SFB 403 Discussion Paper (97-2)

Communication Services Supplied by Intermediaries in Information Networks: The EDI Example
Falk von Westarp, Sascha Weber, Peter Buxmann, Wolfgang König
Research Project "Competitive Advantage by Networking"
Project B3 "Economics of Standards in Information Networks"
Institut für Wirtschaftsinformatik
J. W. Goethe-Universität
Mertonstr. 17, 60054 Frankfurt am Main
Telephone: + 49 69 798-23318
Fax: + 49 69 798-28585


Information network participants face the decision problem of choosing appropriate communication standards for all of their bilateral communication links. In addition, the question arises whether or not it is more efficient to conduct data or information exchange internally, or whether the responsibility should be outsourced to an intermediary. Using the EDI example, this paper analyzes the possible efficiency increases resulting from the use of such intermediaries and introduces a model to solve the decision problem for a centrally coordinated network. The model is based on linear programming. For a given information network, it determines the optimal set of communication standards for each participant, the optimal number of intermediaries, as well as their optimal service range.



1. Introduction

Most economic activities of todayís world are influenced by the exchange of data or information. The smooth functioning of such exchanges is only made possible by using communication standards. Participants generally reach agreements on such standards in bilateral negotiations. Standardization organizations, corporate associations or individual companies continuously develop new communication standards to improve communication in information networks. An example is the EDI standard EDIFACT, a universal language used for exchanging trade documents, directly developed from the requirements and possibilities of electronic data processing. EDI standards, which originated in the Seventies, are becoming increasingly popular. On account of positive network effects, growing numbers of EDI users lead to increased cost savings potentials by using these standards. Due to the dynamics of such developments, it is necessary for each participant in an information network to evaluate the advantages of EDI implementations repeatedly over the course of time. Next to the choice of implementing the standard or not, alternative actions exist. Currently, in the economic world increased numbers of institutions are emerging that supply the market with services for translating data of various communication standards into an EDI standard (and vice versa). Examples of this are, among others, debis (a fully-owned subsidiary of the Daimler Benz Corporation), the Deutsche Telekom AG, the Deutsche Post AG, IBM, and GEIS (General Electric Information Services). The question is therefore not merely whether an EDI implementation is economically viable, but rather to additionally consider whether services should be outsourced to an external company.

In order to develop an instrument for solving the decision problem, we begin by examining the significance of communication standards for coordinating economic activities. We consider the decision to implement standards as an isolated coordination problem. Using the EDI example, we subsequently point out the potential advantages and disadvantages connected to changing communication standards. By including independent EDI service providers, we present our problem as a typical make-or-buy problem. For this purpose, in section 4 we analyze the efficiency increases made possible by external service businesses in information networks. Our evaluated results serve as the basis for developing a model to coordinate standardization decisions in a centrally coordinated information network. The model is based on linear programming. We conclude the paper with an illustration of our results by optimizing an exemplary network.


2. Communication Standards and the Coordination of Economic Activities

In todayís world based on the division of labor, organizations are faced with the problem of efficiently coordinating economic activities internally and externally. The basis of coordinating individual activities is the exchange of information, i.e. communication between participants (humans or machines). Information is transmitted through signals by using appropriate media. The content of signals must be interpreted identically by the transmitter and the receiver. To ensure communication, however, uniform regulations must be established which determine the signals, the communication media, and the signal interpretation. We refer to such regulations, e.g. technical terminology, communication protocols (TCP/IP) or standards for structuring the content of data (EDIFACT), as communication standards.

Participants of information networks, i.e. networks primarily based on communication, have a choice of different communication standards. The selection of such standards is subject to the efficiency rule, that is, weighing the costs against the benefits. Their implementation and use, on the one hand, induce resource consumption and, on the other hand, cost savings potentials. Selecting the optimal set of communication standards is therefore viewed as a separate coordination problem, referred to as the standardization problem [Buxmann 1997].

The need for coordination arises whenever interdependencies exist between activities [Malone/Crowston 1994]. Implementing and using individual communication standards represent such economic activities. Interdependencies between these and other activities result from the budget constraint. If the budget is invested in the implementation of a specific standard, these resources cannot be used for any other purpose. Next to the interdependency in the context of resource shortage, the standardization problem additionally considers interdependencies between activities due to so-called positive network effects. That is, the benefits a person or an organization derives from using a communication standard also depend on the extent to which it is used by other participants of the information network [Katz/Shapiro 1985]. An example is the e-mail technology which provides a quick and low cost exchange of information. The more communication partners are accessible with this standard, the higher the potential cost and time savings. Therefore, the evaluation of own activities depends on the activities of others.

To solve the standardization problem of an information network, the above mentioned interdependencies must be taken into consideration. The combination of communication standards for each individual node, which is determined for given volumes of exchanged data, induces the cost optimum of the entire network. However, the optimality of a determined solution is continuously questioned over time, since changes in attributes of decision parameters occur. The worldwide continuously decreasing telecommunication rates are a good example of this. In addition, the amount of data to be transmitted is consistently increasing. Furthermore, new communication standards constantly extend the existing set, and must be examined with regard to their efficiency. Since such future developments are difficult to predict, it is especially important to review the validity of the existing solutions when new standards emerge. Within this context, EDI standards are presently an especially innovative area. Using the EDI example, we begin by identifying and quantifying the decision parameters to solve the described coordination problem. This supplies the basis for the decision model we will subsequently introduce.


3. Decision Determinants of an EDI Implementation

EDI is the exchange of business data (e.g. delivery notes, invoices) between two application systems in a standardized, automated form. The goal is to achieve a more efficient data and information management by reducing processing time and avoiding redundant data entry. EDI standards structure the content of messages with regard to syntax, semantics, and pragmatics by uniformly defining message types and components [Kilian et al. 1994, pp. 42-43]. The EDIFACT standard provides an example. For a description of the United Nations EDIFACT specifications see "http://www.premenos.com/unedifact/". EDI is not concerned with the technical requirements of data exchange, such as those listed in the transfer protocols. These are determined by other communication standards like SMTP, X.400 and FTAM.

A coordinating unit faces alternative courses of action in making the decision to implement an EDI standard. On the one hand, the internal data management can be converted to the new technology. On the other hand, it can outsource the entire data management, or at least a part of it, to an external, independent service provider which offers EDI services, e.g. debis. Thus, in addition to deciding on which standard to use, the problem of a typical make-or-buy decision must be solved.

Table 1 summarizes the benefits and costs of company-provided EDI services [Emmelhainz 1993, Kilian et al. 1994, Neuburger 1994].

The benefits are primarily derived from achieving efficiency goals, i.e. cost reductions induced by rationalization and automation measures. On the one hand, costs that depend on the amount of data transferred can be reduced. Examples are avoiding redundant data entry and error rate improvements. On the other hand, cost savings can be induced that do not directly depend on the data volume, for example, by improving the value-added chain (just-in-time management, etc.). The same differentiation applies to the costs incurred for the implementation and use of EDI. For example, data transmission represents a variable figure, while the costs of hardware and software are for the most part independent from the amount of transferred data. Subsequently, we generalize the advantages and assume the benefits derived from EDI usage consist of reduced variable costs of transmitting data. The costs of implementing the EDI standard will in the following be referred to as set-up costs or standardization costs.

Benefits of EDI
Costs of EDI
  • Avoid redundant data entry
  • Reduced resource consumption (personnel, inventory, storage, etc.)
  • Reduced error rate of data entry
  • Reduction of non-value time
  • Reduced uncertainty in order cycle time
  • Improved supply chain management
  • Reduced transaction costs
  • Contact costs
  • Monitoring costs
  • Costs of hardware and software
  • Personnel costs
  • Consultant costs
  • Costs of data communication
  • Transaction costs
  • Dependency on dominating partners
  • Security problems
  • Incompatibility to previous business partners (insider-outsider problem)
  • Table 1. Benefits and costs of EDI

    In case the EDI services are outsourced to an external company, the high implementation costs of standards can be saved, without restricting communication to other nodes. Additional benefits can be derived from specific functions of external services, which will be discussed in the subsequent section. On the cost side, next to low reorganization expenditures for enabling communication to the EDI service provider, costs are mainly incurred charges depending on the data volume of the purchased services [Emmelhainz 1993, pp. 109-117]. The following chapter analyzes potential efficiency increases resulting from engaging an external service business.


    4. Intermediaries in Information Networks

    The communication standards summarized under the general term EDI are characterized by relatively high set-up costs, together with lower variable costs of data transmission [Emmelhainz 1993]. Experience shows that a critical limit exists with regard to the amount of data, after which the EDI is economically viable. Due to the high dynamics of the development of new communication standards and, in the same context, the short life-cycle of standards, the set-up costs must be relatively quickly compensated by variable cost savings. Therefore, especially for smaller companies with small amounts of data, the problem is whether or not such a change is worthwhile. Transaction costs can present a further barrier. These can result from companiesí concerns of losing their independence or existing business connections. In the past, for example, the automotive industry showed how market dominating manufacturers pushed their own EDI standard onto their suppliers. In complex information networks such as the automotive sector, this leads to a problem we refer to as the insider-outsider problem. Supposing a supplier joins the communication standard of one of its customers, it automatically becomes more difficult to support the already existing communication links to his other customers and suppliers. The reason is that sustaining the technology of old standards continues to incur costs dependent on the amount of data, however, these are then allocated to a smaller amount of data. As an example, one can imagine what it means to continue keeping up the telex standard for communicating with some partners. In addition, as with all standards, the uncertainty still remains if, and to what extent, the EDI standard will prevail in the business world. This leads to the fact that network effects that, as we already mentioned, influence the benefits of EDI, can only be estimated.

    EDI induces significant cost savings potentials, however, its implementation is costly and combined to different uncertainties. Therefore, institutions have emerged that mediate between organizations with EDI and those without. Such intermediaries provide each information network participant with the EDI advantages, without the participants having to implement the standard themselves.

    We generally define intermediaries as independent mediators between associated (networked) participants. They transform data outputs of one participant into the appropriate data inputs of the other by eliminating incompatibilities between output and input attributes. A communication standard regulates the specific characteristics of a specific amount of data attributes. Transforming data from one standard into another is therefore nothing more than adapting incompatible data attributes. By describing the transformation process in the subsequent section, the services and functions of intermediaries will be elaborated.


    4.1. Transforming technical and organizational attributes

    The main issue of transforming technical and organizational data attributes is to uncouple both the transmitter and the receiver in terms of time and technical aspects. For direct data exchange between two business partners, the transmitterís computer system reaches the receiverís system, for example, by using a telephone line. Both systems must be compatible. This refers to technical attributes, on the one hand, such as the transmission speed determined by the communications protocol. On the other hand, organizational attributes, such as the systemís operating time, must be synchronized for data transmission. Suppose a manufacturer has several customers, possibly using different computer or communication systems, maybe even located in different time zones, then the result is a complex, costly system of direct point-to-point connections.

    Assuming an intermediary takes over basic responsibilities, such as receiving, storing, and forwarding electronic data, then many of the connections necessary for direct communication can be eliminated. In a formal approach, this relation can be derived from the law of contact cost reduction [Balderston 1958, Baligh/Richartz 1967]. An information network with N participants, each sustaining communication links to every other participant, results to N(N-1)/2 communication links. By employing an intermediary, the number can be reduced to N. Figure 1 illustrates the connection.


    Figure 1. Contact reduction


    For N>3 a reduction of edges results for the whole communication system, independent of the amount of data to be transmitted; the number of communication contacts decreases correspondingly. The problems of synchronization in terms of time and technical compatibility are reduced. However, the intermediary must be capable of supporting the technical and organizational conditions of each participant in order to coordinate the reception and forwarding of data.

    Other services offered by intermediaries of information networks are connected to data security. In exchanging business data, EDI users or users of other communication standards are often concerned with the security of sensitive data. In addition to authenticity, integrity, and confidentiality [Europäische Kommission 1997], a significant issue is system security. An intermediary specialized in forwarding data generally has a more solid know-how in the area of data security than individual users, and furthermore has improved technical possibilities to take any necessary security measures.


    4.2. Transforming content attributes

    Next to services in the technical area, intermediaries of EDI are primarily concerned with transforming content attributes of data. Two forms can be differentiated. Format transformation consists of the intermediary translating incoming electronic data with company-specifically formatted contents into the structure of an EDI standard and forwarding the data to a business partner using EDI. Vice versa, after implementing EDI in the own organization, an intermediary can be used to communicate with existing business partners in the previously used format. This reduces the insider-outsider effect. In cases where the transmitter and the receiver are using completely different storage or transferring mediums, an intermediary additionally conducts a medium transformation. Syntactic, semantic and pragmatic elements are determined and applied to the structure of the new medium. The presence of both technical and content attributes becomes clear, since the storage medium is also used for the physical data transfer. Medium transformation enables EDI data to be forwarded to the according receiver by mail, fax, e-mail, or telephone. An example for a supplier of such services is the Deutsche Post AG. The opposite case is much costlier, but also possible.


    4.3. Other potential efficiency increases

    In addition to the potential efficiency increases mentioned above, intermediaries in information networks can economies of scope as well as economies of scale. The latter are especially attainable by grouping data while processing different services. Examples are clustering advantages for storage or transfer capacities or transferring grouped data during times of low network rates. In such a competitive market, the savings are passed on to the customer through lower prices.

    By determining his service range, an intermediary decides for which communication standards and for which storage mediums he wishes to transform data into which EDI standards. All of these standards must be supported in his own organization and, additionally, he must have the appropriate transformation instruments available. The choice of service range depends on the consumers, whose needs the intermediary aims at satisfying.


    5. Optimal Number of Intermediaries in Information Networks and their Optimal Service Range

    In the previous sections we described participantsí decision criteria of an EDI implementation. We now introduce a model which optimizes the use of communication standards in a centrally coordinated information network. Such a model for coordinating standardization decisions in consideration of the buy alternative must illustrate the above mentioned criteria. The goal is to determine the optimal selection of communication standards for each participant as well as the optimal number of intermediaries and their range of transformation services for a given information network. As a basis, we begin with a model which optimizes standardization in an information network without considering intermediation [Buxmann 1996, Buxmann 1997]. The basic idea of the model is that by implementing a new communication standard for all or some of the participants, set-up costs must be borne. However, variable communication costs can be reduced, which can lead to an efficiency increase for the entire network. In order to consider services of intermediaries in such a model, the diversion and transformation of data and the costs involved must additionally be illustrated.


    5.1. Formal approach

    The information networks considered can formally be described as nodes connected by edges. The nodes represent participants (organizations, humans, or machines) faced with the decision of extending their existing set of communication standards by installing one or several more. The costs incurred for implementing new standards are assigned to the nodes. The edges represent the communication links between participants. The data flow along the edges, transmitted by using a specific standard, and the connected variable transfer costs are edge-related figures. Each node can perform intermediary functions if it is equipped with the standards to be transformed. In the business world, this assumption is confirmed by companies using company-internal know-how to supply data transfer services on the market. An example is debis, which, in addition to handling the data communication of its parent corporation Daimler Benz, supplies the service externally. If data is not directly transferred from a transmitting node to a receiving node, i.e. if an intermediary is used, a series of edges is passed through. For the transformation of data from one standard into another, variable transaction costs are incurred for the node conducting intermediary functions. These are the charges to the respective intermediary, dependent on the amount of data transferred. In the United States, for example, these costs range between $0.20 and $0.75 per 1,000 characters [Emmelhainz 1993, pp. 112-117]. Implementing a standard in a node is only worthwhile if the sum of variable cost savings outweighs the set-up costs. A trade-off therefore exists between node- and edge-related costs. A similar economic efficiency consideration is also made for the buy alternative. The costs of a company internal implementation are weighed against the intermediary costs, i.e. the costs of having data transformed into the specific standard. A third alternative is, of course, maintaining the status quo of a node. Since coordinating standardization activities centrally seeks a solution optimal for the entire network, all possible constellations of standardization decisions of each participant must be compared. This shows that positive network effects are considered, since every possible constellation of such effects is implicitly enumerated in the solution process and therefore included in the comparison of alternatives. The interdependencies described in section 2 (budget constraint and network effects) are thus incorporated in the model.

    Our model optimizes an existing information network, consisting of N nodes with S available standards to choose from. The participants are already equipped with certain standards. When a node i (iÎ {1,...,N}) is equipped with a standard s (sÎ {0,...,S-1}), costs of stancostsis are incurred. The amount of information transferred in the communication link between nodes k and l (k,lÎ {1,...,N}) is indicated by flowkl. A flow along the edge between nodes i and j (jÎ {1,...,N}) incurs variable costs of comcostsijs, dependent on the standard s being used. If an information flow transmitted in node i is transformed from standard s into standard t (or vice versa), variable costs of trancostsist result. Nodes and edges are not subject to any capacity restrictions, i.e. no quantitative restrictions apply for redirecting or transforming data flows.

    The action variables of the model are, on the one hand, the binary variable XisÎ {0,1}, assuming the value 1 if node i is equipped with standard s. On the other hand, the variable FijklsÎ R0+ (with i¹ j and k< l), indicates the data flow transmitted from node k as the sender to node l as the receiver (subsequently referred to as "communication link (k,l)"), passing through edge (i,j) on its way to the target node.

    The binary variable YijklsÎ {0,1} (with i<j and k<l) serves as an auxiliary variable of the model. It indicates whether a flow takes place along edge (i,j) using standard s for the communication link (k,l). Another auxiliary variable, TiklstÎ {0,1} (with k<l and s<t), indicates whether in node i a flow of the communication link (k,l) is transformed from standard s into standard t (or vice versa).

    The model is static, i.e. fixed costs must be compensated by potential variable cost savings within a specified service life time of the standard considered. This assumption is based on the already mentioned dynamics in connection with the development of new standards, which make long-term planning unrealistic. The (first best) cost optimum is determined from an overall viewpoint of the network.

    The following target function determines the solution of the coordination problem for centrally coordinated networks:


    The constraints of the linear problem are illustrated in the appendix. The solution resulting from the optimization can be used to determine the optimal total costs of the information network, the use of communication standards in each node, as well as the standards and the series of edges of each data transfer. In addition, the nodes functioning as intermediaries, the extent of data transformation, as well as the intermediaryís range of standards can be determined. The following section illustrates an example of optimizing a given network.


    5.2. Example of optimizing a network

    To illustrate the scope of the model described, we conclude with an example of optimizing a given network. Figure 2 depicts an information network in its initial state.
    Figure 2. Initial state of the network


    To simplify, we demonstrate a network with five participants (individuals, organizations or machines), represented by nodes 1 through 5. Determined data flows take place along the communication links between the participants. The determined amount of data is listed along the edges. For example, 900 data units are exchanged between participant 1 and participant 4. In the initial state of the network each node is equipped with a standard Z. Node 2 is additionally equipped with EDI, symbolized by the shaded area in figure 2. The entire communication of the network is based on the standard Z, since node 2 does not have a partner to communicate with using EDI. Variable costs are incurred along the edges for exchanging information. Table 2 lists their amounts for both standards.


    volume of data
    var. comcosts (Z)
    var. comcosts (EDI)
    Table 2. Variable edge-related figures


    The exchange of information between nodes 3 and 4, for example, incurs costs of 15,000* 3=45,000 monetary units. The costs of the initial state therefore total 77,900 monetary units for the data exchange. At the same time, these costs represent the total costs of the information network, since no costs are incurred for implementing a new standard.


    Figure 3. Optimum without intermediation

    Figure 3 presents the optimal standardization solution of the network without intermediaries. Based on the situation described above, we evaluate whether it is economically viable to implement EDI in the individual nodes. Table 3 lists the respective standardization costs incurred.


    stancosts (EDI)
    var. trancosts
    Table 3. Node-related costs


    The standardization costs of node 2 of course equal zero, since EDI is already available. Nodes 1 through 4 are equipped with EDI in the optimum, and use this standard to communicate. Standardization costs totaling 16,500 monetary units result. Node 5, on the other hand, continues to communicate with its partners using standard Z. The costs of transferring data amount to 44,200 monetary units, and therefore total costs of 60,700 monetary units result.

    Figure 4 shows the optimal solution of the network under consideration of intermediation. Considering the initial state, each node can take over intermediary functions. Table 3 lists the edge-specific variable costs incurred for transforming data.


    Figure 4. Optimum with intermediation


    In the optimal solution of this example, less nodes implement the EDI standard than in the solution without an intermediary. Only nodes 1 and 4 are additionally equipped with EDI. Node 1 serves as the intermediary. It directs data flows between nodes 3 and 4 and nodes 4 and 5. Along edge (1,4), for example, flow34=15,000 data units results, that is assigned to the communication link between nodes 3 and 4. The intermediaryís service consists of transforming data from standard Z into the EDI standard, and vice versa. The participating nodes are charged a total of (15,000+5,000)* 0.1=2,000 monetary units in transformation costs. It is clear that in this example, even with redirecting data, lower total costs result for the entire information network.


    6. Summary and Concluding Remarks

    To optimize communication in information networks individual standardization decisions must be coordinated. Using the EDI example, this paper demonstrates how intermediaries can increase efficiency. The decision model we introduced solves the standardization problem in centrally coordinated information networks in consideration of the possibility of outsourcing. It illustrates the relevant cost parameters and incorporates interdependencies between economic activities which underlie the coordination problem. Therefore, it serves as an effective instrument to support the decision making process of such problems. For a given network the optimal solution reveals which participant should provide which intermediary services. In addition, the optimal standard and the optimal path of each data flow is determined.

    Our model provides various points for future extensions. Possible extensions could include dynamic aspects as well as introducing edge and node capacities. Furthermore, the effects on the overall network of entirely new intermediary nodes emerging could be examined. With the rapid developments of the internet in mind, we are also currently examining the extent to which the results can also be applied to decentralized networks.


    7. Appendix

    Constraints of the linear problem:


    (Un-)Balanced Data Flow in Nodes:

    For each communication link (k,l) an information flow flowkl is transmitted from node k into the network (1) as well as an information flow being received from every node l (2).


    (1)  (with k<l)

    (2)  (with k<l)


    For each node (i¹ k) and (i¹ l) the sum of all incoming flows of the communication link (k,l) must equal the sum of all outgoing information flows of this link.


    (3)  (with i¹ k and i¹ l and k<l)


    Node Standardization as a Constraint of Yijkls<0

    Low communication costs can only be achieved between two nodes i and j if both nodes are using the same standard s.


    (4)  (with i<j and k<l)


    Indication of Necessary Transformations by Yijkls

    Inequation (5) ensures that Yijkls assumes the value one as soon as a flow of the communication link (k,l) using the standard s flows along edge (i,j).



    (with i<j and k<l and M³ )


    Activating Transformation Costs

    Transformation costs must only be borne if a flow of the communication link (k,l) passing through node i is changed from standard s to standard t (or vice versa).



    (with i¹ k and i¹ l and k<l and s<t)


    8. Bibliography

    Balderston, F. E. (1958). Communication Networks in Intermediate Markets. Management Science, Vol. 4, pp. 154-171.

    Baligh, H.H. / Richartz, L.E. (1967). An Analysis of Vertical Market Structures. Management Science, Vol 10, pp. 667-689.

    Buxmann, P. (1996). Standardisierung betrieblicher Informationssysteme, Wiesbaden.

    Buxmann, P. (1997). The Standardization Problem. Proceedings of the Symposium on Operations Research, Jena.

    Emmelhainz, M.A. (1993). EDI - A Total Management Guide, 2nd ed., New York.

    Europäische Kommission (1997). Sicherheit und Vertrauen in elektronische Kommunikation. Ein Europäischer Rahmen für digitale Signaturen und Verschlüsselung, KOM (97) 503, Brüssel.

    Katz, M.L. / Shapiro, C. (1985). Network Externalities, Competition, and Compatibility. The American Economic Review, Vol. 75 (3), pp. 424-440.

    Kilian, W. et al. (1994). Electronic Data Interchange (EDI), Baden Baden.

    Malone, T.W. / Crowston, K. (1994). The Interdiciplinary Study of Coordination. ACM Computing Surveys, Vol. 26 (1), pp. 87-119.

    Neuburger, R. (1994). Electronic Data Interchange. Einsatzmöglichkeiten und ökonomische Auswirkungen, Wiesbaden.