) a limited number and (c) all the product or compo- nent types, corresponding to the degrees of no, limited and total exibility, respectively. In the model, the process exibility is split between two variables, namely the assembler and supplier exibility. As far as the logistics exibility is concerned, it involves dierent SC logistics strategies, which can be adopted in the procurement and distribution processes. The selection of a certain type of supplier can depend, for instance, on either the critical role of the component or the complexity of the logistics. Dierent procurement policies are then considered in the model, namely the single, the double and the multiple sourcing. This means that each assembler can purchase the needed components from (according to an increasing value of exibility): one (no), more than one (limited), all the available suppliers (total), respectively. Similarly, dierent distribution policies can be considered. In particular, each assembler can deliver its products to the local market (no), to more than one market (limited) and/or to all the markets (total). 3. SC exibility congurations Referring to the SC exibility considered in the previous section, we dene an SC conguration as a set of customers (nal markets), manufacturing sites (suppliers and assemblers), each releasing a certain number of components or product types (process exibility), and logistics links between suppliers and assemblers as well as between assemblers and markets (logistics exibility). Therefore, an SC conguration is based on four exibility variables, as reported in table 1. It is worth noting that, under some working hypo- thesis (demand market mean is equal to assembler/ supplier capacity, all the products have the same prot margin, distribution and/or supply channels have the same cost, assembler/supplier set-up and idle time are not considered), not all the combinations of the values of the SC exibility variables generate feasible SC congurations. First of all, referring to the intermediate degree of exibility, according to the denition of limited exi- bility provided by Jordan and Graves (1995) and considering a generic system made by N plants and N markets (each market demanding the same N products), N2 limited exibility congurations can be considered. Each of them is characterized by Q products manufac- tured by each plant, with 2 Q<N. When a higher Q is considered, the exibility of each plant increases. However, for the sake of brevity, we consider just the limited exibility conguration with Q2, since it provides the best costbenet tradeo, at least for a single stage of the SC (i.e. supplierassembler or assemblermarket). This implies that, referring to the process exibility, every supplier (assembler) produces two components (products), while referring to the logistics exibility, every supplier (assembler) can deliver its components (products) to two dierent assemblers (markets). Referring to the distribution part of the SC, made up of assemblers and markets (in gures 1 and 2, an example is reported with N5), in correspondence of the no exibility option, every assembler produces just one product (from A to E). In this case, every assembly plant provides every market with its own product. This involves distribution links between every assembler and market (total option) (gure 1a). In the case of the limited exibility option, given that every assembler produces a limited number of products, it is necessary to connect every assembler with all the markets, in order to serve all of them (gure 1b). In correspondence with the total exibility, every plant assembles all the pro- ducts. In this case it can adopt dierent distribution strategies, namely choosing among no, limited and total options (gure 2). Referring to the procurement part of the SC, made up of suppliers and assemblers, it is possible to extend the above considerations. In particular, in correspondence with the no exibility conguration every supplier plant produces just one component, while in the total exibil- ity conguration every supplier manufactures all the components. In between, there is the limited exibility conguration, where every supplier produces two com- ponents. In this part of the SC, since the assembler plants demand dierent products according to their planned exibility, the logistics links between suppliers Operations planning and exibility in a supply chain 23 and assemblers do not depend only on the supplier exibility, but also on the assembler one. Therefore, in this situation it is necessary rst to consider all the combinations resulting from the supplier and the assembler exibility, and then it is possible to infer the logistics options that are available. Nine combinations are then obtainable by amalga- mating the supplier and the assembler exibility (gure 3). In correspondence with the combination 1, 2, 3, 4, 6, 7 and 8, only one option of links between the suppliers and the assemblers is possible, namely no, limited, total, limited, total, total and total, respectively. In correspondence with combination 5, suppliers can choose to deliver only to the local assemblers (no), or to all the assemblers that require the components (limited). Table 1. The SC exibility variables. Type Variable Options Process exibility Assembler exibility No, limited, total Process exibility Supplier exibility No, limited, total Logistics exibility Procurement exibility No, limited, total Logistics exibility Distribution exibility No, limited, total A/B/C/D/E A/B/C/D/E A/B/C/D/E A/B/C/D/E A/B/C/D/E A/B B/C C/D D/E E/A Assemblers Markets Products Products (a) A/B/C/D/E A/B/C/D/E A/B/C/D/E A/B/C/D/E A/B/C/D/E A B C D E Assemblers Markets Products Products (b) Figure 1. Constrained logistics options for no and limited assembler exibility. A/B/C/D/E A/B/C/D/E A/B/C/D/E A/B/C/D/E A/B/C/D/E A/B/C/D/E A/B/C/D/E A/B/C/D/E A/B/C/D/E A/B/C/D/E Assemblers Markets Products Products A/B/C/D/E A/B/C/D/E A/B/C/D/E A/B/C/D/E A/B/C/D/E A/B/C/D/E A/B/C/D/E A/B/C/D/E A/B/C/D/E A/B/C/D/E Assemblers Markets Products Products A/B/C/D/E A/B/C/D/E A/B/C/D/E A/B/C/D/E A/B/C/D/E Assemblers Products Products A/B/C/D/E A/B/C/D/E A/B/C/D/E A/B/C/D/E A/B/C/D/E Markets Figure 2. Dierent logistics exibility options for total assembler exibility. Total 7 PF: total 8 PF: total 9 PF: no, limited, total Limited 4 PF: limited 5 PF: no, limited 6 PF: total No 1 PF: no 2 PF: limited 3 PF: total Supplier flexibility No Limited Total Assembler flexibility Figure 3. Feasible combinations of the procurement exibility (PF) with the supplier and assembler exibility. 24 D. Aprile et al. In the case of combination 9, all three logistics options are available. Based on the above considerations about, on the one side, feasible combinations of assembler and distribu- tion exibility (gures 1 and 2) and, on the other side, feasible combinations of supplier, assembler and procurement exibility (gure 3), it is possible to dene 22 SC exibility congurations. In fact, given that when the assembler is characterized by total exibility, three dierent distribution options are available, combina- tions 3 and 6 generate three dierent congurations each (3.1, 3.2, 3.3; 6.1, 6.2, 6.3), and the combina- tion 9 generates nine SC congurations (9.1 to 9.9). In gure 4, conguration 1 and conguration 9.1 are depicted as an example. 4. The operations planning model Let us consider a two-stage SC (i.e. suppliers, assemblers and nal markets), which produces P products and serves N customer markets located in dierent geographic areas. Each stage involves several production units (K suppliers and M assemblers), which are localized in dierent geographic places. Each market i demands D pi product units for each product family p. Demand D pi is a stochastic variable (gure 5). For the sake of simplicity, we assume that the assembler for each product p needs to purchase a single component from the supplier. Therefore, the supply stage totally produces P components. Let AC j be the (total) production capacity of the jth assembler and x p ij (with x p ij 0 if j do not make p) the demand share expressed by the ith market, referred to the pth product, assigned to the jth assembler. The production system of the assembler is considered as a exible ow line, that can manufacture any production mix required, so that the quota of AC j (namely AC p j ) allocated from the jth assembler to the pth product depends on the demand volume and mix required. Let SC k be the supply capacity of the kth supplier and x p jk (x p jk 0 if k do not make p) the supply capacity share of the pth product supplied by the kth supplier to the jth assembler. AC j and SC k are stochastic variables. According to assembly stage, the quota of SC j (namely SC p j ) allocated from the kth supplier to the pth product depends on the demand volume and mix required. The SC operations planning problem consists in dening the assignment of the market demand and of the supply production capacity to the assemblers, so as to optimize the SC performance in a given planning horizon. The performance analysed is the expected lost sales in the planning horizon. Therefore, the problem can be formulated in order to maximize the total produced quantity in the planning horizon, which is equivalent to minimize the lost sales. A/B/C/D/E A/B/C/D/E A/B/C/D/E A/B/C/D/E A/B/C/D/E A B C D E a b c e d Suppliers Assemblers Markets Components Products Products Configuration 1 A/B/C/D/E A/B/C/D/E A/B/C/D/E A/B/C/D/E A/B/C/D/E A/B/C/D/E A/B/C/D/E A/B/C/D/E A/B/C/D/E A/B/C/D/E a/b/c/d/e a/b/c/d/e a/b/c/d/e a/b/c/d/e a/b/c/d/e Suppliers Assemblers Markets Components Products Products Configuration 9.1 Figure 4. Two examples of SC congurations. AC j SC k D pi
x ik x ij
Supply stage Assembly stage Markets Figure 5. The SC model. Operations planning and exibility in a supply chain 25 The following linear programming model is dened: max z X M i X N j X P p x p ij ~ DD pi 1 subject to X P p X M i x p ij ~ DD pi A ~ CC j , 8j 2 X M i x p ij ~ DD pi
X K k x p kj S ~ CC k 0, 8j and 8p 3 X M i x p ij 1 8i and 8p 4 X P p X N j x p kj 1 8k 5 x p ij , x p kj 0 8i, 8j, 8k and 8p 6 Constraint (2) involves that the sum of the demand allocated to each assembler does not exceed the assembler production capacity. Constraint (3) balances the supply with the demand. In fact, it assures that the total quantity provided by the suppliers (i.e. the supply) is equal to the actual total quantity produced by the assemblers (i.e. the demand). Constraints (4), (5) and (6) are necessary to achieve a feasible solution. 5. Performance analysis A performance analysis is carried out to compare dier- ent SC exibility congurations subject to demand and capacity variability, assuming NMP5. This com- parison is based on the consideration that the proper choice among dierent degrees of exibility depends on a costbenet analysis, which takes into account, on the one hand, higher investments and management costs required by the exibility increase and, on the other hand, related better operations planning performance. This tradeo is just qualitatively addressed in this model, which is aimed at assessing and comparing the performance of limited exibility SC congurations that implement approximately 25% of the exibility totally achievable (i.e. one additional product or component over four per plant). The expected lost sales in each SC conguration are obtained by solving T times (T10,000) the stochastic linear programming model proposed in section 4. The SC is subject to two dierent sources of uncertainty, namely demand and capacity variability. In particular, two dierent patterns of demand vari- ability are taken into account: volume and mix variability. When demand variability is limited to product volume, the demand mix is equal to 20% for each product in each market. When mix variability is considered, the product volume is equal to 1,000, and each product share is determined by the normalization of each random product share value. Moreover, two dierent degrees of demand uncertainty are considered (i.e. high and low variability). In table 2, the dierent demand uncertainty patterns are shown. Two dierent patterns of uncertainty for the assem- bler and supplier capacity are also considered. They are modelled by adopting a uniform distribution as shown in table 3. Combining the uncertainty degrees of demand and capacity, four dierent settings are dened for every demand variability pattern (i.e. volume and mix). The analysis of results is carried out in two steps. First (subsection 5.1) we focus on the congurations characterized by higher logistics exibility in both procurement and distribution. In this way, it is possible to analyse the eect of process exibility, regardless of the inuence of logistics exibility. The analysed SC congurations are reported in table 4. Subsequently (subsection 5.2), to investigate the eect of the logistics exibility, we compare the performance achieved by the SC congurations characterized by the Table 2. Demand variability and uncertainty patterns. Demand variability pattern Variable Probability distribution function Volume Low variability (LV_D) Total market demand: Uniform [900, 1,100] High variability (HV_D) Total market demand: Exponential [1,000] Mix Low variability (LV_D) Product share: Uniform [18, 22] High variability (HV_D) Product share: Exponential [20] 26 D. Aprile et al. two extreme degrees of logistics exibility for each possible combination of supplier and assembler exibi- lity (table 5). In this case, only the two extreme settings are considered, characterized by low demand and capacity uncertainty and high demand and capacity uncertainty, respectively. 5.1 Eects of the process exibility For each demand variability pattern, and for all the uncertainty settings, the SC performance improves as the process exibility increases (tables 6 and 7). However, limited exibility congurations perform quite similarly to the total exibility ones, so that the former ones have to be preferred to the latter ones regardless of the exibility costs. Below we discuss in detail the results for each pattern of demand variability in each setting. 5.1.1 Demand volume variability. The congurations characterized by limited degrees of exibility (C5.2, C6.3 and C8) achieve the same performance as the most exible conguration (C9.9) in all the uncertainty settings (table 6). Then, it is not benecial for a limited exible SC (C5.2) to increase the exibility of either the supply or the assembly stage (C6.3, C.8 and C9.9). It is also possible to group the SC congurations in four classes, each of them characterized by similar performances, namely C1; C2 and C3.3; C4 and C7; C5.2, C6.3, C8 and C9.9. In setting (1), lost sales are quite low and the dierence between the best and the worst performance is limited to 2.15%. In setting (2), due to the increase of demand variability, the performance considerably decreases, but the dierence between the best and the worst value is still quite low (4.86%). Notice that a high degree of process exibility at the supplier sites is more benecial than at the assembler sites. In fact, C2 and C3.3, characterized by no supplier exibility, achieve a worse performance than C4 and C7, which are characterized by no assembler exibility. In setting (3), when the demand variability is low and the production capacity variability is high, the lost sales increase, but less than in setting (2). However, the dierence between the best and the worst perfor- mance is higher (7.88%). Opposite to setting (2), in this case C2 and C3.3 perform much better than C4 and C7, which perform very close to the total exible conguration. Therefore, to cope with high capacity uncertainty, high degrees of process exibility are more appropriate at the assembler sites. In setting (4), there is a general decrease of perfor- mance, since lost sales are high as well as the dierence between the best and the worst performance (7.11%). Table 4. SC congurations and process exibility. SC congurations Supplier exibility Assembler exibility Procurement exibility Distribution exibility C1 No No No Total C2 No Limited Limited Total C3.3 No Total Total Total C4 Limited No Limited Total C5.2 Limited Limited Limited Total C6.3 Limited Total Total Total C7 Total No Total Total C8 Total Limited Total Total C9.9 Total Total Total Total Table 5. SC congurations and logistics exibility. SC congurations Procurement exibility Distribution exibility C3.1 Total No C3.3 Total Total C5.1 No Total C5.2 Limited Total C6.1 Total No C6.3 Total Total C9.1 No No C9.5 Limited Limited C9.9 Total Total Table 3. Capacity uncertainty patterns. Capacity Probability distribution function Low variability (LV_C) Uniform [900, 1,100] High variability (HV_C) Uniform [500, 1,500] Operations planning and exibility in a supply chain 27 5.1.2 Demand mix variability. When the demand mix varies, the congurations C5.2, C6.3, C8 and C9.9 achieve an equivalent performance only in the case of low demand variability. The same result holds for the couples of congurations C2 and C3, C4 and C7 (table 7). However, the dierences among these groups of congurations remain negligible as the demand variability increases. In setting (1), lost sales are low. As the SC exibility increases, the performance improvement is quite low. In fact, the dierence between the best and worst result is 2.43%. Compared to setting (1), in setting (2) the SC per- formance signicantly decreases in all the congura- tions, but the eect of process exibility is very considerable, since it improves the SC performance as the SC exibility degree increases. Moreover, as observed in the case of demand volume variability, the increase of the process exibility at the supplier sites is more benecial than on the assembler ones. In setting (3), the eect of production capacity variability can be observed. In this setting, SC perfor- mance is similar to the case of demand volume vari- ability. However, here the dierence between the best and the worst performance (8.05%), even if higher than in setting (1), is lower than in setting (2) (17.11%). As for the demand volume variability, the results show that it is better to increase the process exibility at the assembler sites rather than at the supplier ones. In setting (4), a general decrease of the SC perfor- mance is measured. Here the performance dierence between the less and most exible conguration rises to 17.88%, emphasizing the benecial impact of exi- bility on the SC, at least as the limited degree of exibility is reached. 5.2 Eects of the logistics exibility To evaluate the eects of the logistics exibility, we compare the performance achieved by the SC congurations C3.1 and C3.3; C5.1 and C5.2; C6.1 and C6.3; C9.1, C9.5 and C9.9 (see table 5). The results are shown in table 8. As the logistics exibility increases, lost sales decrease for all SC congurations in all the settings and for both demand variability patterns. For both SC congurations C3 and C6, character- ized by total process exibility at the assembler stage, the increase of the distribution exibility from no (3.1 and C6.1) to total (3.3 and C6.3) yields its main Table 6. Lost sales performance under volume variability. (1) (2) (3) (4) SC congurations LV_D & LV_C HV_D & LV_C LV_D & HV_C HV_D & HV_C C1 4.35% 25.28% 13.91% 28.84% C2 3.06% 22.68% 6.70% 23.77% C3.3 3.10% 22.68% 6.70% 23.75% C4 3.05% 20.49% 12.60% 23.76% C5.2 2.20% 20.42% 6.03% 21.69% C6.3 2.20% 20.42% 6.03% 21.69% C7 3.05% 20.48% 12.60% 23.73% C8 2.20% 20.42% 6.03% 21.69% C9.9 2.20% 20.42% 6.03% 21.69% Table 7. Lost sales performance under mix variability. (1) (2) (3) (4) SC congurations LV_D & LV_C HV_D & LV_C LV_D & HV_C HV_D & HV_C C1 4.19% 22.82% 13.81% 26.70% C2 2.84% 18.65% 6.59% 19.56% C3.3 2.84% 18.65% 6.59% 19.48% C4 2.85% 16.90% 12.54% 19.64% C5.2 1.76% 6.32% 5.76% 9.51% C6.3 1.76% 6.21% 5.76% 9.20% C7 2.85% 16.87% 12.54% 19.55% C8 1.76% 5.90% 5.76% 9.20% C9.9 1.76% 5.71% 5.76% 8.82% 28 D. Aprile et al. benets in the demand volume variability pattern, while the impact is far lower in the demand mix one. The performance of SC congurations C5.1 and C5.2 is quite good for both variability settings, so that the increase of the procurement exibility from no (5.1) to limited (5.2) provides considerable benets also in the demand mix variability pattern. As expected, the best performance is achieved by the SC conguration C9.9, characterized by total process and logistics exibility. However, the performance of SC conguration C9.5, characterized by limited logistics exibility (in both procurement and distribution), is similar to C9.9. Therefore, C9.5 qualitatively appears as a conguration with a better tradeo between logistics exibility costs and SC performance. Comparisons between SC congurations character- ized by similar performance but dierent degrees of process, procurement and/or distribution exibility are possible, but they require considerations about the costs of the dierent types of exibility. However, it can be qualitatively observed that the SC congura- tion C5.2, with a limited degree of procurement exi- bility (and limited process exibility at both supplier and assembler stages), provides SC performance very close to more exible SC congurations, showing again the advantages of limited exibility. 5.3 Discussion of the results From the performance data analysis, it is possible to infer the following guidelines about the selection of a proper exibility degree for SC congurations which t the model hypotheses: (a) The limited process exibility (at either supplier or assembler stages) yields tangible benets to the SC performance. In fact, an increase of the process exibility degree from limited to total does not improve the SC performance in the case of demand volume variability, while the improve- ment is not relevant in the case of demand mix variability. The SC conguration with limited process exibility at both supplier and assembler stages then provides a good tradeo between SC performance and cost. (b) Higher process exibility in the supply stage allows high demand variability to be faced, producing better SC performance. On the cont- rary, in the case of high capacity uncertainty, higher process exibility in the assembler stage is more benecial. These interesting behaviours, in particular, can be observed in correspondence of any demand variability pattern (mix or volume). (c) An increase of the logistics exibility degree, within the feasible congurations, improves the SC performance in every situation, even though the improvement can be insignicant in some cases. The limited logistics exibility results are very eective, in both cases of limited and total process exibility. 6. Conclusions This paper has addressed the topic of SC exibility by focusing on two types of exibility playing a critical role in SC management, namely process and logistics exibility. Two main aspects have emerged from the study. The rst aspect concerns the design of SC congura- tions, based on several options of exibility. The SC conguration feasibility, resulting from the combina- tion of dierent degrees of process and logistics exibility, has been investigated. The study has pointed out that when the SC is taken into account, the process Table 8. Lost sales performance. Volume Mix (1) (4) (1) (4) SC congurations LV_D & LV_C HV_D & HV_C LV_D & LV_C HV_D & HV_C C3.1 3.92% 39.04% 3.33% 21.36% C3.3 3.06% 23.75% 2.84% 19.48% C5.1 3.63% 26.83% 3.37% 17.13% C5.2 2.20% 21.69% 1.76% 9.51% C6.1 3.62% 38.93% 2.82% 14.25% C6.3 2.20% 21.69% 1.76% 9.20% C9.1 5.02% 44.83% 4.15% 20.76% C9.5 2.20% 25.40% 1.76% 8.89% C9.9 2.20% 21.69% 1.76% 8.82% Operations planning and exibility in a supply chain 29 exibility limits the choices of the SC logistics strategies that can be adopted in supply and distribution stages. The second aspect concerns the evaluation of the SC performance in correspondence with dierent degrees of exibility. A model for the optimization of the lost sales of a multi-product SC, made up of several customers, assembling facilities and suppliers sites, has then been dened. It has been used to compare dierent SC congurations under demand variability and production capacity uncertainty. The analysis of the results has stressed that SC congurations characterized by limited degrees of both process and logistics exibi- lity are often more convenient than the corresponding total exibility ones, given that they achieve similar SC performance but involve lower nancial eorts and complexity. A further interesting result is that the higher the demand uncertainty, the more benecial the exibi- lity at the supply level, while the higher the capacity uncertainty, the more benecial the exibility at the assembler stage. In particular, the impact of the demand variability on the upstream SC stages recalls the well-known bullwhip eect, which stresses the increase of variability moving upstream along the SC (Lee et al. 1997). Further developments of this study seem to be promising. For instance, it can be interesting to consider more SC stages or to introduce levels of asymmetry in the SC congurations, in order to extend the analysis to more kinds of SC congurations and to explore further result generalizations. The relationship between the degrees of SC exibility and the bullwhip eect could be a further aspect to be investigated. Dierent SC exibility dimensions could also be included in the study, and dierent SC performances could be measured (e.g. back orders, service level, inventory stocks). In particular, the quantication of the tradeo between exibility benets and costs could also be addressed in order to provide management with a more precise evaluation tool. This quantication seems especially useful in those cases where SC congurations character- ized by dierent degrees of process, procurement and/or distribution exibility provide similar SC performance, thus requiring the cost evaluation of the dierent types of exibility. References Albino, V. and Garavelli, A.C., Limited exibility in cellular manufacturing systems: a simulation study. International Journal of Production Economics, 1999, 6061, 447455. Ballou, R.H., Gilbert, S.M. and Muckherjee A., New manage- rial challenger from supply chain opportunities. Industrial Marketing Management, 2000, 29, 719. Bhatnagar, R., Chandra, P. and Goyal S.K., Models for multi-plant coordination. European Journal of Operational Research, 1993, 67, 141160. Bowersox, D.J. and Closs, D.J., Logistical Management: The Integrated Supply Chain Process, 1996 (McGraw-Hill). Browne, J., Dubois, D., Rathmill, K., Sethi, S.P. and Stecke, K.E., Classication of exible manufacturing systems. The FMS Magazine, 1984, 2(2), 114117. Christopher, M., Logistic and Supply Chain Management, 1992 (Pitman). Das, S.K. and Nagendra, P., Selection of routes in a exible manufacturing facility. International Journal of Production Economics, 1997, 48, 237247. De Groote, X., The exibility of production processes: a general framework. Management Science, 1994, 40(7), 933945. DSouza, D. and Williams, D., Toward a taxonomy of manufacturing exibility dimensions. Journal of Operations Management, 2000, 18(5), 577593. Duclos, L.K., Lummus, R.R. and Vokurka, R.J., A concep- tual model of supply chain exibility, in Proceedings Paper presented at the 32nd Decision Science Institute Annual Meeting, 1720 November, San Francisco, 2001. Garavelli, A.C., Performance analysis of a batch production system with limited exibility. International Journal of Production Economics, 2001, 69(1), 3948. Garavelli, A.C., Flexibility congurations for the supply chain management. International Journal of Production Economics, 2003, 85(2), 141153. Gerwin, D., Manufacturing exibility: a strategic perspective. Management Science, 1993, 39(4), 395410. Giannoccaro, I. and Pontrandolfo, P., Models for supply chain management: a taxonomy, in Proceedings of the POM-2001 Conference: POM Mastery in the New Millennium, 30 March to April 2, Orlando, Florida, 2001. Giannoccaro, I. and Pontrandolfo, P., Inventory management in supply chains: a reinforcement learning approach. International Journal of Production Economics, 2002, 78(2), 153161. Giannoccaro, I., Pontrantrandolfo P. and Scozzi B., A fuzzy echelon approach to inventory management in supply chains. European Journal of Operational Research, 2002, 149, 185196. Gupta D. and Buzacott, J.A., A goodness test for operational measures of manufacturing exibility. International Journal of Manufacturing Systems, 1996, 8, 233245. Ho, Y.C. and Moodie, C.L., Solving cell formation problems in a manufacturing environment with exible processing and routing capabilities. International Journal of Production Research, 1996, 34(10), 29012923. Hyun, J.-H. and Ahn, B.H., A unifying framework for manufacturing exibility. Manufacturing review, 1992, 5(4), 251260. Jordan, W.J. and Graves, S.C., Principles on the benets of manufacturing process exibility. Management Science, 1995, 41(4), 577594. Lee, H.L., Padmanabhan, V. and Whang, S., The bullwhip eect in the supply chains. Sloan Management Review, 1997, 38(3), 93102. Sarker, B.R., Krishnamurthy, S. and Kuthethur, S.G., A survey and critical review of exibility measures in manufacturing systems. Production Planning & Control, 1994, 5(6), 512523. 30 D. Aprile et al. Sarmiento, A.M. and Nagi, R., A review of integrated analysis of productiondistribution systems. IIE Transactions, 1999, 31, 10611074. Sethi, A.K. and Sethi, S.P., Flexibility in manufacturing: a survey. International Journal of Flexible Manufacturing, 1990, 2(4), 289328. Thomas, D.J. and Grin, P.M., Coordinated supply chain management. European Journal of Operational Research, 1996, 94(1), 115. Upton, D.M., The management of manufacturing exibility. California Management Review, 1994, Winter, 7289. Vickery, S., Calantone, R. and Droge, C., Supply chain ex- ibility: an empirical study. The Journal of Supply Chain Management, 1999, August, 1624. Viswanadham, N. and Srinivasa Raghavan, N.R., Flexibility in manufacturing enterprises. Sadhana, 1997, 22(2), 135163 (http://lcm.csa.iisc.ernet.in/ raghavan/news adhana. html). Vokurka, R.J. and OLeary-Kelly, S., A review of manufactur- ing exibility empirical research, Journal of Operations Management, 2000, 18(4), 485501. Whang, S., Coordination in operations: a taxonomy. Journal of Operations Management, 1995, 12, 413422. Domenico Aprile Born in Lecce on August 29th 1972, in 2001 he graduated in Computer Science Engineering at the University of Lecce, Italy. In 2002 he got a MS in Supply Chain Management at the University of Brescia, Italy. After that, for a short period he stayed at the Signal Lux Company S.p.A., Milan (Italy), working on Advanced Planning/Production System/Scheduling (APS). Since 2003 he has been attending the Ph.D. program in Advanced Production Systems at the Politechnic University of Bari, Italy, where he is doing research on two main streams: supply chain exibility, and bullwhip eect. Achille Claudio Garavelli Ph.D. in Engineering Management, he has been Assistant Professor at the University of Basilicata, Italy, since 1994. Visiting scholar in 1996 at the University of South Florida (Tampa, USA), he was formerly Associate Professor at the University of Lecce, Italy, and now at the Polytechnic University of Bari, Italy. His main teaching and research areas concern operations management, knowledge management, organisation networks. He is involved in many national and international research projects and he is author of more than 70 papers published on national and international journals and conference proceedings. Ilaria Giannoccaro Born on October 9th, 1974, in 1998 she got her laurea degree in Mechanical Engineering and in 2001 her Ph.D. in Business Engineering at the Tor Vergata University of Rome (Italy). She is now Assistant Professor at the Polytechnic University of Bari, Italy. She is author of more than 30 papers published on international journals and conference proceedings. Her main research interests include supply chain management, inventory management, and supply chain contracts. She is currently studying the application of multi-agent systems to the analysis of industrial clusters. Operations planning and exibility in a supply chain 31