European Management Journal Vol. 22, No. 1, pp. 12–26, 2004 2004 Published by Elsevier Ltd.Pergamon

Printed in Great Britain0263-2373 $30.00doi:10.1016/j.emj.2003.11.016

Elucidating ElusiveEnsembles:The Strategic Value ofInformal InternetNetworksRIAN VAN DER MERWE, Curtin University of Technology, AustraliaLEYLAND PITT, Rotterdam School of ManagementPIERRE BERTHON, Bentley College, Massachusetts

For many years, sociologists have conceptualizedand debated the value of ‘social capital’: theresources embedded in an individual’s social net-work. More recently, the notions of network organi-zations and strategic alliances have becomeimportant forms of entrepreneurial venture. Animportant and often overlooked form of alliance onthe Internet is what we call the ‘elusive ensemble’.These associations of individuals and organizationsare typically undocumented, difficult to identify bythird parties, and even pass unrecognized by theirconstituent members. They consist of informalInternet networks that are extremely valuablebecause of the strategic social capital embedded inthem. Drawing on Social Network Theory, thispaper outlines a methodology for viewing and valu-ing informal Internet networks, and in doing sooffers guidance on improving the strategic efficacyof these elusive ensembles. 2004 Published by Elsevier Ltd.

Keywords: Elusive ensembles, Informal Internetnetworks, Social capital, Social network theory,Strategic alliances, Network organizations

Introduction

Sociologists have for many years explored the valueof social capital, or the resources embedded in social

European Management Journal Vol. 22, No. 1, pp. 12–26, February 200412

relations and social network structures (Lesser, 2000;Lin, 2001a, 2001b). In recent years, organizationalforms, such as strategic alliances and network organi-zations have been an important new form of entrepr-eneurial venture, particularly in business markets(Achrol, 1997; Achrol and Kotler, 1999). As newforms of cooperation, these organizational forms rep-resent attempts by existing firms to exploit strategicopportunities as collaborators, with each bringingparticular resources and skills to the alliance (Lewis,1990; Lorange, 1990). Typically, there are outcomesfor all parties, with positive or negative results beingshared, according to some predetermined procedure.Academics have rushed to study these new organiza-tional forms, for apart from their intrinsic interest,they represent increasingly important structures inthe global business environment.

The rise of the Internet has facilitated new forms ofstrategic alliance. An important and often overlookedtype of Internet association is what we call the ‘elus-ive ensemble’. These alliances are generally unac-knowledged, difficult to identify and can often beunknown, even to the members of the alliance them-selves. They consist of informal links and social net-works on the Internet — unspoken connections andcodes of conduct that are extremely valuable becauseof the social capital embedded in them. Examples ofthe elusive ensemble are the networks of intercon-nected websites dedicated to various music artists.Websites relating to a particular artist may link back

ELUCIDATING ELUSIVE ENSEMBLES

and forth between official and unofficial fan sites,and they may even link to e-commerce companieswhere fans can buy the artist’s merchandise — e.g.Amazon.com or CDNow.com.

This paper aims to find ways to formally identify andexpand these associations in entrepreneurial waysthat will improve their effectiveness, usefulness andultimately, their social capital, translated into finan-cial returns. We argue that the elusive ensemble isan important part of the different forms of strategicalliances that may exist between organizations. Theframework we use to study these networks is socialnetwork theory.

In general, social network theory explains the mech-anisms and structures that individuals use toaccumulate power in social settings, leading to therelated construct of social capital (Scott, 1991; Wass-erman and Faust, 1994). This can be viewed as anasset benefiting the actors who have possession of itand constituting a resource available to an actordefined by its function (Coleman, 1990). Contingentto the theory of social networks is the theory of struc-tural holes (Burt, 1992). Structural hole theory viewssocial capital as entrepreneurial opportunitiesinherent in structural disconnects. ‘Structural holesare disconnections or non-equivalencies betweenplayers in the arena… Structural holes are entrepren-eurial opportunities for information access, timing,and control’ (Burt, 1992).

This discussion on elusive ensembles and social net-work theory leads us to the general definition of theresearch problem: To what extent can social networktheory and its related constructs be used to identify andenhance elusive ensembles as an important form of stra-tegic alliance on the Internet?

This paper explores this problem by investigating thefollowing specific research questions:

1. Can social network theory and social capital theory beused to explore the formation of elusive ensembles onthe Internet?

2. Can social network analysis be used to identify keynodes in the network?

3. Once the key nodes in the network have been identified,can the distinguishing characteristics of these actors bedescribed in managerially useful ways in order toincrease the strategic value of the association? Morespecifically:

(a) Can those nodes that are at the center of elusiveensembles be used to influence a large network ofnodes more effectively?

(b) Can structural hole theory be used to find entrepren-eurial opportunities in these networks?

It is necessary to understand the basics of networktheory as well as to have an idea of what this theory

European Management Journal Vol. 22, No. 1, pp. 12–26, February 2004 13

can be used for, before an in-depth study to answerthese questions can be commenced. This paper startswith a review of social network theory, social capitaltheory and structural hole theory. It continues todescribe how these theories can be applied to net-works on the Internet, and elusive ensembles in parti-cular. A small network is analyzed to develop ourmethodology and to test the applicability of thetheory, and an example of a large elusive ensemblenetwork in the service industry is then presented andanalyzed. Areas for future research are identified,and in closing conclusions are drawn about mana-gerial implications and the limitations of the study.

Social Network Theory and the Theoryof Structural Holes

Social network theory is based on the network concept,which ‘emphasizes the fact that each individual hasties to other individuals, each of whom in turn is tiedto a few, some or many others, and so on’(Wasserman and Faust, 1994). It is possible to modelthese relationships between individuals to representthe structure of the network. These models use actorsand ties to represent networks. Actors are defined as‘discrete, individual, corporate, or collective socialunits …, e.g. people in a group, departments withina corporation, public service agencies in a city, etc.’(Wasserman and Faust, 1994). Ties are the relation-ships between actors. Through network analysis tech-niques, it then becomes possible to ‘study the impactof the structure on the functioning of the groupand/or the influence of this structure on individualswithin the group’ (Wasserman and Faust, 1994).Social network theory thus provides an alternative tothe notion that actors function independently, thatthey make their own choices without considerationof the actors around them with which they have tiesor relationships. Moreover, its strong theoretical andmathematical basis provides a framework for testinghypotheses about social relationships and theirinfluences.

Closely related to social network theory is the theoryof ‘social capital’ which deals with the intrinsic valueof network structures. In general terms, social capital‘consists of resources embedded in social relationsand social structure, which can be mobilized whenan actor wishes to increase the likelihood of successin a purposive action’ (Lin, 2001b). Lin (2001a) pointsout two important components concerning this defi-nition. First, resources are embedded in social relationsrather than in the individual. The properties of the net-work and an actor’s position in that network aremore important than the actor himself. Second, accessand use of these resources are dependant on an actorbeing aware of their presence. If an actor is not awareof ties or relationships between him and other actors,he cannot use the resources available to him. Social

%a large network of non-

redundant contacts will

provide the biggest

information benefits

ELUCIDATING ELUSIVE ENSEMBLES

capital then seems not to exist, and will only comeinto existence for that actor once he becomes awareof it. The ability to identify networks and key role-players in these networks will therefore make it poss-ible to identify social capital where the relevantactors may be completely unaware of them. Our goalin finding the key actors in a network can thereforebe likened to gauging the social capital of a networkand finding value in networks where it was not pre-viously observed. From the definition, it is inferredthat social capital depends first on an actor’s positionin the network (are they in the right place to accessthe resources?), and second, it depends on the natureof the resources in the network (are the resourcesworth accessing?).

The many benefits of social capital strengthen ourgoal to explore the value of elusive ensembles. Adlerand Kwon (2000) summarize some of these benefitsas follows:

❖ Social capital provides actors in the network withaccess to broader sources of information at lowercosts.

❖ Social capital provides actors in the network withextended power and influence.

❖ Social capital facilitates solidarity between actors,as strong networks encourage compliance withrules and customs without the need for formalcontrols.

An important extension of social network theory isRonald Burt’s theory of ‘structural holes’. This theoryaims to explain ‘how compe-tition works when players haveestablished relations withothers’ (Burt, 1992), and arguesthat networks provide twotypes of benefits: informationbenefits and control benefits.Information benefits refer to whoknows about relevant infor-mation and how fast they find out about it. Actorswith strong networks will generally know moreabout relevant subjects, and they will also knowabout it faster. According to Burt (1992), ‘players witha network optimally structured to provide thesebenefits enjoy higher rates of return on their invest-ments, because such players know about, and havea hand in, more rewarding opportunities’. Controlbenefits refer to the advantages of being an importantplayer in a well-connected network. In a large net-work, central players have more bargaining powerthan other players, which also means that they can, toa large extent, control many of the information flowswithin the network.

Burt’s theory of structural holes aims to enhancethese benefits to their full potential. A structural holeis ‘a separation between non-redundant contacts’(Burt, 1992). The holes between non-redundant con-tacts1 provide entrepreneurial opportunities that can

European Management Journal Vol. 22, No. 1, pp. 12–26, February 200414

enhance both the control benefits and the informationbenefits of networks. Figure 1 (page 16) shows agraphical representation of this definition.

We will now look at how structural holes can facili-tate the optimization of information benefits and con-trol benefits.

Structural Holes and InformationBenefits

There are several ways to optimize structural holes ina network to ensure maximum information benefits:

❖ The Size of the Network. The size of a networkdetermines the amount of information that isshared within the network. You have a much bet-ter chance to receive timely, relevant informationin a big network than in a small one. The size ofthe network is, however, not dependant merely onthe number of actors in the network, but the num-ber of non-redundant actors. The utility of a net-work with reference to its size can be described bya function known as Metcalfe’s Law. Robert Met-calfe, founder of 3COM Corporation and thedesigner of the robust Ethernet protocol for com-puter networks, observed that new technologiesare valuable only if many people use them.Specifically, the usefulness or utility of the net-work equals the square of the number of users

(Metcalfe’s Law, 1999). Themore people use a piece ofsoftware, a network, a parti-cular standard, a game, or abook, the more valuable itbecomes and the more newusers it will attract, increas-ing both the utility and thespeed of its adoption by still

more users. Therefore, a large network of non-redundant contacts will provide the biggest infor-mation benefits.

❖ Efficient Network. Efficiency in a network is con-cerned with maximizing the number of non-redundant contacts in a network in order to max-imize the number of structural holes per actor inthe network. It is possible to eliminate redundantcontacts by linking only with a primary actor ineach redundant cluster. This saves time and effortthat would normally have been spent on main-taining redundant contacts.

❖ Effective Networks. Effectiveness in a network isconcerned with ‘distinguishing primary from sec-ondary contacts in order to focus resources on pre-serving primary contacts’ (Burt, 1992: p. 21). Build-ing an effective network means buildingrelationships with actors that lead to themaximum number of other secondary actors,while still being non-redundant.

ELUCIDATING ELUSIVE ENSEMBLES

❖ Weak Ties. In his 1973 paper entitled ‘Thestrength of weak ties’, Mark Granovetter(Granovetter, 1973) developed his theory of weakties. The theory states that because a person withstrong ties in a cluster more or less knows whatthe other people in the cluster know (e.g. in closefriendships or a board of directors), the effectivespread of information relies on the weak tiesbetween people in separate clusters. ‘Weak ties areessential to the flow of information that integratesotherwise disconnected social clusters into abroader society’ (Burt, 1992). Structural holesdescribe the same phenomena as weak tiesbecause both emphasize the need forentrepreneurs to fill the gap between differentclusters and non-redundant contacts. However,structural hole theory goes one step further andstresses that what makes the gap important is notthe weakness of the tie but the structural hole overwhich it spans. Building and maintaining weakties over large structural holes enhances infor-mation benefits and creates even more efficientand effective networks.

To achieve networks rich in information benefits, itis necessary to build large networks with non-redun-dant contacts and many weak ties over structuralholes. Some of these information benefits are:

❖ More contacts are included in the network, whichimplies that you have access to a larger volumeof information.

❖ Non-redundant contacts ensure that this vastamount of information is diverse and inde-pendent.

❖ Linking with the primary actor in a cluster impliesa connection with the central player in that cluster.This ensures that you will be one of the first peopleto be informed when new information becomesavailable.

Structural Holes and Control Benefits

Once structural holes are identified and the networkis optimized to provide maximum information bene-fits, an important question is how these benefits canbe used to capitalize on the opportunities in the net-work. Control benefits answer this question. Struc-tural holes not only provide information benefits,they also give actors a certain amount of control innegotiating their relationships with other actors. Tounderstand the role of structural holes in this regard,it is necessary to understand the concept of tertiusgaudens. Taken from the work of George Simmel, thetertius gaudens is defined as ‘the third who benefits’(Simmel, 1923). It describes the person who benefitsfrom the disunion of two others. For example, whentwo people want to buy the same product, the sellercan play their bids against one another to get a higherprice for the particular product. Structural holes are

European Management Journal Vol. 22, No. 1, pp. 12–26, February 2004 15

the setting in which the tertius gaudens operates. Anentrepreneur stepping into a structural hole at theright time will have the power and the control tonegotiate the relationship between the two actorsdivided by the hole, most often by playing theirdemands against one another. A relevant case inpoint is that of Calyx & Corolla. Established in 1988,this company stepped into the structural holebetween flower growers and flower-buying cus-tomers by aiming to deliver flowers directly fromgrower to customer within 12 days of cutting. Nor-mally, growers deliver their flowers to different sel-lers who then sell the flowers to customers. Theseflowers are generally 1–2 weeks old. Owner RuthOwades said: ‘I envisioned a table with three legs,and Calyx & Corolla was only one of them. Thesecond was the best flower growers available, andthe third was Federal Express, the number one aircarrier’ (Wylie, 1991). Customers place their ordersby mail, phone, fax or the Internet. These orders aretransmitted electronically to growers, who then useFedEx to ship the completed order (Watson et al.,2000). As the tertius gaudens between grower and cus-tomer, Calyx & Corolla has control over the relation-ship between them. By cutting the flower shops outof the equation, they can negotiate price and distri-bution with a large amount of power because theyare connected directly to the customer.

Linking Information and ControlBenefits — Structural Holes and SocialCapital

Where structural holes provide a platform for tertiusstrategies, information is the substance with whichthe strategy is performed (Burt, 1992). Accurate,timely and relevant information delivered betweentwo non-redundant contacts at the right time createsan immense opportunity to negotiate and control therelationship between these actors. That is the powerof structural holes, and that is why the theory is sorelevant for business networks on the Internet. Thenext section will elaborate on this.

The concept of structural autonomy binds the theorytogether in an attempt to identify the actors in a net-work that are best positioned for the information andcontrol benefits that a network can provide. Anactor’s goal is first to have no structural holes at theirend of the relationship (as these holes could be usedby other actors to develop relationships for theirbenefit, or even to replace the actor); and second tohave as many structural holes at the other end ofone’s relationships with actors (to optimize theentrepreneurial opportunities available to one). ‘Play-ers with relationships free of structural holes at theirown end and rich in structural holes at the other endare structurally autonomous’ (Burt, 1992).

The different benefits of structural hole theory makes

ELUCIDATING ELUSIVE ENSEMBLES

it instrumental in the creation and development ofsocial capital in networks. The information and con-trol benefits described by this theory can identify andexpand the intrinsic value of networks. If we want tofind the hidden value of elusive ensembles, thesethree constructs — social network theory, social capi-tal theory and structural hole theory — are essentialresearch tools.

Developing and Testing theMethodology — Exploring SocialNetwork Theory and BusinessNetworks

Can social network theory and the theory of struc-tural holes be applied to networks on the Internet?And how would this be useful to practitioners andresearchers? This paper addresses these questions byfirst applying the theory to a simple network as anexample. Our example is Waterford Wedgwood(www.waterfordwedgwood.com), a large manufac-turing company in the crystal and ceramics industry.

Without elaborating on the mathematics, we explainsocial network analysis briefly to enable a betterunderstanding of the network analysis process as itis used in this research. We focus on very simple net-work structures, namely one-mode directional andnon-directional networks. One-mode networks aremade up of a single set of actors, e.g. people ororganizations. Directional ties in a network have anorigin and a destination, for example, person Aregards person B as a close friend, which does notnecessarily mean that person B regards person A asa close friend as well. Non-directional ties have nodirection, for example, if person A lives near personB, it automatically implies that person B lives nearperson A. We will use graph theoretic notation torepresent maps of the social networks. These graphsare also referred to as sociograms (for a more detaileddescription of graph theory and other theoreticalapproaches to social networks, see Wasserman andFaust (1994)). The sociogram in Figure 2 shows thedifferent notations for directional and non-direc-tional ties.

Figure 1 A Graphical Example of a Structural Hole.

European Management Journal Vol. 22, No. 1, pp. 12–26, February 200416

Figure 2 Directional vs. Non-directional Networks.

An important concept relevant to our discussion isthat of redundancy. Contacts are redundant when theylead to the same people, and therefore to the sameinformation (Burt, 1992). Consider the two networksin Figure 3.

In Figure 3, there are four links with actors in thenetwork on the left. In the network on the right, thereare eight links with other actors. However, some ofthe actors in the network on the right are connectedto each other, and therefore lead to the same contactsand the same information. Each of the four connec-tions between these eight actors consequently resultsin one redundant link. Both networks in Figure 3therefore have four non-redundant links. Eventhough the network on the right has more apparentlinks than the one on the left, four of these linksare redundant.

When social networks are constructed and sociog-rams set up, network analysis techniques can be usedto identify key attributes and characteristics of thenetworks. Table 1 outlines the characteristics andconstructs that are used in this study.

The actor dynamics especially (centrality, promin-ence and prestige) can be very useful for businessnetworks on the Internet, as this will identify themost influential actors in the network. To enable net-work analysis for networks on the Internet, we class-ify distinct websites as the actors or nodes in the net-work. Further, we classify the hyperlinks providedon one website to go to one of the other websites asthe relationships or links between these nodes. Foreach of the websites in question, it is thus asked doeswebsite X link to website Y? It is important to note thatif X links to Y, it does not necessarily mean that Ylinks to X. We are therefore constructing a one-modedirectional network.

An important issue that needs to be addressed here

Figure 3 Redundant vs. Non-redundant Links (Takenfrom Burt, 1992: p. 17).

ELUCIDATING ELUSIVE ENSEMBLES

Table 1 Outputs of Social Network Analysis

Characteristic Description

Actor prominence The ties of the actor make it particularly visible to other actors in the network. These actors are identifiedusing centrality measures and prestige measures.

Actor centrality Different measures that investigate the amount of ties an actor is involved in:

❖ Degree centrality measures the proportion of actors that are adjacent to a particular actor.❖ Closeness centrality measures how close an actor is to all the other actors in the network.❖ Betweenness centrality identifies actors that are between many actors in their linkages with each

other.

Actor prestige Different measures that investigate which actors are the object of extensive ties, thus focusing solely onthe actor as a recipient:

❖ Degree prestige measures the relative amount of actors that link to a particular actor.❖ Proximity prestige only measures actors adjacent to a particular actor.

Cohesive subgroups Subsets of actors among whom there are relatively strong, direct, intense, frequent ties. Also referred toas cliques.

(summarized from Wasserman and Faust, 1994)

is the question of whether or not passive hyperlinksbetween websites accurately reflect the relationshipbetween those organizations. Does a hyperlink implya strategic alliance between two organizations? Ingeneral, the answer is no. In fact, an argument canbe made that these external links are often completelyrandom and left to the whim of web designers. How-ever, the fact remains that a hyperlink does mean aflow of information between two organizations, andthat is our focus point. A link may be random, butif it exists, a flow of information exists. We argue thatsuch an ‘elusive’ link can in fact become a strategicalliance by identifying and enhancing the intrinsicsocial capital created by that link. We thereforeassume that a hyperlink is the best possible indi-cation of an elusive ensemble between two organiza-tions.

The Waterford Wedgwood website was studied toidentify all the hyperlinks linking to pages outsidethe Waterford Wedgwood site. Ten external linkswere found. Each of these 10 websites was then vis-ited in turn to see (1) whether or not they link backto Waterford Wedgwood and (2) to which externalwebsites each of them link in turn. The external web-sites identified in this iteration were then visited tocheck whether or not they link back to the site fromwhich the link originated. The study was thus perfor-med by identifying external links up to the third levelaway from Waterford Wedgwood.

The links were used to set up a matrix that showshow the websites are related to each other. Thismatrix (sociogram) is shown in Table 2. Each of the16 websites was numbered as shown. A 0 entry in aspecific cell indicates that no link exists, and a 1 indi-cates that a link does exist.

The next step in the analysis was to construct a soci-

European Management Journal Vol. 22, No. 1, pp. 12–26, February 2004 17

ogram to get a clearer view of the network structure.Two main clusters of websites were also defined,namely, (1) websites relating to Waterford Wedg-wood and (2) websites relating to collectors of crystaland crystal-related objects. The Wedgwood Collectorswebsite falls into both categories, even though it isshown only in the first category on the sociogram.This website will be discussed later, as it plays animportant role in the network analysis. A visual mapof the network is shown in Figure 4. The numbers inthe nodes refer to the websites as they are numberedin Table 2.

The rest of this section will show how social networkanalysis and structural hole theory is useful and rel-evant in the analysis of this network in particular,and to business networks on the Internet in general.

Applying Social Network Theory

Social network theory enables us to identify the cen-tral actors, most prominent actors and most pres-tigious actors in this network. Complex mathematicalmanipulation of the network is necessary to calculatethe characteristics of these actors accurately, but withthis network, we merely aim to develop and test themethodology. When one looks at the network, it ispossible to make some preliminary judgments onactor characteristics. Waterford Wedgwood and Wedg-wood seem to be the central actors because they areinvolved in the most links. These websites are alsothe most prominent actor, because directly andindirectly it is the most accessible to the other actors.Wedgwood is the most prestigious actor, as it is therecipient of the most number of links (five incominglinks, compared to three incoming links each for Wat-

ELUCIDATING ELUSIVE ENSEMBLES

Table 2 Waterford Wedgwood Network of External Links

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

Waterford Wedgwood 1 – 1 1 1 1 0 0 0 0 0 1 1 1 1 1 1Waterford Crystal 2 0 – 1 1 1 1 1 0 0 0 0 0 0 0 0 0Wedgwood 3 1 1 – 0 0 0 0 1 1 1 0 0 0 0 0 0Rosenthal 4 0 0 0 – 0 0 0 0 0 0 0 0 0 0 0 0All-Clad 5 1 0 0 0 – 0 0 0 0 0 0 0 0 0 0 0WWICOP 6 0 0 0 0 0 – 0 0 0 0 0 0 0 0 0 0Waterford Visitor Centre 7 0 0 0 0 0 0 – 0 0 0 0 0 0 0 0 0Wedgwood Collectors 8 0 0 0 0 0 0 0 – 0 0 0 0 0 0 0 0Wedgwood Museum 9 0 0 1 0 0 0 0 0 – 0 0 0 0 0 0 0Wedgwood Hotel Ware 10 0 0 0 0 0 0 0 0 0 – 0 0 0 0 0 0Wedgwouldn’t 11 0 0 1 0 0 0 0 0 0 0 – 0 0 0 0 1Wedgwood Imagebank 12 1 1 1 1 1 0 0 0 0 0 0 – 0 0 0 0Versace 13 0 0 0 0 0 0 0 0 0 0 0 0 – 0 0 0Keith Murray Collectors 14 0 0 0 0 0 0 0 0 0 0 0 0 0 – 0 0Clarice Cliff Collectors 15 0 0 0 0 0 0 0 0 0 0 0 0 0 0 – 0Macromedia Flash 16 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 –

Figure 4 Visual Map of the Waterford Wedgwood Network.

erford Wedgwood, Waterford Crystal, Rosenthal andAll-Clad).

The question is, of course, how this information isrelevant and how it can help managers andresearchers in their understanding of business net-works. When the important actors in a network areknown, it simplifies the communication within thatnetwork. It becomes possible to eliminate unimport-ant, redundant links, and enables one to focus onlyon the actors with the most information benefits. Thisfrees up time and resources that can be used to buildrelationships with important actors in other net-works, thus ever increasing the number of non-redundant links and resulting in efficient and effec-tive networks with maximum information benefits.This aspect links closely with structural hole theoryas we have seen, and the next section shows howthe theory was used to analyze the network currentlyunder discussion.

European Management Journal Vol. 22, No. 1, pp. 12–26, February 200418

Applying Structural Hole Theory

Applying structural hole theory to this network isinsightful. Figure 4 is a good example of a structur-ally autonomous network. There are virtually nostructural holes at Waterford Wedgwood’s side of thenetwork, but on the other side of the relationship(actors 13–16), there are quite a few structural holes.Consider now, for example, actors 14 and 15. Theseactors, along with actor 8 (Wedgwood Collectors) aremore or less in the same market, which is the ceramiccollectors items market. Closer inspection of the sitesof actors 14 and 15 revealed that there is another cer-amic collectors site connected to them (Susie CooperCollectors, www.susiecooper.co.uk), and that thesethree sites are all connected by an online store thatsells their products called Age of Jazz(www.ageofjazz.com). Figure 5 shows a simplifiedWedgwood Waterford network with the additionalwebsites included.

ELUCIDATING ELUSIVE ENSEMBLES

Figure 5 Simplified Visual Map of Waterford Wedgwood Network.

Figure 5 shows how Age of Jazz identified the struc-tural holes between Keith Murray Collectors, ClariceCliff Collectors and Susie Cooper Collectors and capi-talized on the opportunity to be the tertius gaudensbetween them. The structural hole between Wedg-wood Collectors and the other collector sites is one thatcan still provide opportunities for entrepreneurs.

As this example is analogous to countless other net-works on the Internet, it seems therefore that struc-tural hole theory can be very useful in the study ofbusiness networks on the Internet. The key is to findclusters of similar sites in the sociogram and then tofind the structural holes within these clusters, andeven between different clusters. For example, frominspection, it is clear that Age of Jazz is the prominentactor in the ceramic collectors cluster shown in Figure5. There may be deep structural holes between thisnode and nodes in other specialized shopping clus-ters. Linking Age of Jazz with the prominent actors inthose networks will add useful non-redundant linksto the network, enlarging the network and providingenhanced information benefits for the whole net-work.

This simple network enabled us to develop a meth-odology to explore our research questions relating toelusive ensembles:

❖ Construct a sociogram of the network using web-sites as nodes and hyperlinks as links betweenthe nodes.

❖ Use social network analysis to identify the mostprominent actors in the network.

❖ Use structural hole analysis to identify possibleentrepreneurial opportunities in the network.

❖ Interpret these results to enhance social capital andexplore the strategic value of the associations thatwere identified.

We now apply this methodology to a large elusiveensemble in the service industry.

European Management Journal Vol. 22, No. 1, pp. 12–26, February 2004 19

Social Networks and Structural HolesApplied to Elusive Ensembles on theInternet

After successfully testing the application of socialnetwork theory and structural hole theory on a smallvirtual network, the methodology is now applied toa large elusive ensemble on the Internet. We investi-gate the informal online connections formed by Citig-roup Australia (www.citigroup.com.au), an inter-national banking company specializing in personal,corporate and investment banking.

As with the Waterford Wedgwood example, the Citi-group Australia website was studied to identify allthe hyperlinks linking to pages outside the site. Tenexternal links were found. Each of these 10 websiteswas then visited in turn to see (1) whether or notthey link back to Citigroup Australia and (2) to whichexternal websites each of them link in turn. The exter-nal websites identified in this iteration were then vis-ited to check whether or not they link back to the sitefrom which the link originated. A sociogram matrixwith 47 nodes was then constructed (see AppendixA). A visual map of the network, constructed withsoftware called Cyram Netminer 1.1.5, is shown in Fig-ure 6.

Analysis of the data occurred in three stages:

❖ Prominence analysis (centrality measures andprestige measures)

❖ Cohesive subgroups (clique analysis)❖ Structural hole analysis

This process is progressive as well as recursive —each stage is important not only to the next stage, buteach stage also calls the analyst back to the previousstage. For example, prominence measures largelydetermine the nodes on which structural hole analy-sis will be performed, and structural hole analysis isinterpreted using both the prominence measures andclique analysis.

ELUCIDATING ELUSIVE ENSEMBLES

Figure 6 Citigroup Network.

Each of these stages will now be discussed. Twosoftware packages were used to perform the analysis,namely UCINET V and STRUCTURE (for these andother software packages that can be used for socialnetwork analysis, see http://www.heinz.cmu.edu/project/INSNA/soft—inf.html). UCINET V is one ofthe most popular, comprehensive and user-friendlysocial network analysis tools. STRUCTURE was writ-ten by Ronald Burt specifically to do structural holeanalysis.

Prominence Analysis

Three centrality measures were computed, namely,Degree centrality, Closeness and Betweenness. Degreecentrality calculates a normalized value of theamount of actors adjacent to a particular actor. Thisis an important measure because it provides ‘anintuitive notion of how well connected a point iswithin its local environment… the correspondingagent is central in the sense of being ‘well-connected’or ‘in the thick of things’’ (Scott, 1991). Because thisis a directional network, we divided this measureinto two parts, namely out-degree and in-degree.Out-degree computes the number of links sent toanother actor, while in-degree refers to the number oflinks received by each actor (Valente, 1995). Results forthe top actors in each category are shown in Tables 3and 4.

It is interesting to note that Salomon Smith BarneyAustralia with 35 outgoing links has only twoincoming links. It is clear that these measures should

European Management Journal Vol. 22, No. 1, pp. 12–26, February 200420

Table 3 Out-Degree Centrality

Website Out- NormalizedDegree Out-Degree

Salomon Smith Barney Australia 35 76.087SSB Access 26 56.522Citigroup Australia 10 21.739Axiss Australia 8 17.391Australian Prudential Regulation 7 15.217AuthorityCitigroup Global 6 13.043Citibank SSB 5 10.870SSB Global Equities On-line (GEO) 5 10.870Citibank Global 4 8.696Diners Club International 4 8.696

therefore be analyzed separately, and in conjunctionwith the measures to follow, before any conclusionsare made about the level of centrality of a particularactor. Also note that Citigroup Australia is present inboth tables, as would be expected.

Closeness measures how close an actor is to all theother actors in the network. Wasserman and Faust(1994) elaborates by saying that ‘an actor is central ifit can quickly interact with all others… actors occu-pying central locations with respect to closeness canbe very productive in communicating information toother actors’. The measure finds actors with theshortest communication paths to the others. Again,different measures for Out-Closeness and In-Close-ness were calculated. The top actors in each categoryare shown in Tables 5 and 6.

ELUCIDATING ELUSIVE ENSEMBLES

Table 4 In-Degree Centrality

Website In- NormalizedDegree In-Degree

Citibank Global 10 21.739Australian Bureau of Statistics 5 10.870Australian Taxation Office 5 10.870Citibank Australia 5 10.870Citigroup Australia 5 10.870Australian Stock Exchange 4 8.696Bloomberg 4 8.696Citibank SSB 4 8.696Citigroup Global 4 8.696Citigroup Private Bank 4 8.696Diners Club International 4 8.696(Australia)Reuters 4 8.696Salomon Smith Barney Global 4 8.696

Table 5 Out-Closeness Centrality

Website Out-Closeness

Salomon Smith Barney Australia 77.966SSB Access 65.714Citigroup Australia 56.098Diners Club International (Australia) 37.705Australian Indigenous Leadership Centre 36.220Learning Links 36.220Diners Club International 28.571

Table 6 In-Closeness Centrality

Website In-Closeness

Salomon Smith Barney 3.498Citibank Global 3.415Citibank Australia 3.410Citigroup Private Bank 3.410Citibank SSB 3.407Citibank e-Business 3.395Citibank Icard 3.367Personal Banking for Overseas Employees 3.367Worldwide Securities Services 3.365

The third and last dimension of centrality that wasmeasured is betweenness. Betweenness identifiesactors that are between many actors in their linkageswith each other. This measure is important because‘a point of relatively low degree [centrality] may playan important ‘intermediary’ role and so be very cen-tral to the network… The betweenness of a pointmeasures the extent to which an agent can play the‘broker’ or ‘gatekeeper’ with a potential for controlover others’ (Scott, 1991). In this sense, it is stronglyrelated to structural hole theory. The actors that

European Management Journal Vol. 22, No. 1, pp. 12–26, February 2004 21

scored the highest on this measure are shown inTable 7.

As expected, Citigroup Australia, as the originatingwebsite, has the highest betweenness centrality.These measures are significant because they effec-tively identify where in the network, the social capi-tal is the highest. These central actors have the mostpower to influence those around them, and they aretherefore the most valuable. These nodes are invalu-able in turning an elusive ensemble into a strategicalliance.

Clique Analysis

Cohesive subgroups, or cliques, are subsets of actorsamong whom there are relatively strong, direct,intense, frequent ties. The method of n-cliques wasused to analyze the subgroups in this network. Intraditional clique analysis, a clique is defined as ‘asubset of nodes, all of which are adjacent to eachother, and there are no other nodes that are alsoadjacent to all the members of the clique’(Wasserman and Faust, 1994). n-Cliques is differentin that it specifies a cutoff value, n, as ‘the maximumpath length at which members of the clique will beregarded as connected’ (Scott, 1991). A basic cliqueanalysis of adjacent actors is therefore an n-cliquewith n = 1. We decided on an often-used value of n= 2. A 2-clique ‘is one in which the members are con-nected either directly (at distance 1) or indirectlythrough a common neighbor (distance 2)’ (Scott,1991).

Of the many different ways to analyze n-cliques indirectional relations, we used recursively n-connectedcliques. This means that there is a path of length n orless from i to j, and a path of length n or less from jto i, and the path from i to j uses the same nodes andarcs as the path from j to i, in reverse order(Wasserman and Faust, 1994). A UCINET V analysisusing n = 2 and a minimum set size of three actorsidentified six 2-cliques, shown in Table 8.

As expected, Citigroup Australia features extensively

Table 7 Betweenness Centrality

Website Betweenness

Citigroup Australia 184.400Salomon Smith Barney Australia 141.683Citibank 57.583Citibank SSB 49.533Diners Club International (Australia) 47.000Citigroup 25.517Citibank Australia 19.700SSB Global Equities On-line (GEO) 6.233Axiss Australia 4.833Diners Club International 3.000

ELUCIDATING ELUSIVE ENSEMBLES

Table 8 n-Clique Analysis

No. Actors in 2-clique

1 Citibank GlobalCitibank AustraliaCitibank e-BusinessCitibank SSBCitigroup Private Bank

2 Citibank GlobalCitibank AustraliaCitibank I-card

3 Citibank GlobalCitibank SSBWorldwide Securities Service

4 Australian Indigenous Leadership CentreCitigroup AustraliaDinersclub AustraliaLearning LinksSSB Connect

5 Citigroup AustraliaDinersclubDinersclub Australia

6 Citigroup AustraliaSSB ConnectSSB Access

in these cliques. As with centrality measures, but in abroader sense, cliques show the areas of the networkwhere social capital is the highest. As mentionedearlier, these measures are vital in identifying thestrategic significance of certain actors.

Structural Hole Analysis

The final stage of this network analysis is StructuralHole Analysis. Burt’s (1992) STRUCTURE softwarewas used for this purpose. The analysis was perfor-med on the Citigroup Australia website, which is thestarting point of this network and naturally also oneof its most central actors. The analysis calculates sev-eral measures, the most important of which are:

❖ Oi, the lack of holes around the actor itself.❖ cij, the constraint of absent primary holes. An

actor’s (i) entrepreneurial opportunities are con-strained to the extent that another of his contacts(q), with whom he has a strong relationship,invested heavily and therefore also has a strongrelationship with actor j. To put it another way,‘your entrepreneurial opportunities are con-strained to the extent that you have invested thebulk of your network time and energy in relation-ships that lead back to a single contact’ (Burt,1992). cij must therefore be low for structural holes(entrepreneurial holes) to exist.

European Management Journal Vol. 22, No. 1, pp. 12–26, February 200422

❖ pij, the ‘proportion of i’s network time and energyinvested in each relationship’ (Burt, 1992).

Structural holes occur around a specific actor whenhe has a large Oi (no or few structural holes aroundthe actor himself); and there is a large pij and a smallcij at the other end of a relationship with anotheractor (large amount of time and energy invested inthe relationship, small constraint on the actor’sentrepreneurial opportunities). These conditions giverise to what Burt (1992) calls the hole signature of anactor. This visual representation of the differencebetween pij and cij provides an easy way to identifythe structural holes around a specific actor.

Structural hole analysis for Citigroup Australiarevealed an Oi = 1, which means that there are nostructural holes around this actor. Further analysisrevealed the following values for pij and cij, shown inTable 9.

The signature hole for Citigroup Australia is shown inFigure 7. A large band in the hole signature indicatesan opportunity relationship, ‘relationships in which theplayer has the greatest room to negotiate, and so con-trol’ (Burt, 1992). Narrow bands indicate constraintrelationships, where ‘the player is most out of control’(Burt, 1992). A third category, residual relationships, isreserved for instances where a small amount of timeand energy is invested in a particular actor (small pij).These relationships are usually ignored, as ‘there islittle to protect and little to gain by alleviating con-straint’ (Burt, 1992). It should be noted, however, thatthese actors could potentially have value to the mainplayer if the relationship is developed further.

From the table and figure, it is clear that the biggestentrepreneurial opportunities exist between CitigroupAustralia and Diners Club International (Australia),Australian Indigenous Leadership Centre and LearningLinks, respectively. These are the opportunity relation-ships in this network.

Interpretation and Discussion

The two important outcomes from this analysisenable us to answer research question 3 posed in theIntroduction: Once the key nodes in the network havebeen identified, can the distinguishing characteristics ofthese actors be described in managerially useful ways inorder to increase the strategic value of the association?

First, the prominence analysis calculated three meas-ures that can be used effectively to identify the mostimportant actors in the network. By implication,this will develop the social capital of that network,which will define elusive ensembles more clearly asstrategic assets to an organization. The specificmeasure(s) used in the interpretation depends largelyon the nature of the network and what the marketer

ELUCIDATING ELUSIVE ENSEMBLES

Table 9 Structural Hole Analysis for Citigroup Australia

Website pij cij pij�cij

A Salomon Smith Barney Australia 0.0450 0.1333 0.0883B Diners Club International (Australia) 0.0194 0.1333 0.1139C Australian Indigenous Leadership Centre 0.0178 0.1333 0.1155D Learning Links 0.0178 0.1333 0.1155E SSB Access 0.0163 0.0667 0.0504F Citigroup Global 0.0070 0.0667 0.0597G Citigroup Private Bank 0.0059 0.0667 0.0608H Citibank Australia 0.0053 0.0667 0.0614I Citigroup Asset Management Australia 0.0053 0.0667 0.0614J Australian Institute of Aboriginal and Torres Strait Islander 0.0044 0.0667 0.0623

StudiesK Citibank SSB 0.0044 0.0667 0.0623

Figure 7 Hole Signature for Citigroup Australia.

intends to accomplish in the network. For example,if the intention is to introduce information to a net-work of websites as fast as possible, In-Degree cen-trality, In-Closeness centrality and Betweenness wouldbe the best measures to use, as these measures indi-cate which actors are most likely to be reached whensomeone visits one of the sites included in the net-work. Studying these measures for the CitigroupAustralia network, the measures will therefore leadus to introduce new information on websites like Cit-ibank Global, Citigroup Australia, Salomon Smith Barneyand Salomon Smith Barney Australia. The 2-cliqueanalysis confirms these results, as most of these sitesare part of one or more cliques. Centrality measuresare also strongly linked to the diffusion of inno-vations. Valente (1995) notes ‘For networks, struc-tural centrality is associated with more rapid dif-fusion for advantageous innovations and slowerdiffusion for more risky/uncertain innovations’. Cen-trality is therefore not only an important measure todecided which actors are the most influential in a net-work, it also directly affects the speed with which newinformation is disseminated.

Second, the structural hole analysis identified threepossible entrepreneurial opportunities in the net-work, namely between Citigroup Australia and Diners

European Management Journal Vol. 22, No. 1, pp. 12–26, February 2004 23

Club International (Australia), Australian IndigenousLeadership Centre and Learning Links, respectively.These opportunity relationships can now be investi-gated to see whether or not it is practically possibleto find a tertius strategy. As with centrality measures,these opportunity relationships can develop thesocial capital of the network, and the elusive ensem-ble then becomes strategic for the organizations inquestion. In this case, the most interesting opport-unity exists between Citigroup Australia and DinersClub Australia. The fact that Diners Club is a subsidi-ary of Citigroup also raises some interesting ques-tions. If these companies are in business together,why is there a structural hole between them? A typi-cal tertius gaudens strategy in this case will involve athird party organization that puts itself between thesetwo organizations and negotiates the relationshipbetween them. An entrepreneurial opportunity forthe third party organization may be to serve as inintermediary not only between these two organiza-tions, but between Citigroup and its other subsidiariesas well. This new network will free up resources forCitigroup that has previously been tied up in manag-ing relationships between and among subsidiaries.The third party organization will then spend most ofits resources on managing these complex relation-ships. Citigroup will have effectively outsourced theirnetwork management function, freeing up valuableresources that can be used for core business functions.

Every network will have a different structural holeanalysis output, and each network should be ana-lyzed and interpreted separately. The Citigroup caseshows that although it is impossible to provide a setblueprint for interpreting the results, each networkand its analysis will raise enough questions to pointthe researcher/manager in the right direction toidentify entrepreneurial opportunities.

Conclusions and ManagerialImplications

This paper establishes an answer to our first researchproblem: can social network theory be used to

ELUCIDATING ELUSIVE ENSEMBLES

explore the formation of elusive ensembles on theInternet? Our research clearly shows how this can bedone. We also begin to answer the other two researchproblems. It showed how maps or sociograms can becreated to represent elusive ensembles on the Inter-net; it showed how social network theory can enableus to identify important actors in a network; and itshowed how social network theory, combined withthe theory of structural holes, can be used by man-agers to identify and enhance the social capital intheir networks by identifying entrepreneurial opport-unities in their own market and beyond. The paperultimately shows that through proper network analy-sis, elusive ensembles can constitute an importantform of strategic alliance.

One limitation of the research is that only one net-work was used to illustrate the procedure, and theremay therefore be a limitation to the generalization ofthe findings. Future research could apply thismethod to more (and even larger) networks in orderto establish whether or not the findings of this paperare accurate. Further research will also expand onthese preliminary findings and seek ways to for-malize the argument for proper use in any industry.

As mentioned in the paper, and as illustrated in theCitigroup Australia case, there are two main outcomesfor the different network analyses. First, prominenceanalysis can be used to identify the most importantactors in a network. This information is essential formanagers who want to disseminate informationacross a network as quickly as possible. Prominencemeasures not only allow managers to identify theactors that will have the most influence on the net-work, it also shows them which actors to target ifthey want to distribute the information in the net-work in the fastest way possible. This observationfinds a strong tie-in with and is thoroughly backedup by Diffusion of Innovation theories (see Valente,1995; Rogers, 1995). Second, managers can use thisinformation to identify entrepreneurial opportunitiesin the network, and it will also make them aware ofthe opportunities that may exist around them-selves — opportunities that can be exploited by thirdparty organizations.

These two outcomes enable researchers and prac-titioners to enhance the social capital of informalassociations on the Internet. In a world that is increas-ingly global and virtual this is an important contri-bution to our knowledge of business marketing andstrategy.

European Management Journal Vol. 22, No. 1, pp. 12–26, February 200424

Appendix A Citigroup Network

(See next page)

Note

1. Redundancy will be discussed in detail in the next section.Contacts are redundant when they lead to the same actorsin a network, and therefore to the same information.

References

Achrol, R.S. (1997) Changes in the theory of interorganiza-tional relations in marketing: toward a network para-digm. Journal of the Academy of Marketing Science 25(1),56–72.

Achrol, R.S. and Kotler, P. (1999) Marketing in the networkeconomy. Journal of Marketing 63, 146–164.

Adler, P.S. and Kwon, S.W. (2000) Social capital: the good,the bad, and the ugly. In Knowledge and Social Capital —Foundations and Applications, ed. E.L. Lesser. Butterworth-Heinemann, Boston.

Burt, R.S. (1992) Structural Holes: the Social Structure of Compe-tition. Harvard University Press, Cambridge, MA.

Coleman, J.S. (1990) Foundations of Social Theory. Harvard Uni-versity Press, Cambridge, MA.

Granovetter, M.S. (1973) The strength of weak ties. AmericanJournal of Sociology 78, 1360–1380.

Lesser, E.L. (2000) Leveraging social capital in organizations.In Knowledge and Social Capital — Foundations and Appli-cations, ed. E.L. Lesser. Butterworth-Heinemann, Boston.

Lewis, J. (1990) Partnership for Profit — Structuring and Manag-ing Strategic Alliances. Free Press, New York.

Lin, N. (2001a) Building a network theory of social capital. InSocial Capital — Theory and Research, eds N. Lin, K. Cookand R.S. Burt. Aldine de Gruyter, New York.

Lin, N. (2001b) Social Capital — A Theory of Social Structure andAction. Cambridge University Press, Cambridge.

Lorange, P. (1990) Building successful strategic alliances. LongRange Planning December, 34–46.

Metcalfe’s Law (1999) From http://searchnetworking.techtarget.com/sDefinition/0,,sid7—gci214115,00.html.Accessed on May 13, 2002. Last updated on Nov 25, 1999.

Rogers, E.M. (1995) Diffusion of Innovations (4th ed.). FreePress,

Scott, J. (1991) Social Network Analysis — A Handbook. SagePublications, London.

Simmel, G. (1923) The Sociology of Georg Simmel. Free Press,New York.

Valente, T.W. (1995) Network Models of the Diffusion of Inno-vations. Hampton Press, New Jersey.

Wasserman, S. and Faust, K. (1994) Social Network Analysis —Methods and Applications. Cambridge University Press,Cambridge.

Watson, R.T., Zinkhan, G.M. and Pitt, L.F. (2000) Object-orien-tation: a new perspective on strategy. In Academia/IndustryWorking Conference on Research Challenges.

Wylie, D. (1991) Calyx & Corolla. Harvard Business School.

ELUCIDATING ELUSIVE ENSEMBLES

Ap

pe

nd

ixA

.C

itig

rou

pN

etw

ork

European Management Journal Vol. 22, No. 1, pp. 12–26, February 2004 25

ELUCIDATING ELUSIVE ENSEMBLES

RIAN VAN DER LEYLAND PITT, 8,MERWE, 8, Bretby Close, Bretby close, Carine, Aus-Carine 6020, Australia. E- tralia, E-mail: lpitt@rsm.nlmail: merchantrec-ords.com Leyland Pitt is Professor of

Marketing at the ErasmusRian Van der Merwe Graduate School of Busi-recently completed a Ph.D. ness, Rotterdam School ofat Curtin University of Management, and AdjunctTechnology, Australia. His Professor of Marketing inprevious publications and the Ph.D. programme in e-current research centers on commerce, Lulea University

Internet marketing, electronic commerce and strategic of Technology, Sweden. He is widely published.alliances.

PIERRE BERTHON,Department of Marketing,Bentley College, 175 ForestStreet, Waltham, MA02452, USA. E-mail: pber-thon@bentley.edu

Pierre Berthon is Professorof Marketing at Bentley Col-lege. His research focuses onelectronic commerce, mar-keting information pro-

cessing, organization and strategy and managementdecision making. A number of his published papershave won awards in the US and UK.

European Management Journal Vol. 22, No. 1, pp. 12–26, February 200426