Inventory configurations and drivers: An international study of assembling industries

Inventory configurations and drivers: An international studyof assembling industries$

Krisztina Demeter a, Ruggero Golini b,n

a Department of Logistics and Supply Chain Management, Corvinus University of Budapest, Hungaryb Department of Engineering, Università Di Bergamo, Viale Marconi 5, 24044 Dalmine (BG), Italy

a r t i c l e i n f o

Article history:Received 24 October 2012Accepted 30 October 2013

Keywords:ConfigurationsInventory strategyIMSS

a b s t r a c t

In recent decades, inventory reduction has been a key objective of companies in various industries and isparticularly important in the current crisis. Inventory is closely related to a company's production systemand supply chain and a one-way strategy towards zero-inventory can be inapplicable or too general. As amatter of fact, there is a complex relationship between inventory types (input, WIP and output) and thefactors causing or affecting them. On the basis of three editions of a survey in different assemblyindustries (IMSS) carried out in 2001, 2005 and 2009, we demonstrate in this paper that the actualconfigurations that companies adopt, as well as the factors behind the chosen configurations, are stableand consistent over time, in terms of the levels of each type of inventory. We also show that not all of thecompanies are stuck in a configuration; with the right measures, they can reduce the stock of inventoryand become more competitive.

& 2013 Elsevier B.V. All rights reserved.

1. Introduction

Inventories are costly. For instance, in the automotive industry,the level of inventory to be financed compared to the level of saleswas 4.4% in Japan and 8% in the US from 1990 to 1993 (Liebermanand Asaba, 1997). Nevertheless, inventories are unavoidable parts ofoperations. Inventories are generally kept in order to: (a) prepare forfuture operations (anticipation inventory); (b) cover usage betweentwo supplies (decoupling inventory); (c) provide economies of scalein production and deliveries (cycle inventory) and (d) buffer againstdemand uncertainties (buffer inventory) (Slack et al., 2007). Each ofthese purposes is applicable for each type of inventory: input, work-in-process (WIP) and finished goods (FG). However, certain factorsaffect these inventory types differently.

First, regular shortages can lead to higher levels of inputinventories (Kornai, 1979; Kraljic, 1983). Consequently, companiesmay buy and pile up raw materials to avoid production stoppages.In contrast, companies in competitive product markets keeprelatively high levels of FG inventories to provide products tocustomers at any time. Other key factors, related to the supply

chain (Jones and Riley, 1985) or the production characteristics(Demeter and Matyusz, 2011), as well as the strategy of firms(Ward et al., 1996), lead to different inventory configurations, thatis, to different ratios of input, WIP and FG inventories.

Understanding the drivers behind these configurations is funda-mental to setting up an inventory optimisation strategy because thedrivers can have very different impacts on inventory levels. Forinstance, higher product variety can increase each type of inventory:more kinds of input are needed, more kinds of FG are produced andproduction processes become more complex, resulting in higher WIPinventories. On the other hand, changing the decoupling point(Olhager, 2003) by allowing customer orders to enter ahead intothe production flow can change the ratios of inventory types bydecreasing the ratio of FG inventories and increasingWIP inventories.

As a result, identifying the typical inventory configurations and thedrivers behind them provide a powerful strategic tool to inventorymanagers. Inventory managers can understand the constraints of theirinventory reduction efforts and make their decisions more logically.

Inventory characteristics may vary among industries, however.Process industries (with non-discrete products) and assembly indus-tries (with discrete products) differ greatly from one another (Dennisand Meredith, 2000). While the variety of inputs in the processindustries is usually low and there are only a few places within theproduction process to keep inventories, the variety of FG can be high.Sometimes these processes are referred to as analytic, meaning thatsingular input is processed into many separate outputs (Meredith,1992). Nevertheless, there are large differences even within theprocess industry (Dennis and Meredith, 2000). However, in the

Contents lists available at ScienceDirect

journal homepage: www.elsevier.com/locate/ijpe

Int. J. Production Economics

0925-5273/$ - see front matter & 2013 Elsevier B.V. All rights reserved.http://dx.doi.org/10.1016/j.ijpe.2013.10.018

$We would like to thank the reviewers (and especially the second reviewer) fortheir constructive comments that led to significant improvements of the paper. Thisresearch project was partially funded by Fondazione Cariplo within the FYRE(Fostering Young REserchers) program.

n Corresponding author. Tel.: þ39 035 2052 048.E-mail addresses: [email protected] (K. Demeter),

[email protected] (R. Golini).

Please cite this article as: Demeter, K., Golini, R., Inventory configurations and drivers: An international study of assembling industries.International Journal of Production Economics (2013), http://dx.doi.org/10.1016/j.ijpe.2013.10.018i

Int. J. Production Economics ∎ (∎∎∎∎) ∎∎∎–∎∎∎

www.sciencedirect.com/science/journal/09255273

www.elsevier.com/locate/ijpe

http://dx.doi.org/<?tbklnk=







assembly industry, the input variety is usually high, there are severalpoints to keep WIP and the variety of FG can also be high. Assemblyindustries are called synthetic because many materials come togetherto form a singular discrete output. We focus on the assemblyindustry in this paper.

The aim of this paper is to illuminate the typical inventoryconfigurations and their relationship with internal and externalfactors within the assembly industry. We start with a literaturereview to identify possible factors related to different types ofinventories. Next, using data from the 2001, 2005 and 2009International Manufacturing Strategy Surveys, we identify differ-ent inventory-based configurations and relate these configurationsto explanatory factors. Finally, we discuss our findings and drawconclusions.

2. Literature review

Inventories have a clear impact on the financial performance offirms (Capkun et al., 2009). Chikán (2011, 2009) reported thatinventories “serve as strategic tools in achieving customer satisfactionand profit simultaneously”, but that they can be “efficiently managedonly as parts of the supply chain, jointly with other company functions”.These two statements show, on one side, the strategic importance ofinventories and, on the other side, the growing complexity ofeffective inventory management. As a matter of fact, inventorydecisions must be coordinated among company functions (e.g.,purchasing, manufacturing, logistics and marketing). These functionsare responsible for different types of inventories, namely input (i.e.,material and components), WIP and FG. Moreover, while WIPinventory represents an intra-firm buffer, input and FG inventoriesare called inter-firm buffers because they must be coordinated withsuppliers and customers (Lieberman and Demeester, 1999). Thus,discussing inventory management issues inevitably requires invol-ving both intra- and inter-company factors.

In recent decades, lean management has had the largest impacton inventory management practices and performance (Capkunet al., 2009). Following lean principles, companies can reduce thelevel of different types of inventories. A trend of WIP inventoryreduction has occurred in many industries, starting in the Japaneseautomotive industry (Lieberman and Demeester, 1999).

Chen et al. (2005) extended their analysis to other inventorytypes and found a similar pattern of reduction (2%, on average,between 1981 and 2000). However, the FG inventories tended tostay more stable compared to input and WIP, both of whichsignificantly decreased. The stability in FG inventories was con-firmed by Rajagopalan and Malhotra (2001), who found that inputand WIP inventories significantly decreased from 1961 to 1994 inmost industries, while FG decreased in only a few industries (suchas the electric, electronic equipment, rubber, leather, food andtobacco industries).

Lean management has some limitations; however, particularly inregard to product variety and high geographical distances(Cusumano, 1994). Moreover, being too lean can be counterproduc-tive. There is a broad range of literature on lean versus agile supplychains (Bruce et al., 2004; Cagliano et al., 2004; Goldsby et al., 2006;Mason-Jones et al., 2000; Naylor et al., 1999), which claim thatinventories in fast markets can create higher responsiveness andavoid lost sales. Moreover, in global contexts characterised by risksand uncertainties, inventories can play a buffer role in hedgingpossible supply or production disruptions (Juttner et al., 2003).

Additionally, empirical evidence has demonstrated that beingtoo lean is not always good. Chen et al. (2005) show that firmswith very high inventories have poor long-term stock returns.However, very low inventory is not associated with higherfinancial performance.

Consequently, some companies keep some inventory where itis more strategic. This decision depends upon some characteristicsof the market, of the supply chain or of the company (see Fig. 1). Atthis stage, it is important to specify that, even if we take intoaccount supply chains (which have a large impact on business unitinventories), our unit of analysis is the business unit and we do notconsider inventory optimisation issues at the supply chain level.

Adopting a contingency-based perspective, as advocated for ourdiscipline (Sousa and Voss, 2008), we checked the literature for thefactors that can influence the aforementioned types of inventories. Inparticular, we identified four groups of factors (Fig. 1):

1) Market factors (both supplier and customer sides) that give theexternal context for company operations (Porter, 1980) andinventory decisions

2) Internal operations, which provide the internal context throughexisting operating technologies, processes and procedures

3) The characteristics of the supply chain the company belongs to,which determines the type and variety of materials andproducts it buys, produces and sells, as well as the partnerrelationships themselves

Business strategy, which has an impact on the prioritiescompanies follow in their internal and external decisions. It shouldbe emphasised that the factors above are logically interrelated. Forinstance, market factors affect not only input and FG inventory,they may also influence WIP, as more refined needs of customersmay drive implementation of manufacturing customization, whichmeans interference into WIP. Nevertheless, we focus on the mostimportant impacts in our paper.

These direct or mediated impacts on inventories (e.g., Cachonand Olivares, 2010) will be discussed in the following sections.

2.1. Market factors

Market factors influence input and FG inventories. The powerof partners, and thus the conditions of services and products theyprovide/expect, depend on the competition on the market.

In resource-constrained systems, for example, if the number ofsuppliers is limited or uncertainty of deliveries is high, customerstend to buy larger amounts at one time and are more ready tosubstitute. In these markets, it is easy to sell and difficult to buy.The ratio of FG is low for suppliers, while the ratio of input is highfor customers. In demand-constrained systems, the situation is theopposite (Chikan, 1996; Kornai, 1979).

Furthermore, material cost plays an important role in inventorydecisions. If the material cost is high compared to other costs (e.g.,work or overhead), then holding inventory of any type, even ofraw materials, will be more costly (Beamon, 1998), forcing com-panies to reduce the inventory. Material costs depend heavily onmarket characteristics, such as the number and bargaining powerof suppliers (Porter, 1980).

Supplier Customer

The company

Internal operations

Business strategy

Customer

CustomerSupplier

Supplier

MARKETMARKET

Input

Work-in-process

Output

supply chain

Fig. 1. Factors influencing business unit inventories.

K. Demeter, R. Golini / Int. J. Production Economics ∎ (∎∎∎∎) ∎∎∎–∎∎∎2





Certainly, particular customers/suppliers can have furtherrequirements towards/from suppliers/customers, such as deliverylead time, delivery reliability, or frequency of delivery. Thesefactors, which finally define the characteristics of customer ser-vice, however, are not general characteristics of markets and canbe addressed by the business strategy of companies (see later) andby individual contracts with partners.

2.2. Internal operations

The customer order penetration point, which “defines the stagein the manufacturing value chain, where a particular product islinked to a specific customer order” (Olhager, 2003, p. 319), as wellas the nature of the manufacturing process (jobbing, batch or massproduction) determine the level of WIP and influence the level ofinput and output inventories.

The literature has identified a close relationship between theposition of the decoupling point, sometimes called the orderpenetration point (Olhager, 2003) and the level of the differenttypes of inventories. In particular, Naylor et al. (1999) identify fivedifferent possibilities in which the decoupling point can bepositioned corresponding to different types of inventory buffers.These types are: buy-to-order (the stock is at the supplier's),make-to-order (internal raw materials and components stock),assemble-to-order (internal WIP stock), make-to-stock (internalfinished products stock) and ship-to-stock (the stock is at custo-mer's). Olhager (2003) uses four points: engineer to order (ETO),make to order (MTO), assemble to order (ATO) and make to stock(MTS) – a more accepted and common categorisation in theliterature. As he explains, shifting the decoupling point backwardreduces or eliminates WIP buffers and reduces the risk of obsoles-cence of inventories. Shifting forward, however, increases the levelof WIP, due to more items being forecast-driven. The relationshipbetween the decoupling point and inventory types was empiri-cally tested in the manufacturing industry by Demeter andMatyusz (2011). They found partial confirmation of the theoreticalassociation. In this study, make-to-order companies tend to havehigher input inventories and make-to-stock companies havehigher FG inventories. However, assemble-to-order companieshave lower input levels but not higher WIP inventory. This showsthat other explanatory factors have to be taken into account for aclearer picture.

The type of production process can also impact inventory ratiosand levels. Following the product-process matrix (Hayes andWheelwright, 1979), the number of different raw materials andcomponents in job shops can be very high, although the amount ofeach is low. As we move towards the assembly line process, thenumber of different inputs does not necessarily fall, but theamount of each is definitely higher due to the mass productionnature. WIP is different. Due to high complexity, WIP level can behigh in job shops, while likely decreasing in well-designedassembly line processes. Finally, the customers in job shops areusually known at the start of the process, so the FG can bedelivered quickly, while assembly lines frequently work on fore-casts and thus keep FG inventories. Empirical evidence shows(Demeter and Matyusz, 2011) that companies organised in jobshops have higher WIP, while companies with dedicated lines havelower WIP. No specific effect is found, however, for companieswith cellular layout.

2.3. Supply chain characteristics

Supply chain position (upstream or downstream), coordinationwith partners or partner proximity can influence the ratio ofinventory types.

Following the principles of the bullwhip effect, demand varia-bility is expected to be higher in the upstream stages, producinghigher inventory levels (Lee et al., 1997). Thus, the supply chainposition can impact inventories (Beamon, 1998). However, thisevidence is not conclusive, as some authors (Cachon et al., 2007)have found the opposite: the supply stages close to the finalmarket have had higher demand variability. Using up bufferinventories downstream along supply chains can help to set offthe amplifications of short-term demand increase (Scott andWestbrook, 1991).

Information sharing within and between companies is vital inmanaging inventories efficiently (Cachon and Fisher, 2000; Leeet al., 2000). According to traditional theories, inventories can bebetter managed in vertically integrated firms thanks to easierinternal rather than external communication and coordination(D'Aveni and Ravenscraft, 1994). Nevertheless, technology andmanagerial improvements currently give more room for externalinformation sharing. Using computer-based information technol-ogies and systems such as vendor-managed inventory (VMI) andefficient customer response (ECR), collaborative planning andforecast (CPFR) systems can reduce supply chain costs up to 35%(Cachon and Fisher, 2000). Further benefits of ICT include reducedlead times and smaller delivery batch sizes due to more efficientand less costly purchasing orders. Information is especially valu-able when there is enough capacity to react to information in thecase of high demand uncertainty. A simulation analysis of the beergame (Van Ackere et al., 1993) has shown that complete informa-tion sharing between supply chain participants can reduce costs(including inventory costs) to as low as 40% of the original beergame setup.

The level of vertical integration, determined by the level ofoutsourcing and/or the ratio of materials purchased from themarket instead of making in house, can have essential impact onthe ratio and level of inventory types (Kakabadse and Kakabadse,2002). Both tendencies can result in higher input material ratios,as companies have more complex and higher level of purchasing.Also, they can reduce WIP, since many parts are bought instead ofproducing them.

We have to note here, that many companies are part ofmultinational company networks. As such they can use theirnetworks' distribution centres, share up to date confidentialinformation and rely on each other in making various parts ofthe same products. This network level optimisation can havesignificant impact on inventories at subsidiaries, but they arebeyond the scope of our analysis. Since these networks can beconsidered as internal supply chains for companies, their impactwill anyway appear in the results.

Finally, the level of globalisation of purchasing and distributioncan play an important role in managing inventories. In general, thelonger the distance, the longer the lead time necessary to receiveor ship freights. As shown by classical inventory models and byempirical studies (Lieberman et al., 1999; Rumyantsev andNetessine, 2007), the longer the lead times, the higher theinventories. This effect on inventories is quite typical in globalsupply chains in which suppliers and customers are located faraway from each other (Golini and Kalchschmidt, 2011a, 2011b;Kouvelis and Gutierrez, 1997). Moreover, when dealing with globalsourcing, companies often protect themselves from supply varia-bility and disruption through higher inventories (Stratton andWarburton, 2006).

2.4. Business strategy

The final group of factors identified in the literature encompassbusiness strategy. There are several product-related strategies thatcompanies can follow to win orders from potential customers.

K. Demeter, R. Golini / Int. J. Production Economics ∎ (∎∎∎∎) ∎∎∎–∎∎∎ 3





The basic strategies most often used in operations strategyliterature are price/cost, quality, time and flexibility (Dangayachand Deshmukh, 2001; Flynn et al., 1999; Kim and Arnold, 1996;Spring and Boaden, 1997). These strategies heavily affect opera-tions decisions and inventory configurations.

Price-based competition, for example, requires low-cost opera-tions. Mass production can serve such a strategy because unit costscan be lower due to the more efficient use of resources. Massproduction can reduce WIP inventories (Demeter and Matyusz,2011) by providing assembly lines between working stages, whichdo not allow extra inventories. These companies also make effortsto push inventory holding responsibilities to their suppliers if theycan reduce their costs (Fazel, 1997) and accept output shortages inan effort to keep low finished goods inventories. By consequence,these companies strive to reduce the level of each inventory typeto reduce overall inventory holding costs.

Wider product range causes higher production switching timeand higher inventories (Cachon and Olivares, 2010). These com-panies must usually keep higher finished goods inventories toserve any product a customer wants, higher WIP to produce largevariety and higher input inventories to serve the needs of variousproducts. In general, a higher product range requires higher levelsof each type of inventory.

A focus on fast deliveries can impact inventories differently.Use of inventory reduces the response time of systems, so keepinghigher levels of inventories can help achieve fast deliveries. On theother hand, lean improvements can eliminate or reduce dramati-cally the level of inventories by accelerating flows, and thus reducethe overall level of inventories. The level of inventories is one ofthe hot issues in lean vs. agile systems (Bruce et al., 2004; Caglianoet al., 2004; Goldsby et al., 2006; Mason-Jones et al., 2000; Nayloret al., 1999).

Flexibility has several dimensions and the tools that serve it candiffer (Gerwin, 1993). The impact of flexibility on inventories,however, is ambiguous. According to Cachon and Olivares: “Produc-tion flexibility allows a firm to track production more closely to sales,thereby yielding a lower optimal level of safety stock for a firm” (Cachonand Olivares, 2010). However, closer tracking of production to sales isnot enough for flexibility. Flexibility usually also requires keepingbuffer resources – including inventories – in some points to be moreresponsive to changes in customer demand (Newman et al., 1993).

Finally, more innovative products have a higher risk of obsoles-cence and higher profit margins (Rumyantsev and Netessine,2007). Products in the early stages of the lifecycle (or innovativeproducts) usually have higher margins and less predictabledemand (Fisher, 1997). The cost of items leftover in any type ofinventory due to product change can be high (Callioni et al., 2005).As soon as sales rise, demand becomes more stable and lessfragmented, therefore lowering inventories (Cachon and Olivares,2010).

3. Research questions

As seen in the literature, there are different inventory types andmany factors that can affect them. It is too complex to consider allof the possible relationships among the inventories and theirdrivers simultaneously. Therefore, our aim is to identify (1) typicalratios (i.e., configurations) of inventory types and (2) the keydrivers behind these typical ratios.

Moreover, examining inventory configurations rather thanseparate inventory types provides companies with a frameworkin which they can easily identify themselves and obtain a deeperand strategic thinking of the dynamics and relationships betweeninventory types. Thus, instead of taking into account several minoror major influencing factors, the company can focus on a few real

drivers behind their specific inventory configurations. Further-more, knowing other configurations, the company can have aclearer target of where it wants to go.

As a matter of fact, companies and researchers strive to find thebest allocation of inventories in order to reach some objectives(e.g., minimise costs, avoid stock-outs) and the optimal solutioncan change from company to company. This means that not all ofthe companies can work in a complete low-inventory regime, butthey must keep some inventories in recurrent situations wherenecessary (Jones and Riley, 1985; Naylor et al., 1999). Therefore, weexpect that analysis of the empirical data will reveal recurrentconfigurations. Thus, our first research question is:

RQ1: What are the main configurations of inventory levelsof input (i.e., raw materials and components), WIP and output(i.e., finished goods)?

Although the existing literature revealed some minor changesin the ratios of input, WIP and output inventories (Chen et al.,2005; Chikan, 1996), over time their levels, rather than their ratios,experienced radical reductions (Chen et al., 2005; Lieberman andDemeester, 1999) due to lean efforts made in the late 1980s andearly 1990s. Nevertheless, there is no literature on the level ofinventories from the last decade. Thus, our second researchquestion addresses this issue:

RQ2: Are inventory configurations stable during time?Finally, inventory configurations are affected by different inter-

nal or external factors. Based on the literature, production char-acteristics (Demeter and Matyusz, 2011), supply chain features(Beamon, 1998; D'Aveni and Ravenscraft, 1994; Golini andKalchschmidt, 2011a, 2011b; Rumyantsev and Netessine, 2007)and strategic objectives (Cachon and Olivares, 2010) all impactvarious inventory types, so they must also account for inventoryconfigurations. Thus, our third research question is:

RQ3: What are the factors explaining these different inven-tory configurations?

4. Methodology

To pursue our research goals, we used data collected from thelast three editions of the global network of the InternationalManufacturing Strategy Survey (IMSS), 2001 (IMSS 3), 2005 (IMSS4) and 2009 (IMSS 5). This project, originally launched by theLondon Business School and Chalmers University of Technology,studies manufacturing strategies within the assembly industry(ISIC 28-35 classification) through a common questionnaire admi-nistered simultaneously in many countries by local researchgroups (Lindberg et al., 1998). The main research goal of theproject is to investigate the relationships among strategic prio-rities, manufacturing practices, improvement programs, perfor-mance and contingent variables.

Companies are usually selected from local databases and theoperations, production or plant manager is contacted regarding hisor her interest in the research. If the respondent agrees, thequestionnaire is sent out. When necessary, a reminder is sentafter several weeks. Questionnaires that are returned are con-trolled for missing data, which are handled case by case, usually bycontacting the company again. Finally, all data are grouped into aunique database, which is further controlled by the projectcoordinator and distributed to all partners. For the differenteditions of the survey, the response has always been higher than17%: 33% in 2001, 22% in 2005 and 17% in 20091. The structure of

1 The response rate is calculated as the ratio between the total number ofanswers collected and the number of questionnaires sent to companies.






the questionnaire has remained the same over time. The firstsection of the questionnaire is related to the business unit(gathering general information, such as company size, industry,production network configuration, competitive strategy and busi-ness performance), whereas the other sections refer to the plant(focusing on manufacturing strategies, practices and perfor-mance). Although the structure of the questionnaire has remainedthe same with every edition, some questions have been updated orremoved and new questions have been added by the design team,which is composed of a pool of international researchers, to avoidresearchers' country biases (Van de Vijver and Leung, 1997).

We excluded cases that did not provide the necessary informa-tion for the purposes of this study. We also excluded companiesthat provided answers classified as outliers2. Therefore, the overallsample used in this study consists of 331 firms from 2001, 518firms from 2005 and 495 firms from 2009.

The distribution of the four samples in terms of country,industry and size is shown in Tables 1–3.

The three samples are similar in terms of size, industry andcountry. However, they do not consist of the same firms (only avery limited number of firms participated in two or more editionsof the survey). Therefore, among the longitudinal studies, we canconsider this as a trend analysis, in which the population issampled and analysed at different moments of time. This type ofstudy is useful when the focus is on aggregate, rather thanindividual, change over time (Taris, 2000).

In this study, we are interested in the general behaviour of thesample, namely, inventory configurations and the key factorsbehind them. Therefore, our model is not purely longitudinalbecause we are interested in not only those firms that participatedin all three editions but in the results of the three completesamples (Menard, 2002). The three samples are drawn from thesame set of industries and with the same guidelines across thevarious editions.

In terms of measures, the levels of inventories are askedseparately for material and components, work-in-process andfinished goods in terms of days of production (on average) carriedin the different inventories.

Based on these three variables, a two-step cluster analysis wasperformed. First, hierarchical cluster analysis, based on squaredEuclidean distance and the Ward method, was used to identify themost suitable number of clusters and the cluster centroids. Thehierarchical cluster analysis suggested four clusters. Next, the K-means clustering algorithm was used to iteratively assign eachfirm to a cluster (Ketchen and Shook, 1996).

First, we checked whether the industry (ISIC code) and thecompany size affected our clusters. We found that the clusters wereevenly represented in each sub-industry, even though companiesfrom ISIC 29–30 (manufacture of machinery, equipment, office,accounting and computing machinery) are over-represented incluster 4, while companies from ISIC 34–35 (manufacture of vehiclesand transport equipment) are overrepresented in cluster 5 (seeAppendix A). This last result is not surprising and shows that ourdata are coherent with the literature. As a matter of fact, cluster5 represents low-inventory companies (see the “Results” section)and ISIC 34-35 includes the automotive industry, which is known forearly diffusion of lean management. On the other hand, thedistribution of the clusters is not dependent on the size (seeAppendix A). The fact that there is no relationship to size can beexplained by the fact that we measure the level of inventories interms of days of production and not in absolute terms.

Moreover, as we will show in the “Results” section of the paper,the clusters tend to change their relative composition over time(that is, along the different IMSS versions), highlighting interestingtrends.

Next, we checked differences between clusters on the variablesdescribed in Table 4. We picked the items available in the threeeditions of the survey (except for vertical integration and theposition of the decoupling point that were not available forIMSS 3) to have at least one item for each group of factorsdescribed in the literature review (market, internal operations,

Table 1Country distribution over the four samples (in brackets the column percentage).

IMSS 3 (2001) IMSS 4 (2005) IMSS 5 (2009)

Hungary 44 (13%) 47 (9%) 58 (12%)Italy 38 (11%) 30 (6%) 33 (7%)The Netherlands 7 (2%) 47 (9%) 41 (8%)Brazil 31 (9%) 15 (3%) 32 (6%)Sweden 10 (3%) 65 (13%) 0 (0%)China 3 (1%) 29 (6%) 38 (8%)Germany 21 (6%) 16 (3%) 31 (6%)Denmark 26 (8%) 25 (5%) 14 (3%)Belgium 14 (4%) 23 (4%) 25 (5%)USA 6 (2%) 25 (5%) 30 (6%)United Kingdom 32 (10%) 12 (2%) 11 (2%)Spain 15 (5%) 0 (0%) 31 (6%)Argentina 12 (4%) 28 (5%) 0 (0%)Ireland 23 (7%) 11 (2%) 4 (1%)Australia 26 (8%) 10 (2%) 0 (0%)Estonia 0 (0%) 16 (3%) 16 (3%)Canada 0 (0%) 15 (3%) 15 (3%)Other countriesa 23 (7%) 104 (20%) 116 (23%)

Total 331 (100%) 518 (100%) 495 (100%)

a Others: Turkey, Taiwan, Croatia, Korea, New Zealand, Switzerland, Venezuela,Japan, Romania, Israel, Portugal, Mexico, Greece, Norway.

Table 2Industry distribution over the three samples (in brackets the column percentage).

ISIC Codes IMSS 3 (2001) IMSS 4 (2005) IMSS 5 (2009)

28 114 (34%) 203 (39%) 178 (36%)29-30 78 (24%) 110 (21%) 128 (26%)31-32 88 (27%) 103 (20%) 102 (21%)33 27 (8%) 19 (4%) 28 (6%)34-35 24 (7%) 83 (16%) 59 (12%)

Total 331 (100%) 518 (100%) 495 (100%)

ISIC Code (Rev. 3.1): 28: Manufacture of fabricated metal products, exceptmachinery and equipment; 29: Manufacture of machinery and equipment notclassified elsewhere; 30: Manufacture of office, accounting, and computingmachinery; 31: Manufacture of electrical machinery and apparatus not classifiedelsewhere; 32: Manufacture of radio, television, and communication equipmentand apparatus; 33: Manufacture of medical, precision, and optical instruments,watches and clocks; 34: Manufacture of motor vehicles, trailers, and semi-trailers;35: Manufacture of other transport equipment.

Table 3Size distribution over the three samples (in brackets the column percentage).

IMSS 3 (2001) IMSS 4 (2005) IMSS 5 (2009)

Small 156 (47%) 285 (55%) 248 (50%)Medium 79 (24%) 99 (19%) 95 (19%)Large 96 (29%) 134 (26%) 152 (31%)

Total 331 (100%) 518 (100%) 495 (100%)

Small: less than 250 employees, Medium: 251–500 employees, Large: over 501employees.

2 Omitted companies: Days of inventories 4¼70; Company sizeo10 employ-ees, missing ISIC code or company size.






supply chain and business strategy). Three other variables, namely,company size, the ISIC code (i.e., industry) and the IMSS version,were used as control variables.

First of all we checked that the factors were independent. Weonly found a correlation higher than .5 between local sourcing andlocal distribution. Next we performed a confirmatory factor analy-sis keeping all the variables alone and putting together only LocalSourcing and Local Distribution into a unique factor. The analysisholds (χ2/df¼3.74; RMSEA¼ .045; CFI¼ .984; NFI¼ .98) witnessingthat we can consider the variables independently. Given therelatively low correlation between local sourcing and local distribu-tion and the theoretical difference in the two concepts, we also leftthe two variables separated.

We checked the significance of the results by means ofMANOVA analysis (with LSD test3 to check differences betweencouples of groups) supported by non-parametric tests (Kruskal–Wallis) to avoid issues related to non-normal distributions. MAN-OVA analysis takes into account control variables (i.e., industry,size and IMSS version) that are not only related to the clusters, butcan affect the variables analysed. For instance, firm size might berelated to the power and the position of a company in the supplychain (Benton and Maloni, 2005; Choi and Hartley, 1996).

5. Results

In this section we present the results of our analysis. First, somepreliminary analyses are shown. Next, we present results andanswer our three research questions. All of the results arediscussed in the context of the existing literature presented inthe “Discussion” section.

Before examining our research questions, we performed pre-liminary analyses to check the behaviour of our data. Table 5reports the average values of the different types of inventories inthe three study periods. We found a decreasing trend from 2001 to2009 (even though this trend was more evident between 2001 and2005) and the inventory level was higher than the other levels.This is in line with previously cited works in similar industries(e.g., Chen et al., 2005) supporting the validity of our data.

5.1. Configurations by inventory type ratios (RQ1)

The key results of the two-step cluster analysis are representedin Table 6. We can see that several, but not all, of the possiblecombinations of high or low inventory are represented. The twoextremes, in terms of stock levels, are the high–high–high (HHH)group with high inventories of each type and, on the opposite side,the low–low–low (LLL) group with low inventories of each type. Inbetween, we find the HHL, LLH and LHL configurations. Looking atthe average values, we can empirically define “low” as less than 23days and “high” as more than 34 days of production carried ininventory. Interestingly, there are no figures between these twoupper-low and downer-high thresholds. Looking at the sampleaverage, we notice, as previously highlighted (see Table 5), ahigher level of materials and components compared to the othertwo types of inventory.

5.2. Change in configurations over time (RQ2)

To check if the clusters are stable over time (i.e., they arerepresented in every edition of the survey), we verified in Table 7the number of companies belonging to each cluster for everyedition of the survey. We can see that in every edition, there are arelevant number of companies for each cluster. Interestingly, theLLL cluster is the most populated, followed by HLL, LLH, LHL and,lastly, HHH. However, the relative composition changed over time.As shown in Fig. 2, the share of LLL increased mainly at theexpense of HHH. HHL and LLH slightly decreased while LLHremained quite stable.

5.3. Factors behind configurations (RQ3)

To answer our third research question, we checked if significantdifferences among clusters existed on the set of variables selectedon a literature basis (Table 4). The MANOVA analysis is controlledby industry (i.e., ISIC code), size (logarithm of the number ofemployees) and time (i.e., IMSS version). As a robustness checkand to overcome possible non-normality issues, we calculated theKruskal–Wallis statistic without any control variable. The two testsproduced similar results (Table 8), demonstrating that controlvariables are not very relevant. As some missing values weremissing, the analysis shows a slightly different number of casesconsidered per variable. However, except for the position in the

Table 4Explanatory variables included in the analysis.

Group of factors Variable Description

Market factors Material cost Percentage of product cost that is direct represented by materials/parts/componentsInternal operations Type of production Calculated continuous variable that ranges from 0 (one of a kind production) to 1 (mass production) (see Appendix A for

details)Position of the decouplingpoint

Calculated variable that ranges from 0 (engineer to order) to 1 (make to stock) (see Appendix A for details)

Supply chaincharacteristics

Position in the supplychain

Calculated variable that ranges from 1 (upstream, closer to raw material suppliers) to �1 (downstream, closer to finalcustomers) (see Appendix A for details)

Vertical integration Calculated variable that ranges from 0 (low level of vertical integration) to 2 (high level of vertical integration)Information sharing withcustomers

Mean of two 1–5 Likert scale (1: no adoption; 5: very high adoption) items “Share inventory level information withcustomers” and “Share production planning and demand forecast information with customers” that are highlycorrelated (Pearson corr.:.463; sig.: .000)

Local sourcing Percentage of suppliers inside the continent where the plant is basedLocal distribution Percentage of customers inside the continent where the plant is based

Business strategy Price as order winner 1-5 Likert scale for the importance of price as an order winnerDeliveries as orderwinner

1–5 Likert scale for the importance of faster and more dependable deliveries as an order winner

Range as order winner 1–5 Likert scale for the importance of product range as an order winnerFlexibility as order winner 1–5 Likert scale for the importance of order size flexibility as an order winnerNew products as orderwinner

1–5 Likert scale for the importance of offering new products innovative products as an order winner

3 LSD uses t-tests to perform all pairwise comparisons between group means.






supply chain and vertical integration that were unavailable in IMSS3 (about one-third of the initial sample), all of the other variableswere calculated on more than 90% of the original sample.

Looking at Table 8, we see that several variables are significantat 1%. These are: position of the decoupling point, type ofproduction, information sharing with customers, new products,range and material cost, price and vertical integration. Position inthe supply chain and local sourcing are significant at 5%. Finally,flexibility, local distribution and deliveries do not differ amongclusters. Therefore, we did not include these last variables in thefollowing analyses.

In the last analysis, we checked the pairwise differences amongclusters (Table 9). We also labelled each cluster accordingly. Theresults by cluster are reported below:

� HLL – upstream specialty producers: they are usually engineer/make-to-order and oriented to a one of a kind/batch production.They are positioned more upstream in the supply chain than anyother clusters and use global sourcing more intensively, most

likely to buy special materials and components not availablelocally or because they are internal suppliers within multina-tional companies obtaining inputs from other subsidiaries orpartners contracted by global buyers. They pay relatively lessattention to product range and new product introduction.

� HHH – integrated low communicators: the key feature of thisgroup is the high level of vertical integration and the relativelyweak information-sharing with customers. However, theyassemble-to-order or make-to-stock, thus they require theaforementioned information. They mainly use batch produc-tion and relatively underemphasise price.

� LLH – innovative mass producers: although their productiontype leans toward mass production and they make many oftheir products to stock; they consider new products andproduct ranges more important order winners than any othergroup. They have relatively intense information sharing withcustomers, most likely to be able to handle the level of FGinventories and both overstock and understock situations.

� LHL – customised high value-added producers: they are theclosest to the engineer- to-order decoupling point and one-of-a-kind production. Although the ratio of material costs is thelowest in this group, price is the least important to them. Thissuggests that the value they add to a product is relatively high,allowing them to obtain relatively high profit margins.

� LLL – focused lean producers: the key characteristics of thisgroup are their intense information sharing with customers, alow level of vertical integration and the relatively high focus onprice. Furthermore, although they tend towards make-to-orderas the HHH – upstream specialty producers and customised highvalue-added producers, their production process is closer tointegrated low communicators and innovative mass pro-ducers. That is how they let the decoupling point penetrate

Table 5Average days of production carried in the inventory by inventory type and year.

VersionIMSS 3 (2001) IMSS 4 (2005) IMSS 5 (2009)

Input

WIP

Finished goods

Table 6The 5 clusters identified (“H” stands for “High” and “L” for “Low”).

Cluster Input WIP Finished goods

Total amount of inventories

Number of companies

HLL 62.6 282

HHH 131.7 67

LLH 64.4 167

LHL 76.4 127

LLL 21.4 701

Sample Average 21.2 13.0 11.9 46.1

Table 7Distribution of companies by cluster and year.

IMSS 3 (2001) IMSS 4 (2005) IMSS 5 (2009) Total

HLL 74 109 99 282HHH 28 27 12 67LLH 50 64 53 167LHL 38 41 48 127LLL 141 277 283 701

Total 331 518 495 1344






the organisation, but they manage the production processefficiently.

6. Discussion

Our study supports previous findings (e.g., Rajagopalan andMalhotra, 2001) on the general declining trend of inventories,which began in the mid-1970s (Chikan, 1996). The total number ofproduction days in our sample covered by inventories was 54 daysin 2001, 46 days in 2005 and 41 days in 2009. Our study is uniquein that while previous studies used macro level (national) data (e.g., Chikan, 1996; Rajagopalan and Malhotra, 2001) or focused theirattention on the changes in one inventory type, industry orcountry4, our results stem from three large scale international,company-based databases from assembly industries (including theautomotive industry).

While the overall inventory levels decreased, the ratio of input– WIP – output inventories stayed more or less constant in oursamples. The ratio of raw materials varied between 44% and 47%,WIP between 27% and 29% and finished goods between 25% and27%. The relative stability of ratios corresponds to the finding ofChikan (1996), who examined the time period of 1967–1989.Chikán notes, however, that the level of finished goods inventoriesdoes not seem to follow the decreasing trend of WIP and raw

material inventories. Rajagopalan and Malhotra (2001) examined asomewhat wider period between 1961 and 1994, and similar toChikán, they detected WIP reduction ratios compared to finishedgoods in some industries, such as machinery or fabricated metalproducts (both included in our sample) worldwide and for theoverall manufacturing industry in the US economy. Chen et al.(2005) also found that the level of finished goods inventories didnot decline, but the level of WIP fell dramatically between 1981and 2000. The weak trend detected by Chikan (1996) and the morevisible trend identified by Rajagopalan and Malhotra (2001) andChen et al. (2005) indicates that in late 1980s and early 1990s,significant changes began taking place in the manufacturingindustry (think about MRP, JIT, lead time reduction efforts),resulting in reduced WIP ratios. However, these changes likelypeaked at the end of 1990s because the ratios stayed the samebetween 2001 and 2009. An additional explanation for thisstability is that efforts originally started in production spread tothe supply chain (Naylor et al., 1999) and caused similar changes ininventories on both the supply and demand sides. Thus, reductionopportunities in each inventory type became similar.

Furthermore, our study shows that there were some rigidinventory configurations during the eight-year period (RQ1 andRQ2). Even if the overall inventory levels fell, the five inventoryconfigurations existed in each time spot. Even the ratio of classesseemed to be relatively stable, with just one exception. Thenumber of companies in the HHH (high inventories in each type)class fell and the number of companies in the LLL (low inventoriesin each type) proportionately increased. That is, there are an

Fig. 2. Relative cluster composition in the different years.

Table 8ANOVA and Kruskal–Wallis test for differences among clusters. Variables are sorted by F-value.

Variable F-value MANOVA Sig. Kruskal–Wallis Sig. Number of observations

Position of the decoupling point 21.775 .000 .000 1294Type of production 10.431 .000 .000 1325Information sharing with customers 8.484 .000 .007 1259New products as order winner 6.858 .000 .000 1307Range as order winner 5.167 .000 .002 1298Material cost 4.617 .001 .000 1219Price as order winner 3.885 .004 .001 1326Vertical integration 3.492 .009 .000 940a

Position in the supply chain 2.874 .022 .088 940a

Local sourcing 2.605 .034 .013 1286Flexibility as order winner 1.278 .277 .361 1306Local distribution 1.127 .342 .575 1282Deliveries as order winner .239 .916 .893 1305

a Data not available for IMSS 3.

4 WIP in Japan in automotive industry: Lieberman and Demeester (1999);finished goods in US in automotive industry: Cachon and Olivares (2010).






increasing number of companies that consider inventory as apossible target for reducing costs to remain competitive. Reducingthe level of vertical integration (which is high in the HHH class) byoutsourcing, or increasing the ratio of purchased materials (bothhave become general trends in recent decades, see Kakabadse andKakabadse, 2002) can make processes and products less complexin these companies. Although D'Aveni and Ravenscraft (1994)argue that higher vertical integration makes inventory controleasier due to better communication, the change in ICT in the lastdecade has made external communication as easy as internalcommunication. Actually, the HHH group seems to be a poorcommunicator, at least toward its customers, which can alsoaccount for higher inventory levels. Consequently, if HHH compa-nies start to make efforts to share information or decrease thelevel of vertical integration, the ratio of inventory types remainsthe same. However, the companies may change class and becomeLLL group members. There is, however, another important differ-ence between HHH and LLL companies. The HHH group has thehighest profit margins, which likely stem from its more complexproducts and production. Increased outsourcing and reducedproduction complexity can lead to dramatic changes in thiscompetitive advantage, rendering price more important. Alto-gether, changing clusters and reducing inventories seems possible,but involves several strategic decisions.

General inventory reduction will not produce class change inthe case of other classes. In other classes, companies just becomebetter within the same class. The most important factors that tiethese companies to their specific class (RQ3) are the position of thedecoupling point, the type of production, the level of informationsharing with partners and the focus on some strategic goals, suchas new products, product variety or price/cost.

Producing a one-of-a-kind product (production type) andengineer-to-order (decoupling point) leads to high WIP inven-tories (Lieberman et al., 1999) due to the complexity of productsand production processes. However, purchasing materials afterproduct design acceptance and delivering products right afterthey are completed, creates low input and output inventories.Therefore, companies with these characteristics belong to theLHL group.

Mass production (production type) makes it easier to controlinput and WIP inventories, according to Demeter and Matyusz

(2011). Due to relatively stable demand, these companies can useframe contracts and withdraw the amounts of input needed fromsuppliers (delivered in many cases directly to the assembly line)(Fazel, 1997), or held in consignment stock. Furthermore, the clearroutes of materials prevent WIP inventory from piling up. Thesecompanies usually make to stock (decoupling point) (whichresults in LLH classification). Furthermore, Demeter and Matyusz(2011) mentioned that mass producers focus more on prices andwhat they can easily support with low costs due to the highvolumes they produce. We did not find a higher price focus in thisgroup statistically, but it received the second-highest grade afterthe LLL group.

The higher level of global sourcing and a relatively highportion of small batch production, induces companies in theHLL group to keep high raw material inventories. Global sourcingimplies longer lead times (Golini and Kalchschmidt, 2011a, 2011b;Kouvelis and Gutierrez, 1997) combined with higher deliveryrisks (Stratton and Warburton, 2006), increasing the amount ofsafety stocks. In contrast, producing in small batches requireskeeping many different types of components in stock. Produc-tion processes; however, are less complex and the variety ofproducts are lower than in the LHL group. Thus, WIP inventoriescan be lower.

There are some factors that cannot be assigned to any specificconfiguration. The ratio of local distribution, focus on deliveriesand flexibility did not help differentiate clusters. This result maystem from the fact that there was not enough diversity in theseaspects in our sample. It does not necessarily mean, however, thatthey are not important factors in inventory management, only thatthey require further analyses.

Altogether, we found that the two most important character-istics of clusters are the decoupling point and the type of produc-tion. Using the averages of our calculated variables, Fig. 3 canbe drawn.

We can see that four configurations (LHL, HLL, LLH, HHH) aremore or less located in the diagonal. Three of these configurationshave one type of inventory higher than other configurations, butthe overall inventory levels are very similar (see Table 6). LHLcompanies keep inventories primarily to decouple productionsteps. HLL companies have inventories to buffer in case of suddenorders and/or to decouple due to high distance suppliers. In LLH

Table 9Estimated means by cluster from the MANOVA analysis. After the variable, the clusters that are significantly different are reported.

HLL (1) HHH (2) LLH (3) LHL (4) LLL (5)

Position of the decoupling point .47 .58 .65 .44 .48Clusters significantly different 2, 3 1, 3, 4, 5 1, 2, 4, 5 2, 3 2,3Type of production .44 .52 .56 .41 .53Clusters significantly different 2, 3, 5 1,4 1,4 2, 3, 5 1,4Information sharing w/t customers 2.96 2.58 3.10 2.97 3.25Clusters significantly different 2, 5 1, 3, 4, 5 2 2, 5 1,2,4New products as order winner 3.07 3.22 3.59 3.14 3.13Clusters significantly different 3 3 1,2,4,5 3 3Range as order winner 3.25 3.43 3.67 3.35 3.30Clusters significantly different 3 1,4,5 3 3Material cost 50.93 49.38 50.89 41.97 51.04Clusters significantly different 4 4 4 1, 2, 3, 5 4Price as order winner 3.80 3.63 3.84 3.61 3.94Clusters significantly different 5 5 2, 4Vertical integration 1.26 1.39 1.29 1.30 1.23Clusters significantly different 2 1, 5 5 2, 3Position in the supply chain .17 � .01 .08 .07 .14Clusters significantly different 2, 3, 4 1, 5 1 1 2Local sourcing 51.56 57.77 57.68 61.60 56.81Clusters significantly different 3, 4, 5 1 1 1






companies, inventories serve as buffer and decoupling. Becausedecoupling plays a role in each of these three configurations, it isunderstandable that the decoupling point exerts such a crucialimpact on the level of inventories.

LLL and HHH are much different. LLL has three times lessinventories than the previous three groups. A deeper look into thedata revealed that these companies share information with theircustomers more than other companies, thus eliminating inven-tories (Cachon and Fisher, 2000). These companies also use kanbansystems more often. These results indicate that LLL companiesmade many efforts towards lean management and that theybenefit from these efforts through reduced inventories. BecauseLLL is the only group that is not on the diagonal, we can argue thatlean management is a paradigm shift not only in operationsmanagement (Schmenner, 2001; Schmenner and Swink, 1998)but in inventory management. Lean management causes suchsignificant changes that it can divert companies from the diagonalposition and reduces inventories considerably. Lean manage-ment impacts positions similarly in the product-process matrix,diverting companies from the diagonal (Ahmad and Schroeder,2002).

The HHH group has twice as much inventory than the otherthree groups on the diagonal. Higher vertical integration makesHHH internal processes as complex as those of LHL companies. Theresulting long lead times make it important to have buffers in rawmaterial and WIP in some points to remain competitive in deliverytimes. The same holds for finished goods. The likely key purpose ofHHH companies is buffering against uncertainties due to long leadtimes. HHH companies can do this due to their higher profitmargins.

The main message of the figure is that companies need aparadigm shift toward swift even flow (lean) to reduce each oftheir inventory types dramatically.

7. Conclusions and future development

Inventory reduction has been a key objective of companiesin different industries. However, a one-way strategy towardszero-inventory can be inapplicable or too general. This isbecause companies must be aware of the factors behind theirinventory configuration, defined in this paper as the amount ofinput, WIP and FG inventory. As such, in this paper we

empirically tested the existence of stable inventory configura-tions in the manufacturing industry and identified the driversbehind them.

Our research shows that the actual configurations that com-panies adopt, in terms of levels of each type of inventory, as wellas the factors contributing to their adoption, were stable andconsistent from 2001 to 2009. We found five permanent config-urations. Although the number of companies per configurationchanged slightly over time, each configuration existed in eachexamined time slot. Configurations reflect the production char-acteristics of a company, particularly in terms of decoupling pointposition and type of production. The results show that produc-tion features are important explanatory factors behind inven-tory levels and ratios, and more important than supply chaincharacteristics.

Among supply chain characteristics, sharing informationwith customers can improve inventory position, which supportsthe idea that information can substitute inventories to someextent.

Results are interesting not only for researchers but also forpractitioners, who will not only better understand their inventoryconfigurations, but will also have a basis for the deployment ofeffective inventory reduction strategies.

Companies willing to reduce their inventory levels mustsignificantly change their production systems and/or their decou-pling points. Minor changes, however, can be achieved withinconfigurations. Future studies should focus on discovering thislatter issue more deeply: what are the practices companies use toachieve the best inventory performance within their own inven-tory configurations? Are there configuration-specific strategiesbehind these configurations?

An important limitation of this study lies in the fact that thesamples varied over time. The countries and companies involvedwere different in each database, which increased the noise in thedata. Further studies may limit analysis of countries involved ineach sample, reducing part of the noise.

Our results are valid only for assembly industries or discreteproduction systems. Because process and assembly industries havevery different features (Meredith, 1992) and because, according toHill and Hill (2009), process choice overwhelmingly impactsoperations management decisions and the level of inventoriestypes.

Finally, we focused on the plant level. However, some plantscan belong to larger multinational networks in which other issuesand factors may exist related to supply chain design as well ascentralisation and decentralisation of purchasing, manufacturingand distribution activities.

Appendix A

A1. Calculated variables

Type of productionIn the questionnaire we ask the percentage of production that

is one of a kind, batch or mass production.Applying the following formula

Production Type¼ ðOne of a kind� 1ÞþðBatch� 2ÞþðMass� 3Þ�100300�100

we could calculate in a [0,1] range the type of production rangingfrom one of a kind (0) to mass (1) production.

Supply chain position and vertical integrationAdapted from: Szász and Demeter (2011)Upstream supply chain position measure is developed based on

the following logic:

1.0

Massproduction

1.0

Prod

uctio

n ty

pe LLLHHH

LLH

LHL

HLLBatch

0.42 0.63

Oneof akind

0.55

0.39

ETO MTO ATO MTSDecoupling point

Fig. 3. Key drivers of inventory configurations. (The size of circles indicates theapproximate level of inventories in days of production).






After standardizing, the upstream position measure is calcu-lated by the following formula:

UpstreamPos¼∑3i ¼ 1u1ð%Þ � disti�100

300�100

Analogously, a downstream supply chain position measure isdeveloped based on the following logic.

Thus values characterising downstream position of manufac-turing companies uniformly fall into the [0,1] interval.

DownstreamPos¼∑4i ¼ 1d1ð%Þ � dist′i�100

400�100

Combining the measurement of upstream and downstream posi-tions we can calculate a measure of overall supply chain position(ScPos) for each manufacturing company. ScPos¼0 means thatdownstream and upstream positions are similar. ScPos40 meansan upstream declination, while ScPoso0 means a downstreamdeclination.

ScPos¼ UpstreamPos�DownstreamPos

Summing the two separate supply chain position measures offers agood approximation of the degree of vertical integration of amanufacturing company. The higher the value of VertInt, the moremanufacturing processes the company embraces.

VertInt ¼UpstreamPosþDownstreamPos

A2. Distribution of the clusters by industry

ISIC Total

28 29-30 31-32 33 34-35

Cluster numberHLL (1) 97 63 77 17 28 282HHH (2) 23 16 19 6 3 67LLH (3) 75 38 35 13 6 167LHL (4) 38 51 28 2 8 127LLL (5) 262 148 134 36 121 701

Total 495 316 293 74 166 1344

ISIC Code (Rev. 3.1): 28: Manufacture of fabricated metalproducts, except machinery and equipment; 29: Manufacture ofmachinery and equipment not classified elsewhere; 30: Manufac-ture of office, accounting, and computing machinery; 31: Manu-facture of electrical machinery and apparatus not classifiedelsewhere; 32: Manufacture of radio, television, and communica-tion equipment and apparatus; 33: Manufacture of medical,precision, and optical instruments, watches and clocks; 34:

Manufacture of motor vehicles, trailers, and semi-trailers; 35:Manufacture of other transport equipment.

A3. Distribution of the clusters by size

Size Total

Small Medium Large

Cluster numberHLL (1) 157 57 68 282HHH (2) 38 12 17 67LLH (3) 88 28 51 167LHL (4) 67 26 34 127LLL (5) 339 150 212 701

Total 689 273 382 1344

Small: less than 250 employees, Medium: 251–500 employees,Large: over 501 employees.

References

Ahmad, S., Schroeder, R.G., 2002. Refining the product-process matrix. Int. J. Oper.Prod. Manage. 22 (1), 103–124.

Beamon, B.M., 1998. Supply chain design and analysis: models and methods. Int. J.Prod. Econ. 55 (3), 281–294.

Benton, W., Maloni, M., 2005. The influence of power driven buyer/seller relation-ships on supply chain satisfaction. J. Oper. Manage. 23 (1), 1–22.

Bruce, M., Daly, L., Towers, N., 2004. Lean or agile: a solution for supply chainmanagement in the textiles and clothing industry? Int. J. Oper. Prod. Manage.24, 151–170.

Cachon, G.P., Fisher, M., 2000. Supply chain inventory management and the value ofshared information. Manage. Sci. 46 (8), 1032–1048.

Cachon, G.P., Olivares, M., 2010. Drivers of finished-goods inventory in the USautomobile industry. Manage. Sci. 56 (1), 202.



http://refhub.elsevier.com/S0925-5273(13)00459-3/sbref1
















Cachon, G.P., Randall, T., Schmidt, G.M., 2007. In search of the bullwhip effect.Manuf. Serv. Oper. Manage. 9 (4), 457–479.

Cagliano, R., Caniato, F., Spina, G., 2004. Lean, agile and traditional supply: how dothey impact manufacturing performance? J. Purchasing Supply Manage. 10 (4–5), 151–164.

Callioni, G., de Montgros, X., Slagmulder, R., Van Wassenhove, L.N., Wright, L., 2005.Inventory-driven costs. Harv. Bus. Rev. 83 (3), 135–141.

Capkun, V., Hameri, A.-P., Weiss, L.A., 2009. On the relationship between inventoryand financial performance in manufacturing companies. Int. J. Oper. Prod.Manage. 29 (8), 789–806.

Chen, H., Frank, M.Z., Wu, O.Q., 2005. What actually happened to the inventories ofAmerican companies between 1981 and 2000? Manage. Sci., 1015–1031.

Chikan, A., 1996. The volume and structure of inventories. Int. J. Prod. Econ. 45 (1),1–7.

Chikan, A., 2011. Managers' view of a new inventory paradigm. Int. J. Prod. Econ.133 (1), 54–59.

Chikán, A., 2009. An empirical analysis of managerial approaches to the role ofinventories. Int. J. Prod. Econ. 118 (1), 131–135.

Choi, T.Y., Hartley, J.L., 1996. An exploration of supplier selection practices acrossthe supply chain. J. Oper. Manage. 14 (4), 333–343.

Cusumano, M.A., 1994. The limits of “Lean”. Sloan Manage. Rev. 35. (27–27).D'Aveni, R.A., Ravenscraft, D.J., 1994. Economies of integration versus bureaucracy

costs: does vertical integration improve performance? Acad. Manage. J.,1167–1206.

Dangayach, G.S., Deshmukh, S.G., 2001. Manufacturing strategy. Int. J. Oper. Prod.Manage. 21 (7), 884–932.

Demeter, K., Matyusz, Z., 2011. The impact of lean practices on inventory turnover.Int. J. Prod. Econ. 133 (1), 154–163.

Dennis, D., Meredith, J., 2000. An empirical analysis of process industry transfor-mation systems. Manage. Sci. 46 (8), 1085–1099.

Fazel, F., 1997. A comparative analysis of inventory costs of JIT and EOQ purchasing.Int. J. Phys. Distrib. Logistics Manage. 27 (8), 496–504.

Fisher, M.L., 1997. What is the right supply chain for your product? Harv. Bus. Rev.75, 105–117.

Flynn, B.B., Schroeder, R.G., Flynn, E.J., 1999. World class manufacturing: aninvestigation of Hayes and Wheelwright's foundation. J. Oper. Manage. 17 (3),249–269.

Gerwin, D., 1993. Manufacturing flexibility: a strategic perspective. Manage. Sci. 39(4), 395–410.

Goldsby, T.J., Griffis, S.E., Roath, A.S., 2006. Modeling lean, agile, and leagile supplychain strategies. J. Bus. Logistics 27 (1), 57–80.

Golini, R., Kalchschmidt, M., 2011a. Managing inventories in global sourcingcontexts: a contingency perspective. In: Proceedings of the 20th IPSERAConference, Maastricht, The Netherlands.

Golini, R., Kalchschmidt, M., 2011b. Moderating the impact of global sourcing oninventories through supply chain management. Int. J. Prod. Econ. 133 (1),86–94.

Hayes, R.H., Wheelwright, S.C., 1979. Link manufacturing process and product lifecycles. Harv. Bus. Rev. 57 (1), 133–265.

Hill, T., Hill, A., 2009. Manufacturing strategy: text and cases. Palgrave Macmillan,London.

Jones, T.C., Riley, D.W., 1985. Using inventory for competitive advantage throughsupply chain management. Int. J. Phys. Distrib. Logistics Manage. 15 (5), 16–26.

Juttner, U., Peck, H., Christopher, M., 2003. Supply chain risk management:outlining an agenda for future research. Int. J. Logistics: Res. Appl. 6 (4),197–210.

Kakabadse, A., Kakabadse, N., 2002. Trends in outsourcing: contrasting USA andEurope. Eur. Manage. J. 20 (2), 189–198.

Ketchen, D.J., Shook, C.L., 1996. The application of cluster analysis in strategicmanagement research: an analysis and critique. Strategic Manage. J. 17 (6),441–458.

Kim, J.S., Arnold, P., 1996. Operationalizing manufacturing strategy. Int. J. Oper.Prod. Manage. 16 (12), 45–73.

Kornai, J., 1979. Resource-constrained versus demand-constrained systems. Econ-ometrica: J. Econometric Soc., 801–819.

Kouvelis, P., Gutierrez, G.J., 1997. The newsvendor problem in a global market:optimal centralized and decentralized control policies for a two-marketstochastic inventory system. Manage. Sci., 571–585.

Kraljic, P., 1983. Purchasing must become supply management. Harv. Bus. Rev. 61(5), 109–117.

Lee, H.L., Padmanabhan, V., Whang, S., 1997. Information distortion in a supplychain: the bullwhip effect. Manage. Sci. 43 (4), 546–558.

Lee, H.L., So, K.C., Tang, C.S., 2000. The value of information sharing in a two-levelsupply chain. Manage. Sci. 46 (5), 626–643.

Lieberman, M.B., Asaba, S., 1997. Inventory reduction and productivity growth: acomparison of Japanese and US automotive sectors. Manage. Decision Econ. 18,73–85.

Lieberman, M.B., Demeester, L., 1999. Inventory reduction and productivity growth:linkages in the Japanese automotive industry. Manage. Sci., 466–485.

Lieberman, M.B., Helper, S., Demeester, L., 1999. The empirical determinants ofinventory levels in high-volume manufacturing. Prod. Oper. Manage. 8 (1),44–55.

Lindberg, P., Voss, C., Blackmon, K.L., 1998. International Manufacturing Strategies:Context, Content, and Change. Kluwer Academic Pub., Dordrecht, TheNetherlands.

Mason-Jones, R., Naylor, B., Towill, D.R., 2000. Lean, agile or leagile? Matching yoursupply chain to the marketplace. Int. J. Prod. Res. 38 (17), 4061–4070.

Menard, S.W., 2002. Applied Logistic Regression Analysis. Sage Publications, Inc,Thousand Oaks, USA.

Meredith, J.R., 1992. The Management of Operations: A Conceptual Emphasis,Fourth edition. John Wiley and Sons, New York. (ed.).

Naylor, B.J., Naim, M.M., Berry, D., 1999. Leagility: integrating the lean and agilemanufacturing paradigms in the total supply chain. Int. J. Prod. Econ. 62 (1),107–118.

Newman, W.R., Hanna, M., Maffei, M.J., 1993. Dealing with the uncertainties ofmanufacturing: flexibility, buffers and integration. Int. J. Oper. Prod. Manage. 13(1), 19–34.

Olhager, J., 2003. Strategic positioning of the order penetration point. Int. J. Prod.Econ. 85 (3), 319–329.

Porter, M.E., 1980. Competitive Strategy: Techniques for Analyzing Industries andCompetitors: with a New Introduction. Free Press, New York.

Rajagopalan, S., Malhotra, A., 2001. Have US manufacturing inventories reallydecreased? An empirical study. Manuf. Serv. Oper. Manage. 3 (1), 14–24.

Rumyantsev, S., Netessine, S., 2007. What can be learned from classical inventorymodels? A cross-industry exploratory investigation. Manuf. Serv. Oper. Manage.9 (4), 409–429.

Schmenner, R.W., 2001. Looking ahead by looking back: swift, even flow in thehistory of manufacturing. Prod. Oper. Manage. 10 (1), 87–96.

Schmenner, R.W., Swink, M.L., 1998. On theory in operations management. J. Oper.Manage. 17 (1), 97–113.

Scott, C., Westbrook, R., 1991. New strategic tools for supply chain management. Int.J. Phys. Distrib. Logistics Manage. 21 (1), 23–33.

Slack, N., Chambers, S., Johnston, R., 2007. Operations Management. Prentice Hall,Edinburgh Gate, UK.

Sousa, R., Voss, C.A., 2008. Contingency research in operations managementpractices. J. Oper. Manage. 26 (6), 697–713.

Spring, M., Boaden, R., 1997. ” One more time: how do you win orders?” A criticalreappraisal of the Hill manufacturing strategy framework. Int. J. Oper. Prod.Manage. 17, 757–779.

Stratton, R., Warburton, R.D.H., 2006. Managing the trade-off implications of globalsupply. Int. J. Prod. Econ. 103 (2), 667–679.

Szász, L., Demeter, K., 2011. Supply chain position and servitization efforts ofcompanies in Eastern and Western Europe. Journal of International Businessand Economics 11, 104–112.

Taris, T., 2000. A Primer in longitudinal Data Analysis. Sage Publications Ltd,Thousands Oaks, USA.

Van Ackere, A., Larsen, E.R., Morecroft, J.D.W., 1993. Systems thinking and businessprocess redesign: an application to the beer game. Eur. Manage. J. 11 (4),412–423.

Van de Vijver, F.J.R., Leung, K., 1997. Methods and Data Analysis for Cross-culturalResearch. Sage Publications, Inc, Thousands Oaks, USA.

Ward, P.T., Bickford, D.J., Leong, G.K., 1996. Configurations of manufacturingstrategy, business strategy, environment and structure. J. Manage. 22 (4),597–626.









































































































































Date post:	30-Dec-2016
Category:	Documents
Upload:	ruggero
View:	212 times
Download:	0 times

Inventory configurations and drivers: An international study of assembling industries

Documents