You are on page 1of 12

Inf Syst Front DOI 10.

1007/s10796-010-9245-1

Strategic choices of inter-organizational information systems: A network perspective


Daning Hu & Sherry X. Sun & J. Leon Zhao & Xinlei Zhao

# Springer Science+Business Media, LLC 2010

Abstract As cooperation in a networked manner increases via various inter-organizational information systems (IOISs), it is important to choose appropriate IOISs for different types of organizations in the network environment. In this study, we analyzed customer-supplier relationships among organizations in five industries using social network analysis (SNA) methods and empirical data, aiming to help organizations strategically choose appropriate IOISs. Three types of customer-supplier networks were identified based on the network centralization comparison rate: customer-centric, supplier-centric and balanced networks. Based on the empirical findings in our analysis, we then propose strategies about how to choose appropriate IOISs for the firms in these networks and discuss the pros and cons of the choices. To the best of our knowledge, this is the first empirical research that applied SNA methods to study customer-supplier networks in the context of inter-organizational information systems. Keywords Social network analysis . Inter-organizational information systems . Customer-supplier networks
D. Hu (*) : S. X. Sun : J. L. Zhao Department of Information Systems, City University of Hong Kong, Kowloon, Hong Kong e-mail: daninghu@cityu.edu.hk S. X. Sun e-mail: xiaoysun@cityu.edu.hk J. L. Zhao e-mail: jlzhao@cityu.edu.hk X. Zhao Management Information Systems Department, University of Arizona, Tucson, AZ, USA e-mail: xzhao@email.arizona.edu

1 Introduction Nowadays organizations are more and more connected through various relationships such as strategic alliances and customer-supplier relationships. In this networked environment, recent research on the business value of Information Technology (IT) has raised an important issue on how multiple organizations leverage IT to create and deliver business value. One of the most well known multi-organizational information technologies is the information system that links an organization to its supplier, distribution channels, or customers. Such systems, called inter-organizational information systems (IOISs), are automated systems shared by two or more organizations (Johnston and Michael 1988). They utilize information or process capabilities in multi-organizations to improve their performances or relationships. The wellknown examples of IOISs include American Hospital Supply Corporation s ASAP, United Airlines Apollo reservations system, and American Airlines reservation system SABRE. However, there are considerable variations in the patterns of inter-organizational relationships supported by different IOISs. In general, IOISs are mainly used to manage three categories of inter-organizational issues: 1) inter-organizational transaction processing, 2) customer relation management (CRM), and 3) supply chain management (SCM). IOISs that manage inter-organizational transaction processing are those which process routine transactions such as backorder and financial payment among two or more organizations. One example is SABRE system that is developed to mainly process routine transactions between airlines and travel agents. CRM related IOISs aim to track and manage interactions such as follow up service and customer support between suppliers and

Inf Syst Front

customers. For instance, hospitals often install CRM software provided by suppliers on their systems to keep inventory stocked and thereby allow more efficient customer-supplier interactions. SCM related IOISs manage logistics business activities among organizations such as tracking the shipment of goods. In addition, different IOISs may have different impacts on the participating organizations and industries (Choudhury 1997). In this study, we do not focus on a specific category of inter-organizational issues but rather study IOIS from a more general perspective customer-supplier perspectivesince all three categories of issues involve interactions and business processes between the customer and supplier companies. In order to make the most of the competitive advantages of IOISs, it is important to understand how to choose appropriate IOISs for a specific set of organizations or an industry. In this study, we aim to address this problem from a customer-supplier network perspective. We adopted social network analysis (SNA) to model and analyze a real-world customer-supplier network which consists of 3,406 organizations over a 7-year period (20022008). More specifically, we analyze the topologies of customer-supplier networks in five major industry sectors: IT, retail, finance, services, and health care. The results from this empirical analysis may provide insights for researchers and practitioners to devise effective strategies in choosing appropriate IOISs for organizations. To the best of our knowledge, this is the first empirical research that applied SNA methods to study customer-supplier networks in the context of interorganizational information systems. The remainder of this paper is organized as follows. In the next section, we review IOIS typology and SNA methods used in this study. The third section introduces the dataset for this study. Then we present the experimental results. After that, we discuss the implications of the results on the strategic choices of IOISs for organizations. Finally, we discuss our conclusions and suggest directions for future work.

This typology includes three types of IOISs available for organizations: (1) Electronic Dyads: Each supplier (customer) establishes individual, bilateral transaction links (electronic dyads) with each of a group of customers (suppliers) for a product or service. Figure 1 (Choudhury 1997) shows an example of an electronic dyad IOIS among five suppliers and customers. For instance, Supplier 1 has built individual, bilateral exchange links with Customers 1 and 2. A well-known example of an electronic dyad IOIS is an electronic data interchange (EDI) system. (2) Multilateral IOISs: A multilateral IOIS serves as an intermediary between an organization and its exchange partners. It allows an organization to communicate, interact, and exchange information and resources over a single inter-organization link. There are two sub-types of multilateral IOISs. The first type is often called electronic market and allows multiple suppliers and customers to interact over a single IOIS (Thomas et al. 1987) (Fig. 2a (Choudhury 1997)). One well-known example is Alibaba.com, which is the dominant online B2B electronic market (Zhao et al. 2008) in China. The other type of IOIS is usually developed by a customer (supplier) to facilitate the comparison of offers from all suppliers (customers). This type of IOIS is called a broadcast sales system (Fig. 2b (Choudhury 1997)). (3) Electronic Monopolies: An electronic monopoly IOIS supports a single source relationship for a product or a set of products. In other words, an electronic monopoly IOIS is a special case of the electronic dyad system that represents the only exchange link for the

Supplier 1

Customer 1

Supplier 2

Customer 2

2 Related studies 2.1 A typology of IOISs To answer the research question of how to choose appropriate IOISs for different organizations, one needs to know what types of IOISs are available. Based on several previous studies on classifications of IOISs, Choudhury (1997) proposed a typology of IOISs that supports one of the most common inter-organization relationshipsthe customer-supplier (i.e., buyer-seller) relationship. We also adopted this typology in our study.
Supplier 3 Customer 3

Supplier 4

Customer 4

Supplier 5

Customer 5

Fig. 1 Electronic dyads

Inf Syst Front

a
Supplier 1 Customer 1

Supplier 2 Electronic Market Supplier 3

Customer 2

analyzing patterns of customer-supplier relationships in different industries; 2) discover additional characteristics of customer-supplier networks that can help organizations choose appropriate IOISs. Our exploratory findings contribute to the theory-building in IOIS research and have practical implications for various industries.

Customer 3

2.2 Inter-organizational networks and social network analysis


Supplier 4 Customer 4

b
Customer 1 (Supplier 1)

Supplier (Customer)

Broadcasting Sales System

Customer 2 (Supplier 2)

Customer 3 (Supplier 3)

Fig. 2 a. Electronic market, b. broadcast sales systems

product(s). With the electronic monopoly IOISs, the customer has a choice of suppliers, but chooses to establish a sole source contract for the product(s). Choudhury (1997) assumes that these three types of IOISs should facilitate and in turn be influenced by three major types of customer-supplier relationships respectively: (1) a customer always purchases specific product(s) from a single supplier (supported by electronic monopoly IOIS); (2) a customer purchases products from one of a group of preferred suppliers (supported by electronic dyads); and (3) a customer searches the entire market for each purchase, and each time may trade with a different supplier (supported by multilateral IOIS). In addition, corresponding to the second and third type of the above customer-supplier relationships, Powell et al. (1990) reviewed sociological and economic literature on exchange and found that transactions may occur in a stable network of exchange partners who have close relationships, such as a strategic alliance, or between customers and suppliers who have impersonal and constantly changing relationships through markets. Based on Choudhurys typology of IOIS and other relevant studies, our research aims to 1) study strategic choices of industry-specific IOISs by

Previous studies on strategic choices of IOISs have addressed the growing importance of network perspectives, especially the customer-supplier relationship (Cunningham and Tynan 1993). Barua et al. (2004) examine a companys customer and supplier-side digitization efforts respectively. The results suggest that most firms are lagging in their supplier-side initiatives relative to the customer-side. However, supplier-side digitization has a strong positive impact on customer-side digitization, which, in turn, leads to better financial performance (Barua et al. 2004). Subramani (2004) analyzed supplier networks and found that IT deployments in supply chains lead to closer customer-supplier relationships (Subramani 2004). Bensaou (1997) empirically examines the cumulative influence of a number of factors in customer-supplier relations, including exogenous factors such as characteristics of environments and endogenous factors such as technology applications. He concludes that such factors conceptually and empirically capture their collective influence on cooperation, thus influence strategic choices. However, few studies have examined how the patterns of customer-supplier relationships in a networked environment affect companies strategic choices of IOISs. In our analysis, we focus on studying the customer-supplier networks to derive effective strategies for IOIS selections and implementations in organizational networks. We adopt social network analysis (SNA) methods to model and analyze a large inter-organization network in which nodes are firms and links are transactions that occur between customers and suppliers. Social network analysis was originally developed by sociologist Jacob Moreno (1934) to investigate the relationship between social structures and personal psychological well-being. He also invented the sociograma diagram of nodes and links used to represent relationships among social actors. In the early development of SNA, various other ad hoc studies in sociology, anthropology and psychology adopted similar concepts and methods. Linton Freeman in his book (2004) about the development of SNA observed that a growing number of researchers contributed to SNA in the 1960s. One of the most important research groups at that time, Harrison White and his students at Harvard University,

Inf Syst Front

Stanley Milgram (six degree of separation) (Milgram 1967), Wellman et al. (1996), elaborated and popularized SNA. Nowadays, with the advance of computing technologies and the availability of massive online data, network analysis methods have been widely used to study largescale organizational networks. At the individual level, Arakji and Lang (2007) examined the impacts of producer-consumer collaboration relationships on innovations at firms in the digital entertainment industry. Their results can be used to devise effective firm strategies for supporting product innovations. Buckner and Cruickshank (2008) applied SNA methods to study the relationships among researchers in network startups to support their research operations. Keith et al. (2008) examined how technology influenced the social network structure within a group. They concluded that individuals who are proficient with technologies tend to be more central in their group advice networks. At the firm level, Zack (2000) argued the importance of using social network analysis for framing and describing the effects of organizational systems on organizational forms and structures. However, he did not provide any formal framework for using SNA to study organizational systems. Beckman et al. (2004) examined factors that affect the firms choices of network partners. They analyzed data on alliance networks for the 300 largest U.S. firms from 1988 to 1993. The results showed that the stability of a firms alliance network structure depends on the type of uncertainty it experienced. The greater the uncertainty that a firm faces alone, the more likely this firm will expand its alliance network. Likewise, the greater the uncertainty that a firms market or industry faces, the more likely that firm will strengthen the ties it presently has with others. Another research conducted by Powell et al. (2005) studied the determinants of the partner selection process for biotechnology firms in the 1990s. Several types of determinants such as preferential attachment and homophily (i.e., people tend to interact with others having similar characteristics) were statistically examined using McFaddens (1980), McFadden and Zarembka (1974) discrete choice model. Carley (2002) developed a framework for computational analysis of social and organizational systems (CASOS) which utilized various computational approaches including agent technology and dynamic social network analysis. However, little research has been done on the customersupplier relationships using SNA methods. 2.2.1 Social network analysis methods In this section, we review several SNA measures we used in this study to analyze customer-supplier networks across different industries. At the individual node level, centrality measures are used to identify key members and interaction

patterns between sub-groups. One of the most commonly used centrality measuresa nodes degreeis defined by Freeman (1979) as the number of direct links this node has. It measures how active a particular node is. A network member with a high degree can be the leader or hub in a network. In addition, several network-level SNA measures such as average degree, clustering coefficient, average path length, and degree distribution are used to describe and distinguish different network topology models. Three models have been employed to characterize complex networks: random graph model, small-world model (Watts and Strogatz 1998), and scale-free model (Barabasi and Alert 1999). In random networks, most of the nodes have roughly the same number of links. Clustering coefficient is usually used to determine the small-world nature of social networks. It is the probability that two nodes with a common neighbor also link to each other. A small-world network usually has a significantly larger clustering coefficient (Watts and Strogatz 1998) than its random model counterpart, indicating a high tendency for nodes to form clusters and communities. A small-world network also usually has a relatively small average path length (i.e., average number of steps along the shortest paths for all possible pairs of network nodes) (Watts and Strogatz 1998). Degree distribution P(k) is the probability distribution that a node has exactly k links. Power-law degree distribution is used to characterize scale-free networks (Wasserman and Faust 1994). In such networks, a small fraction of the nodes have a large number of links while a big fraction of nodes have just a few. This scale-free topology may be caused by the newly joined nodes preferential attachment to the nodes with high degrees (Laender et al. 2000).

3 Data and methods In this study, the customer-supplier transaction data for major U.S. firms is extracted from the Standard & Poor s COMPUSTAT database. COMPUSTAT is a database which provides financial, statistical and market information on companies around the world to institutional investors, bankers, advisors, and analysts in corporate, private equity, and fixed income markets. This database covers more than 88,000 global securities, covering 98% of the worlds market capitalization, and provides nearly 40 years of company data history. According to the FASB (Financial Accounting Standard Board) regulation No. 14, firms need to report certain financial information for any industry segment that comprised more than 10% of consolidated yearly sales, assets,

Inf Syst Front Fig. 3 Sample data of COMPUTSTAT customersupplier dataset


S_ID C_ID 63593 63644 109522 122394 178538 Sales (Milllion) 5680 QEP CO HOME DEPOT 67.193 6066 ADI CORP IBM 96.4 11259 MED GEN WALMART 1.315 12.874 6066 PERFICIENT IBM 5680 ZEP INC HOME DEPOT 66.25 Supplier Customer Product Hardware Data Storage Healthcare Products Consulting Services Cleaning Products Year S_naics C_naics 2005 2004 2003 2006 2005 332212 334112 541611 541519 325612 444110 541519 452990 541519 444110

or profits. The COMPUSTAT dataset used in our study retrieved companies customer-supplier relationships from such reports. It also includes information such as sales, reporting year, and other related disclosures. The whole customer-supplier dataset in the COMPUSTAT database ranges over a 7-year period from 2002 to 2008. Figure 3 shows a sample set of data extracted from the COMPUSTAT customer-supplier dataset. This sample includes five suppliers and three customers. Other information includes total sales between the customer and supplier, reporting year, and their NAICS codes. A row in this data sample indicates a record of the customer-supplier relationship between these two firms. For example, as the second row shows, in 2004 IBM purchased $96.4 million worth of data storage products from ADI Corporation. Table 1 summarizes the COMPUSTAT customer-supplier dataset used in our study. This dataset includes transaction information involving 5,636 supplier firms and 13,065 customer firms from 2002 to 2008. From the COMPUSTAT customer-supplier dataset, we focused sub-datasets in five major industries: Information Technology (IT), retail, service, finance, and health care. The IT industry is included in our analysis to see if IT companies have special patterns in selecting IOISs since IOISs are essentially an IT product. The retail industry is selected since nowadays retail companies heavily rely on information systems to manage their relationship with suppliers in transaction management and supply chain management (e.g., Walmart). The finance industry is selected because most inter-organizational financial transactions today are computer-based and depend on inter-organizational information systems. Service and health care industries are included because they are considered as very promising application areas for inter-organizational information systems (De and Ferratt 1998; Bakos 1991). The companies from these five industries are extracted based on each firms NAICS (North American Industry Classification System) code. Each NAICS code represents a
Table 1 Basic statistics of COMPUSTAT customer-supplier dataset Time period 20022008 Number of suppliers 5,636 Number of customers 13,065 Number of products 5,035

specific industry category. For example, Intel, the worlds largest semiconductor chip maker, has NAICS code 334290 which is in the category of computer and electronic product manufacturing industry. For each industry, we first extracted all the firms in related NAICS categories and their trading partners (customers or suppliers). Then we construct the network using the extracted firms as nodes and the customer-supplier relationships among them as links. In this research, both the quantitative methods and interpretive findings are needed for studying the IOIS choices. The quantitative methods are used to analyze the topologies of customer-supplier networks, while the interpretative findings can provide the contextual interpretations of such topologies and help researchers and practitioners devise effective strategies for choosing appropriate interorganizational information systems.

4 Results 4.1 Basic statistics We conducted our analysis using the functions provided by the most widely used social network analysis software Ucinet 6. Each of the measures/functions is widely used in other social network analysis studies. Table 2 summarizes the basic statistics of the five constructed networks. IT, retail and service networks are relatively larger than other networks since they include firms across multiple realworld industries. This is mainly due to the categorizing method NAICS used. For instance, according to NAICS, the service network includes firms that provide all kinds of professional, scientific, and technical services ranging from tax preparation to interior design. On the other hand, firms in the finance and health care networks usually only focused on businesses within these two industries.
Table 2 Key statistics of the constructed customer-supplier networks across different industries IT # of nodes # of Links 742 862 Retail 273 279 Service 501 469 Health care 87 78 Finance 303 249

Inf Syst Front

4.2 SNA centrality measures We use SNA centrality measures to describe the topology of the five constructed customer-supplier networks. The customer-supplier relationship is a directional link which indicates flows of goods or services exchanged between supplier and customer companies. Thus we need to measure the network for both in links and out links. The average in-degree is actually the average number of suppliers a customer has and the average out-degree is the average number of customers a supplier has. Moreover, the maximum in(out)-degree is the largest number of suppliers (customers) a firm has in the measured network. Table 3 shows the centrality measures of all five constructed networks. We found that, in most of the networks, the average in-degree is larger than the out-degree. This indicates that in general there are more suppliers than customers in these five networks. Particularly, the average in-degree of the retail network is significantly larger than its average out-degree. This may be because retail industry firms mainly serve individuals rather than organizations. In addition, we measured network centralization and maximum degree for both in and out links in those five networks. Network centralization aims to measure the degree to which an entire network is focused around a few central nodes (Kang 2007). It indicates the levels of inequality or variance in the observed network as a percentage of that of a perfect star network of the same size. In the customer-supplier network, if the network centralization for in(out) links is much larger than the network centralization for out(in) links, it means that there is a substantial level of concentration or centralization of links on suppliers (customers). The network positional advantages for suppliers (customers) are rather unequally distributed in this network. This measure has particular importance for interorganizational studies of coordination and leadership. Turk (1977) argues that inter-organizational network centralization can be equated with coordination. Irwin and Huges
Table 3 Centrality measures of customer-supplier networks IT Average in-degree (suppliers) Average out-degree (customers) Maximum in-degree (suppliers) Maximum out-degree (customers) Network centralization for in links (%) Network centralization for out links (%) Network Centralization Comparison Rate (in links/out links)

(1992) define network centralization as the degree to which an inter-organizational network is dominated by a few places. In our study, network centralization for in links (out links) reflects the degree to which a customersupplier network is dominated by supplier (customer) companies. In this study, since we focused on customer-supplier (directed) networks and each node can be a customer or a supplier, one important question is which type of nodes (customer or supplier) has dominant positional advantage in each industry-specific network. To address this question, we construct a new measurenetwork centralization comparison rate which is the network centralization for in links divided by the network centralization for out links. If the network centralization comparison rate is much larger than 1, it means that supplier companies have much higher positional advantages over customer companies. In the context of our study, we defined the supplier (customer) companies having dominant positional advantages in a customer-supplier network as this network has/with a network centralization for in(out) links that is 2 times larger than (its) network centralization for out(in) links.. Thus the five networks are categorized into three typescustomercentric networks, supplier-centric networks and balanced networksusing the network centralization comparison rate . If 0 < a  0:5, the network is a customer-centric network (i.e., customer nodes have dominant positional advantage). If 2 , the network is a supplier-centric network. Networks with 0:5 < a < 2 are defined as balanced networks since neither type of the companies (nodes) has dominant positional advantage over each other. The comparison rates for all five customer-supplier networks are presented in the last row of Table 3. It was found that network centralization comparison rates for the in links of IT, retail and service networks are larger than 2. This means, in those networks, several firms that have many suppliers take the central positions while the remaining firms are peripheral. These networks are customer-centric networks. On the other hand, the network centralization comparison rate of the health care network is 0.49, indicating supplier firms take central positions. Thus

Retail 3.40 1.42 58 14 21.0 4.84 4.34

Service 2.11 1.85 77 11 15.24 2.02 7.54

Health care 1.93 1.93 5 9 4.81 9.65 0.49

Finance 1.69 1.56 11 9 3.42 2.73 1.25

2.3 2.2 44 13 5.82 1.63 3.57

Inf Syst Front

the health care network is a supplier-centric network. In addition, the finance network has a network centralization comparison rate of 1.25, which indicates it is a balanced network with customer and supplier companies having similar positional advantages. 4.3 SNA visualization analysis We used the network visualization tool NetDraw to visualize the topology of all five networks. The observations from

these visualizations also provide evidence to support our findings about customer-/supplier-centric networks and balanced networks. In both Fig. 4a and b, the large blue nodes represent major customer firms while the yellow nodes represent major supplier firms. It is obvious that customer firms dominate in the sample IT network (Fig. 4a) and take most central positions (customer-centric), while supplier firms have more positional advantages in the sample health care network (Fig. 4b) (supplier-centric).

Fig. 4 a. Sample IT customer-supplier network (customer-centric), b. Sample health care customer-supplier network (supplier-centric)

Inf Syst Front Fig. 5 Sample finance customer-supplier network (balanced networks)

As Fig. 5 shows, the finance network has similar numbers of customer and supplier firms with a large number of exchange links. This is also consistent with its centrality measures. The average degrees, maximum degrees and network centralization measures for both in and out links in the finance network do not differ much. Thus, neither customer firms nor supplier firms have positional advantage in the finance network. We defined such networks as balanced networks. 4.4 Component analysis Component analysis mainly aims to investigate the connectivity of the companies in each industry through the supply relationship. Components of a network are subnetworks that are connected within, but disconnected between, sub-networks. The proportion of the largest component to the whole network indicates if there is a core set of companies that are closely inter-connected with each other through the supply relationship in each industry. Such core companies often have advantages in gaining and efficiently distributing information, knowledge and resources because they are closely connected with other core companies. In contrast, fragmentation rate shows the proportion of the network that cannot reach each other. In

addition, the larger the fragmentation rate the larger the proportion of companies that cannot reach each other. Thus large fragmentation rate often indicates that the network has a lot of disconnected nodes or components. These two measures illustrate the overall connectivity of the customersupplier network in an industry. The rationale for component analysis is to derive empirical propositions about IOISs from the results. For instance, networks with large core components have better capabilities in distributing information and resources. From the IOIS perspective, these networks are more suitable for broadcast sales systems since such systems often require good network connectivity to broadcast information to as many companies as possible. In this research, component analysis was conducted on all five customer-supplier networks. The proportion of the largest component to the network and the fragmentation rate are reported in Table 4 for each of the five networks. We found that the proportions of the largest clusters in the three customer-centric networksIT, retail and service are 65%, 60.9% and 71.6% respectively, which accounts for more than 60% of their corresponding networks. On the other hand, the largest clusters in the health care and finance networks only include 32.1% and 35.7%, around 30% of the total number of nodes those networks have. In

Table 4 Results of the component analysis Number of nodes Size of the largest component (nodes) Proportion of the network Fragmentation rate

IT 742 560 65% 43%

Retail 273 170 60.9% 60.1%

Service 501 336 71.6% 54.9%

Health care 87 25 32.1% 86.5%

Finance 303 89 35.7% 87.8%

Inf Syst Front Table 5 Results of the topological analysis Average Clustering Coefficient Average Path Length Global Efficiency IT 0.005 3.34 0.112 Retail 0.001 1.64 0.117 Service 0.001 2.803 0.091 Health care 0.001 0.414 0.059 Finance 0.003 0.878 0.028

addition, the fragmentation rates of the health care network and finance network are larger than the three customercentric networksIT, service and retail networks. These results from the component analysis indicate that these three customer-centric networks are often much more connected than the supplier-centric networks. This implies that broadcast sales systems are suitable for customercentric networks. 4.5 Topological analysis SNA topology measures are developed to describe the topologies of the constructed customer-supplier networks. Table 5 shows the results of these measures on the five constructed networks. The results of topological analysis show that the average clustering coefficients for all networks are quite small, indicating that the firms in these networks have little tendency to cluster. This shows that the interorganizational customer-supplier networks are different from most social networks of individuals which usually tend to cluster together to form small-world networks. One possible explanation is that organizations may have more obstacles to overcome to form closely connected clusters, such as legal issues, competition within the industry, or organizational culture differences. The results also show that the average path lengths of the three customer-centric networks are much larger than the other two. This is consistent with the cluster analysis results. Since the health care and finance networks have many disconnected small clusters with short path length, the average path length of these two networks is much smaller than the customer-centric networks. Global efficiency is the average of the inverses of the lengths of the shortest paths over all pairs of nodes in a network. It is usually used to measure the communication efficiency of the network. In the customer-supplier networks, it measures how efficient the networks are in terms of transferring/exchanging information and products. The results

show that customer-centric networks are more efficient than supplier-centric networks. Table 6 shows the results of linear regression of the degree distributions for all five networks. It was found that all networks follow a power-law degree distribution (Newman 2001), modeled by p(k) k, while k is the degree and p(k) is the probability a node has degree k in the network under study. Most coefficients of determination R2 for the regressions are larger than 0.9 (ranging from 0 to 1), indicating high fitness of the power-law degree distributions. Thus all customer-supplier networks show scale-free features. These results indicate that, for all five networks, they more or less have the structure that a small number of nodes have a large number of links while the majority of nodes only have few.

5 Findings and discussion In this section, we first summarize our findings and then link our findings to the research question about how to choose appropriate IOISs for organizations. & There are three main types of customer-supplier networks: customer-centric, supplier-centric and balanced networks. In customer-centric networks, a small group of customer firms that have many suppliers take the central positions while the remaining firms are peripheral. Such large customer firms have more network positional advantages in terms of access to resources and communication efficiency than other firms in the networks. Suppliercentric networks have similar features with suppliers as the central nodes. In balanced networks, neither customer nor supplier firms have positional advantage. Customer-centric networks are usually larger since they include firms which have business operations across multiple industries, while supplier-centric networks only include firms that focus on business within one industry.

&

Table 6 Results of linear regressions on degree distributions of customer-supplier networks

IT Goodness of Fit R2 Power-Law Distribution Exponent 0.92 2.21

Retail 0.88 2.04

Service 0.97 2.22

Health care 0.78 1.71

Finance 0.91 2.32

Inf Syst Front Table 7 Proposed IOISs for firms in different types of customer-supplier networks Networks Customer-centric networks (e.g., IT, service, retail) Supplier-centric networks (e.g., health care) Balanced networks (e.g., finance) Features Positional advantages for customer firms; more connected; more efficient in terms of information exchange Positional advantages for supplier firms; many disconnected cliques; less efficient Customers and suppliers have similar positional advantage; less connected IOISs Broadcast Sales Systems (customer firm as users) Broadcast Sales Systems (supplier firm as users) OR Electronic Dyad Electronic Market

&

&

&

&

Among the five networks constructed based on NAICS industry code, IT, retail and service are customer-centric networks; the health care network is a supplier-centric network. In the finance network, neither customer firms nor supplier firms have dominant positional advantages. Thus the finance network may represent a network form in which customers and suppliers are in balance in terms of network position and embeddedness. SNA cluster analysis showed that customer-centric networks are much more connected than other networks, indicating they may have more critical relationships which bridge small local clusters to the rest of the network. SNA topological analysis showed that customercentric networks are more efficient than suppliercentric networks in terms of exchanging information and resources. It was also found that all five networks show strong scale-free features. In these customer-supplier networks, a small group of nodes has a large number of links while the rest only have a few links.

of firms. It may not be efficient to use a broadcast sales system to broadcast or collect product information. In such disconnected cliques, electronic dyad IOISs may work better since such systems mainly improve the local efficiency between the customer firm and a group of its preferred suppliers. 3) We suggest firms in balanced networks use electronic market IOISs. In balanced networks both customer and supplier firms are equally distributed and it would have similar effects to broadcast product information from either customer or supplier firms. Therefore, electronic market may be the most appropriate form of IOIS since it treats all participating firms equally and improves transaction efficiency in general.

6 Conclusions and future directions In this paper, we aim to devise effective strategies based on empirical findings to help firms to choose appropriate inter-organizational information systems. We first reviewed the typology of existing IOISs proposed by Choudhury. Then we used social network analysis to model and analyze realworld customer-supplier networks in five service-related industries. Our analysis identified three types of customersupplier networks based on their topological properties: customer-centric networks, supplier-centric networks, and balanced networks. According to their distinctive characteristics and functions, we then propose several strategies in choosing IOISs for these three types of networks. We argue that each proposed strategy may better utilize the advantages offered by its corresponding customer-supplier network and IOIS. Our future work consists of several directions including (1) extending the granularity of the classification of IOIS and the customer-supplier networks, and thereby providing more specific recommendations on IOIS strategies for practitioners, (2) empirically evaluating the effectiveness of the strategies proposed in our study, and (3) investigating the relationships between IOIS choices and firm performance. Our efforts will open a new venue of research in understanding the IT value in networked enterprises by incorporating insights from network analysis of inter-

Our research objective is to utilize the above findings to develop empirically-driven propositions for choosing appropriate IOISs for firms in different types of customersupplier networks. In Table 7, we link the above findings with the three major types of IOISs proposed by Choudhury (1997). It was suggested that 1) broadcast sales systems are suitable for firms in customer-centric networks; 2) broadcast sales systems or electronic dyad systems are suitable for supplier-centric networks; and 3) electronic market systems are suitable for balanced networks. The reasons for these suggestions are summarized as follows. 1) We suggest broadcast sales systems for customer-centric networks because customer firms are central in such networks and have many positional advantages in accessing product information from suppliers and broadcasting the requirements. 2) For similar reasons, we also recommend broadcast sales systems (supplier firm as users) to firms in supplier-centric networks. However, in some industries, the supplier-centric networks are less connected, with many disconnected cliques or even pairs

Inf Syst Front

organization relationships and developing better strategies for choosing appropriate IOISs.
Acknowledgements This research is supported by City University of Hong Kong Start-up Grant (Grant Number 7200102) and Strategic Research Grant (Grant Number 7002257).

References
Arakji, R. Y., & Lang, K. R. (2007) Digital consumer networks and producer-consumer collaboration: Innovation and product development in the digital entertainment industry. Hawaii International Conference on System Sciences. Bakos, J. Y. (1991). Information links and electronic marketplaces: the role of interorganizational information systems in vertical markets. Journal of Management Information System, 8(2), 3152. Barabasi, A.-L., & Alert, R. (1999). Emergence of scaling in random networks. Science, 286(5439), 509512. Barua, A., et al. (2004). An empirical investigation of net-enabled business value. MIS Quarterly, 28(4), 585620. Beckman, C. M., Haunschild, P. R., & Phillips, D. J. (2004). Friends or strangers? Firm-specific uncertainty, market uncertainty, and network partner selection. Organization Science, 15 (3), 259275. Bensaou, M. (1997). Interorganizational cooperation: the role of information technology an empirical comparison of U.S. and Japanese supplier relations. Information Systems Research, 8(2), 107124. Buckner, K., & Cruickshank, P. (2008). Social network analysis as a tool to evaluate the effectiveness of EC funded networks of excellence: The case of DEMO-net. Hawaii International Conference on System Sciences. Carley, K. M. (2002). Computational organization science: a new frontier. Proceedings of the National Academy of Sciences of the United States of America, 99(Suppl 3), 72577262. Choudhury, V. (1997). Strategic choices in the development of interorganizational information systems. Information Systems Research, 8(1), 124. Cunningham, C., & Tynan, C. (1993). Electronic trading. Interorganizational systems and the nature of buyer seller relationshipsthe need for a network perspective. International Journal of Information Management, 13(1), 328. De, P., & Ferratt, T. W. (1998). An interorganizational information system in the health care arena: insights gained from a hierarchical analysis. In Proceedings of the 1998 ACM SIGCPR conference on Computer personnel research (pp. 214223). ACM: Boston, Massachusetts, United States. Freeman, L. C. (1979). Centrality in social networks: conceptual clarification. Social Networks, 1, 215240. Freeman, L. C. (2004). The development of social network analysis. Vancouver: Empirical Press. Irwin, M., & Hughes, H. (1992). Centrality and the structure of urban interaction: measures, concepts, and applications. Social Forces, 71, 1751. Johnston, H. R., & Michael, R. V. (1988). Creating competitive advantage with interorganizational information systems. MIS Quarterly, 12(2), 153165. Kang, S. M. (2007). Equicentrality and network centralization: a micro-macro linkage. Social Networks, 29(4), 585601. Keith, M., Demirkan, H., & Goul, M. (2008). How does collaborative group technology influence social network structure? Hawaii International Conference on System Sciences.

Laender, A. H. F., Liddle, S. W., & Storey, V. C.(2000). Conceptual modelingER 2000 : 19th International Conference on Conceptual Modeling, Salt Lake City, Utah, USA, October 2000. Berlin; New York: Springer. xv, 588 p. McFadden, D. (1980). Econometric models for probabilistic choice among products. The Journal of Business, 53(3), S13S29. McFadden, D., & Zarembka P. (1974). Conditional logit analysis of qualitative choice behavior. In Frontiers in econometrics (pp. 105142). New York: Academic Press. Milgram, S. (1967). Small-world problem. Psychology Today, 1(1), 6167. Moreno, J. L. (1934). Who shall survive? (Vol. 58). Washington D.C.: Nervous and Mental Disease Publ. Newman, M. E. J. (2001). Scientific collaboration networks. I. Network construction and fundamental results. Physical Review E, 64, 06131. Powell, W. W., Staw, B., & Cummings, L. L. (1990). Neither market nor hierarchy: network forms of organization. In Research in organizational behavior (pp. 295336). Greenwich, CT: JAI Press. Powell, W. W., et al. (2005). Network dynamics and field evolution: the growth of interorganizational collaboration in the life sciences. The American Journal of Sociology, 110(4), 11321205. Subramani, M. (2004). How do suppliers benefit from information technology use in supply chain relationships. MIS Quarterly, 28 (1), 4573. Thomas, W. M., Joanne, Y., & Robert, I. B. (1987). Electronic markets and electronic hierarchies. Communications of the ACM, 30(6), 484497. Turk, H. (1977). Organizations in modern life: Cities and other large networks. San Francisco: Jossey-Bass. Wasserman, S., & Faust, K. (1994). Social network analysis: Methods and applications. Cambridge: Cambridge University Press. Watts, D. J., & Strogatz, S. H. (1998). Collective dynamics of smallworld networks. Nature, 393, 440442. Wellman, B., et al. (1996). Computer networks as social networks: collaborative work, telework, and virtual community. Annual Review of Sociology, 22(1), 213238. Zack, M. H. (2000). Researching organizational systems using social network analysis. Proceedings of the 33rd Hawaii International Conference on System Sciences, 7: p. 7043. Zhao, J., Wang, S., & Huang, W. V. (2008). A study of B2B e-market in China: E-commerce process perspective. Information & Management, 45(4), 242248.

Dr. Daning Hu received his Ph.D. degree in Management Information Systems from Eller College of Management, the University of Arizona, Tucson, Arizona, USA. He is currently a research fellow at the Department of Information Systems, City University of Hong Kong. His research interests include business intelligence, financial system risk management, social network analysis and open source communities. Before joining City University of Hong Kong, he worked as a research associate at the University of Arizona. Dr. Sherry Sun received the M.S. and Ph.D. degrees in Management Information Systems from Eller College of Management, the University of Arizona, Tucson, Arizona, USA. She is currently an assistant professor at the Department of Information Systems, City University of Hong Kong. Her research mainly focuses on the development of workflow technology and its applications in electronic commerce, knowledge management, and organizational process automation. Before joining City University of Hong Kong, she worked as a database developer in the Artificial Intelligence Lab at the University of Arizona and a database administrator in Arizona Cancer Center.

Inf Syst Front Dr. J. Leon Zhao is Head and Chair Professor in Information Systems, City University of Hong Kong. He was Interim Head and Eller Professor in the Department of Management Information Systems, University of Arizona, previously. He holds Ph.D. and M.S. degrees from the Haas School of Business, UC Berkeley, M.S. in Engineering from UC Davis, and B.S. from Beijing Institute of Agricultural Mechanization. Leon's research has been supported by NSF, SAP, and other funding agencies. Leon has served as associate editor of Information Systems Research, ACM Transactions on MIS, IEEE Transactions on Services Computing, Decision Support Systems, Electronic Commerce Research and Applications, among other journals. He has co-edited more than ten special issues in various IS journals including Decision Support Systems and Information Systems Frontiers and has chaired numerous international conferences including the 2010 Conference on Design Science Research, the 2009 IEEE Conference on Services Computing, the 2008 IEEE Symposium on Advanced Management of Information for Globalized Enterprises, the 2007 China Summer Workshop on Information Management, the 2006 IEEE Conference on Services Computing, among others. He received an IBM Faculty Award in 2005 for his work in business process management and services computing and was awarded Chang Jiang Scholar Chair Professorship at Tsinghua University by the Ministry of Education of China in 2009.

Mr. Xinlei Zhao is a Ph.D. in Management Information Systems, Eller College of Management, the University of Arizona. His main research interests focus on Modeling and Analysis Methodologies for Business Processes and Workflow, and Service Computing. His research has appeared at academic conferences including WeB, HICSS, and AMCIS.

You might also like