Electronic Commerce: The End of the Beginning
`13th International Bled Electronic Commerce Conference
`Bled, Slovenia, June 19-21, 2000
`
`A Customization Approach for Structured Products in
`Electronic Shops
`
`Armin Stahl, Ralph Bergmann, Sascha Schmitt
`
`University of Kaiserslautern, Department of Computer Science
`Artificial Intelligence – Knowledge-Based Systems Group
`67653 Kaiserslautern, Germany
`Phone: ++49-631-205 3362, Fax: ++49-631-205 3357
`{Stahl, Bergmann, SSchmitt}@informatik.uni-kl.de
`
`Abstract
`
`Customers of electronic shops find more and more support for the search and selection of
`products in the sales systems. Unfortunately, most of the shops do not provide additional support
`with parameterizable or configurable products. Such products could be further customized. One of
`the major problems most customization techniques suffer from is that they require large
`knowledge acquisition effort, which leads to problems in the rapidly changing e-Commerce
`scenario. In this paper, we present a new approach to customization that is particularly suited to
`e-Commerce applications. It assumes that products can be structured hierarchically into sub-
`components. Customization is achieved by incrementally replacing unsuitable sub-components
`through recursively finding best-matching alternative sub-components, using Case-Based
`Reasoning technology for this search process. The presented approach avoids huge portions of the
`knowledge acquisition effort. The approach is implemented as a prototypical system.
`
`1.
`
`Introduction
`
`A sales system that offers products that can be modified to some degree has to provide its
`customers with the possibility to further customize the base products, i.e. to enable the customer to
`tailor the base products according to her or his wishes. Examples of appropriate products, i.e.
`customizable products, are technical equipment like computers [12], designs for electrical
`engineering [10], holiday trips, service products like insurances or investment plans, etc. For this
`paper, we limit our considerations to the customization of complex technical systems.
`Customization is especially important when complex products with a large number of possible
`variants must be supported during sales [7]. Product customization can be realized by the known
`customization techniques. However, the applicability of the different techniques strongly depends
`on the kind of products to be supported, in particular on the number of different variants. A general
`problem of most approaches to customization is that they require a large effort for acquiring the
`customization knowledge, to be represented, e.g., by rules or by operators [7]. This turns into a
`real problem when products have many different variants or if the product spectrum changes
`rapidly so that the customization knowledge must be updated. Other approaches require a
`complete problem solver for product recommendation and sufficient knowledge for this problem
`solver. Such approaches also suffer form intractability, both in terms of computational efficiency
`
`Page 1 of 38
`
`FORD 1106
`
`

`
`Armin Stahl, Ralph Bergmann, Sascha Schmitt
`and knowledge acquisition effort. Hence, for customizing complex products with a large number
`of possible variants, the existing customization approaches are often not suitable in practice.
`In this paper, we present a new approach to customization that is particularly suited to electronic
`commerce applications. This approach avoids huge portions of the knowledge acquisition effort of
`the previous approaches. It assumes products that are structured into sub-components, possibly in a
`hierarchical manner. The knowledge required is knowledge about available pre-configured
`complete products as well as knowledge about available sub-components. Both kinds of
`knowledge are easily available in an electronic commerce setting. After retrieving the best pre-
`configured product with respect to the customer's requirements, the product is customized by
`incrementally replacing sub-components by more suitable sub-components. These new sub-
`components are determined by recursively applying CBR, i.e. similarity-based product retrieval,
`on the level of the sub-components [1, 7]. In the remainder of this paper, this approach is described
`in more detail. We will first describe a typical e-Commerce scenario and analyze the shortcomings
`of existing approaches. Then the general idea of recursive CBR is introduced before it is
`specialized for the purpose of product customization. Finally, we report on the current state of the
`implementation.
`
`2.
`
`Product Customization in e-Commerce Applications
`
`Within the set of possible products, a continuum of products can be identified (see Fig. 1) [12]. It
`classifies different products according to their ability to be customized by a customer. Generally,
`we distinguish between constant and variable products. Constant products are products which
`cannot be modified by the customer. The product is fixed in such a case. Variable products may be
`customized via product parameterization or product configuration. We differentiate between
`different kinds of such a customization.
`
`Covered
`Products
`
`Complexity
`
`Fixed
`Products
`
`Parameterizeable
`Products
`
`Configurable
`Products
`
`Innovative
`Product
`Design
`
`Creative
`Product
`Design
`
`Fig. 1. The Continuum of Products.
`
`2.1
`
`General Product Categories
`
`Fixed Products: At the lower end of this continuum we have fixed products. The product cannot
`be modified and as a result, the sales assistant cannot customize the product. Examples are music
`CDs, books, integrated circuits, etc., or a single computer monitor, where the product is
`completely fixed.
`Parameterizeable Products: Next on
`find products which are
`the continuum, we
`parameterizeable by certain values. These values may be discrete, like the color of a good and also
`they may be continuous values, like the capacity of a storage device. The sales assistant may
`calculate these values or he may use given or existing ones during the sales process. However, the
`product may only be modified by the instantiation of one or more parameters concerning the
`product.
`Configurable Products: Next, we have configurable products which consist of a set of predefined
`components and the knowledge about how components can be connected. Further, all available
`components must be known and also all the knowledge how components are allowed to be
`connected. During the sales process the sales assistant has to configure the product for the
`customer.
`
`2
`
`
`
`Page 2 of 38
`
`FORD 1106
`
`

`
`A Customization Approach for Structured Products in Electronic Shops
`We will not cover innovative product design which produces an artifact significantly different
`from existing ones or even creative product design products which produce a new type of artifact.
`Examples for such tasks are the design of complete assembly lines, a one-family house, or other
`complex products which do not have a very similar prototype in the past.
`
`2.2
`
`Case-Based Product Recommendation: The Scenario
`
`During recent years, the technology known under the term Case-Based Reasoning (CBR) has
`become a successful tool to realize product customization. The main idea behind CBR is the
`assumption that it should be possible to use experiences of the past, called cases, to solve actual
`problems. In an e-Commerce scenario, the cases are represented by the available products or their
`descriptions respectively. The actual problem that has to be solved is given through a set of
`customer demands, i.e. the customer formulates his needs and wishes on a searched product. We
`consider the scenario in which a customer enters a virtual shop to buy a complex technical system.
`Complex technical systems (e.g., PCs) can usually be decomposed into a set of different
`components. We suppose that the shop offers a set of pre-configured standard systems as well as
`the different individual components.
`If the shop provides an intelligent product recommendation agent the first step of the sales process
`is a demand acquisition phase in which the customer states his individual demands on the searched
`product. The result of this demand acquisition phase can be formulated in form of an incomplete
`product description (on a technical level) which we call query. This query is then used to start a
`similarity-based retrieval in order to determine one of the available pre-configured standard
`systems that fulfills the demands as well as possible. However, because of the large number of
`possible product variants, the retrieved product does usually not fulfill the demands exactly.
`Hence, to be able to present the customer a satisfying result, it is necessary to customize the
`product, e.g., by replacing some components by more suitable ones [9]. This is the task of the
`adaptation phase of CBR. In the following, we review several well-known approaches to
`adaptation in CBR and discuss whether they are appropriate for the customization of complex
`products.
`
`2.3
`
`Existing Approaches
`
`Customization Rules and Customization Operators. The knowledge how existing solutions can
`be adapted on actual problems is encoded explicitly. Adaptation rules [3] consist of a set of
`preconditions and a set of actions. Dependent on the evaluation of the preconditions, the actions
`are able to modify a retrieved case to a new target case with respect to the given query.
`Customization operators [7] are very similar to adaptation rules. While the rules will be executed
`after the retrieval automatically, the operators provide more control by the customer. Therefore,
`the retrieved case with an additional set of applicable operators will be presented to the customer.
`If the retrieved product does not fulfill his demands, he can repeatedly choose operators to change
`the given product until he gets a satisfying result.
`The general problem of both approaches is the necessity to define every possible case modification
`that may occur explicitly in the form of preconditions and actions. Therefore, even simple domains
`often require a large number of customization rules or operators to cover all customization
`possibilities. For really complex domains, the huge amount of necessary rules or operators
`prevents the application in practice.
`Configuration From-Scratch. The configuration of a product can also be performed from-scratch
`by classic configuration systems [4] without using CBR. Applying such a system for the
`customization of products in an e-Commerce application often leads to some disadvantages. First,
`if it is impossible to configure a product that fulfills all customer demands exactly, such a system
`rarely finds a suitable alternative solution. Second, in classic configuration systems it is often
`difficult to handle optimality criteria. Therefore, the system can only present any solution (if
`existing), but a high quality of this solution cannot be guaranteed in general. An additional
`problem of the from-scratch configuration is the time-critical aspect, i.e. the configuration process
`usually takes a lot of calculation time. This is often not acceptable in e-Commerce applications.
`
`
`
`3
`
`Page 3 of 38
`
`FORD 1106
`
`

`
`Armin Stahl, Ralph Bergmann, Sascha Schmitt
`Derivational Adaptation. The common property of customization rules and operators is that they
`are used to transform an existing solution to a new, hopefully better, solution. Thus, this approach
`is also called transformational adaptation [13]. Another approach is known under the term
`derivational adaptation. Here, the cases do not represent concrete solutions. Instead, they contain a
`description of the problem solving trace which describes how a concrete solution is generated.
`This known solution trace can then be used by a from-scratch problem solver to generate a
`solution very quickly. Generally, this approach is coupled with the same disadvantages as
`configuration from-scratch, except for better performance. A description of configuration systems
`using derivational adaptation can be found in [6, 5].
`The described approaches for realizing configuration of complex products in e-Commerce
`applications are all coupled with more or less big problems or significant disadvantages. In the
`remainder of this paper, we will present an alternative approach for realizing product
`customization in complex and highly structured domains. A classic example for such a domain is
`the configuration of personal computers.
`
`3.
`
`Product Customization by Incremental Similarity-Based
`Adaptation
`
`To illustrate the functionality of our approach we consider the configuration of personal
`computers. However, the presented approach can be applied for the customization of any
`structured products.
`
`3.1
`
`Domain Representation
`
`Structure of Products. In general, it is possible to represent complex structured products by an
`object-oriented domain model with an additional aggregation hierarchy [11]. In the following, we
`assume a special form of such a compositional structure like shown in Fig. 2.
`
`
`
`Fig. 2. An Example Compositional Structure.
`
`In this structure, the root node represents the whole technical system to be configured, i.e., in our
`example, a complete personal computer. The leaf nodes represent the concrete parts that a PC
`consists of, like the hard-disk, the processor, the CD-ROM, and so on. In a concrete PC, these
`parts are realized by concrete technical components. For example, the processor part can be
`realized by a Pentium-III with 500MHz. The inner nodes of the compositional structure (if
`existing) do not correspond to concrete parts of the technical system. They rather represent abstract
`nodes used for aggregating related parts into kinds of abstract subsystems (e.g., Base-System or
`Storage-System).
`Query. The starting point of the configuration process is the query represented by an incomplete
`instantiation of the compositional structure. When looking at the example query shown in Fig. 3,
`we can interpret the root note as our actual problem, i.e., we are searching a PC with a set of
`special properties. To reach this goal it is necessary to select appropriate components that fulfill
`the properties of the respective part-queries. In our example, one part-query states that the PC
`
`4
`
`Page 4 of 38
`
`FORD 1106
`
`

`
`A Customization Approach for Structured Products in Electronic Shops
`should have a hard-disk with 12GB of capacity. To fulfill this demand we can, e.g., integrate the
`concrete hard-disk ``Maxtor91303D6'' for the hard-disk part in the PC. Thus, we can interpret the
`different leaf nodes of the query as sub-problems that must be solved to solve the overall problem,
`i.e., the configuration of the required PC. If we have found suitable sub-solutions, i.e. suitable
`components, for every part-query, we have to combine these sub-solutions to a final solution for
`the overall configuration problem.
`
`
`
`Fig. 3. An Example Query.
`
`Similarity Measure. To retrieve appropriate PCs from a product database (in the CBR context the
`case base), the assumed CBR-system uses special domain-dependent similarity measures. To
`compute the similarity between a query and a given product we assume a bottom-up strategy using
`the defined compositional structure. First, it is necessary to compute the similarities between the
`part-queries and the corresponding product parts. These local similarities are used for determining
`the global similarity between the query and the product. A more detailed description of such a
`similarity computation can be found in [2].
`Dependencies between Components. Generally, the decomposition of a configuration problem
`like discussed above is coupled with a basic problem. If we decompose a configuration problem
`into the sub-problems of finding suitable components it is not sufficient to handle these sub-
`problems absolutely isolated. The reason are dependencies between the different sub-problems or
`their solutions respectively. In the described application domain such dependencies occur in form
`of technical constraints between the different components. For example, it is impossible to
`combine every kind of processor with any mainboard model because the respective interfaces must
`fit together. To be able to handle these technical restrictions during the configuration process, the
`domain model must include a formal description of the existing constraints. For our approach, we
`suppose the existence of a special constraint system that is able to check whether the constraints of
`a given configuration are fulfilled or not.
`
`3.2
`
`The Configuration Framework
`
`We now describe an approach to realize product customization by a kind of incremental similarity-
`based adaptation. Fig. 4 shows the respective process model that we will discuss in the following.
`Basically, the whole configuration process can be subdivided into two major steps, called base
`product retrieval and adaptation cycle. The task of the first step is the similarity-based retrieval of
`the best available base product from the respective product case-base. The second step is an
`iterative procedure which performs the necessary adaptation of the retrieved base-product if it does
`not fulfill all customer demands.
`
`
`
`5
`
`Page 5 of 38
`
`FORD 1106
`
`

`
`Armin Stahl, Ralph Bergmann, Sascha Schmitt
`
`Base Products
`
`Query
`
`Base Product
`Retrieval
`
`Base Product
`
`Part
`Selection
`
`Weak Part
`
`Validated Product
`
`Components
`
`Component
`Retrieval
`
`Final Product
`
`Product
`Validation
`
`Alternative
`Components
`
`Adaptation
`Cycle
`
`Component
`Exchange
`
`Modified Product
`
`process
`
`product
`
`case base
`
`domain knowledge
`
`control flow
`modification retraction
`information flow
`
`Constraints
`
`Quality
`Measure
`
`Fig. 4. The Configuration Process.
`
`At the beginning of the adaptation cycle, it has to be determined if the base product includes parts
`that do not fulfill the requirements of the query. This can be done by an examination of the local
`similarities between the different product parts and the related part-queries. Product parts with a
`low similarity to the respective part-query are called weak parts. The determination of these parts
`is the task of the part selection process. If the product includes no weak parts at all, the adaptation
`process is finished and the given product is presented to the customer as suggested customization
`result. Generally, we can distinguish between two causes for weak parts. The first one occurs if the
`retrieved standard product does not include components for required product parts. For example, if
`a customer wants a PC with a modem, but the base-product includes no modem at all. Here, we
`can speak of a missing component. The second cause is given if installed components do not fulfill
`the technical requirements of the customer, e.g., if the PC still contains a modem but this modem
`provides only an insufficient data transmission rate. Then, we can speak of a weak component.
`During the adaptation cycle, a selected weak part leads to a new similarity-based retrieval, called
`component retrieval. In this process, the part-query corresponding to the selected weak part is used
`to retrieve a collection of alternative components from the component case-base. The elements of
`this collection are arranged by their similarity to the query in a decreasing order.
`The task of the next process is then the replacement of the weak component (if existing) by the
`first element of the collection of alternative components. If the weak part corresponds to a missing
`component, the first alternative component is introduced as a new component for this part.
`Because of the mentioned order, this alternative component represents the most suitable
`component that is available within the component case-base. The result of the described
`component exchange is a modified product.
`To judge the effects of the component exchange, the modified product must finally be inspected
`during the product validation process. The task of this process is the validation of the modified
`product with respect to two different aspects. On the one hand, it must be guaranteed that the
`modification leads to a consistent product, i.e., to a product fulfilling the technical constraints of
`the respective application domain. Therefore, the assumed constraint system has to check the
`modified product. On the other hand, the modified product should naturally represent a better
`solution as the product before the modification. Therefore, we have to compare the similarity to
`the query of these two products. Either if the modified product is inconsistent or if it has a lower
`
`6
`
`
`
`Page 6 of 38
`
`FORD 1106
`
`

`
`A Customization Approach for Structured Products in Electronic Shops
`similarity to the query as the product before the modification, retracting the last modification
`becomes necessary. The retraction of the last modification is realized by the repeated application
`of the component exchange process. This means, the just now installed component that had led to
`the failure of the product validation is replaced by the next element of the alternative components
`collection. If no available alternative component is suitable to fulfill the mentioned criteria, it
`becomes necessary to restore the situation at the beginning of the current iteration of the adaptation
`cycle. In this case, an adaptation of the respective weak product part is currently impossible.
`If the product validation could be passed successfully, one iteration of the adaptation cycle is
`finished. The result of this adaptation cycle is the validated product. In the next step, it is again the
`task of the part selection process to determine a weak part within this validated product which
`shall be used as the starting point for a new iteration. If it is impossible to determine a weak part
`whose adaptation could perhaps improve the product, the complete configuration process is
`succeeded and the final product can be presented to the customer.
`
`3.3 Controlling the Adaptation Cycle
`
`After giving an overview of our configuration framework, we will now show that the part selection
`process (see Fig. 4) is the most important step of the adaptation cycle. This process is mainly
`responsible for the overall success of the customization procedure. Generally, we can notice that
`the adaptation cycle has to solve a combination of a constraint satisfaction problem (CSP) and an
`optimization problem. On the one hand, it has to find a combination of different components
`representing a working product, i.e., a product that fulfills all constraints of the application
`domain. On the other hand, it has to find an optimal combination fulfilling the demands of the
`customer as well as possible. To solve this special task, it is not sufficient to determine weak
`product parts during the part selection process in any order. It is rather necessary to find adaptation
`orders which will lead to an optimal configuration result. Generally, we can distinguish between
`two different approaches to determine adaptation orders:
`Preservation of Consistency. Up to now, we have assumed that a product modification during
`one iteration of the adaptation cycle must even lead to a consistent product. This means, only
`adaptation orders that preserve the consistency of the product are allowed.
`Temporary Loss of Consistency. In contrast to the previously described approach it is also
`possible to allow the temporary loss of consistency during the adaptation of a base product. That
`means, if the product validation process determines the violation of constraints after the
`component exchange process, this must not necessarily lead to an immediately retraction
`procedure. It is rather possible only to notice the loss of consistency and to continue the whole
`adaptation process without the retraction of the last modification. But, to find a final solution that
`can be presented to the customer it is of course necessary to re-establish the consistency during the
`further adaptation.
`The advantage of the first approach is the quite simple implementation. Besides that, it represents
`an any time algorithm, i.e., it allows the interrupt of the adaptation cycle at any time to obtain the
`best consistent solution that could be formed so far. Nevertheless, with preservation of consistency
`it is often impossible to get optimal configuration results. The reason are the different constraints
`between the components. For example, if we want to install a SCSI-hard-disk we have perhaps
`also to replace the IO-controller. Therefore, the integration of the SCSI-hard-disk will lead to an
`inconsistent PC before the necessary IO-controller is installed, too. If we allow a temporary loss of
`consistency during the adaptation we are able to handle such problems. But, this requires more
`complex algorithms to guarantee the final consistency of the configured product.
`
`4.
`
`Conclusions and State of Implementation
`
`In our opinion, the most important benefit of the presented approach is the capability to realize
`case-based configuration without requiring a lot of additional domain knowledge. In contrast to the
`
`
`
`7
`
`Page 7 of 38
`
`FORD 1106
`
`

`
`Armin Stahl, Ralph Bergmann, Sascha Schmitt
`application of customization rules/operators, our approach needs only the knowledge about
`existing components and the constraints between them. It is not necessary to describe every
`thinkable adaptation-step. The selection of adaptation-steps is controlled by the knowledge that is
`encoded in already existing similarity measures. The additional case information, i.e., the
`description of the available components, is not critical because it is usually accessible in existing
`product databases. Because the presented approach can obviously be classified as transformational
`adaptation, it is also coupled with the respective advantages: It is able to configure products if the
`customer demands are not satisfactory and it can handle implicit optimality criteria through the
`cases that are assumed to be good pre-configured standard products.
`The applicability of our approach is limited if it is deployed for domains with a huge number of
`dependencies between the different components. Then, the adaptation cycle often has to retract
`adaptation steps after product validation because the selected component does not fit into the given
`product due to the given technical restrictions.
`To evaluate the functionality of our configuration approach we have implemented a generic
`prototype for the described configuration process. This prototype consists of an extension of the
`commercial CBR tool CBR-Works which has been developed jointly by the University of
`Kaiserslautern and tec:inno GmbH. To be accessible over the World Wide Web the prototype also
`provides two respective interfaces for the demand acquisition (see Fig. 5) and the result
`presentation.
`
`Fig. 5.
`
`Implemented demo version.
`
`In the current version, we have modeled the already mentioned PC-domain as a first test domain.
`However, in the future it could be useful to apply the system on other application domains to get
`more concrete experiences about the power of the approach. Because of the structure of the
`selected PC-domain, presently we have only implemented a simple adaptation control algorithm
`preserving the consistency of the retrieved product after every adaptation step.
`Therefore, the next step should be the implementation and the evaluation of different more or less
`sophisticated adaptation control strategies to improve the customization process.
`
`8
`
`
`
`Page 8 of 38
`
`FORD 1106
`
`

`
`A Customization Approach for Structured Products in Electronic Shops
`Acknowledgements
`
`Funding for this research has been partially provided by the Commission of the European
`Communities (ESPRIT contract EP 27.068, the WEBSELL project - Intelligent Sales Assistants
`for the World Wide Web). The partners of WEBSELL are tec:inno (prime contractor, Germany),
`IMS (Ireland), IWT Magazin Verlags GmbH (Germamy), Adwired AG (Switzerland), Trinity
`College Dublin (Ireland), and the University of Kaiserslautern (Germany).
`
`References
`
`1. A. Aamodt and E. Plaza. Case-Based Reasoning: Foundational Issues, Methodological
`Variations, and System Approaches, in AICOM, Vol. 7, No. 1, 1994, pp. 39–59. 1994.
`2. R. Bergmann and A. Stahl. Similarity measures for object-oriented case representations. In B.
`Smyth and P. Cunningham (Eds.): Case-Based Reasoning Research and Development, volume
`1488 of Lecture Notes in Artificial Intelligence, pp. 25-36, Springer, 1998.
`3. R. Bergmann, W. Wilke, I. Vollrath, and S. Wess. Integrating general knowledge with object-
`oriented case representation and reasoning. In 4th German Workshop: Case-Based Reasoning -
`System Development and Evaluation, 1996.
`4. R. Cunis, A. Günther, and H. Strecker (eds.). Das PLAKON-Buch. No. 266 in Informatik-
`Fachberichte, Springer, 1991.
`5. H. Holz. Erweiterung eines Konfigurationssystemes um Techniken des fallbasierten Schließens.
`Diploma Thesis, University of Kaiserslautern, Germany. 1996.
`6. K. Pfitzner. Fallbasiertes Konfigurieren technischer Systeme. Dissertation, University of
`Hamburg, Germany. 1993.
`7. S. Schmitt and R. Bergmann. Applying Case-Based Reasoning Technology for Product
`Selection and Customization in Electronic Commerce Environments. Proceedings of the 12th
`Bled Electronic Commerce Conference, Bled (Slovenia), June 7-9, 1999.
`8. S. Schulz. CBR-Works - A State-of-the-Art Shell for Case-Based Application Building.
`Proceedings of the 7th German Workshop on Case-Based Reasoning, GWCBR'99, Würzburg,
`Germany, March 3-5, 1999.
`9. U. J. Timm and M. Rosewitz. Electronic Sales Assistance for Product Configuration.
`Proceedings of the 11th Bled Electronic Commerce Conference, Bled (Slovenia), June 8-10,
`1998.
`10.I. Vollrath. Reuse of Complex Electronic Designs: Requirements Analysis for a CBR
`Application. In B. Smyth, P. Cunningham (Eds.): Advances in Case-Based Reasoning,
`Proceedings of the 4th European Workshop, EWCBR-98. Springer, 1998, pp. 136-147. 1998.
`11.S. Wess. Fallbasiertes Problemlösen in wissensbasierten Systemen zur Entscheidungsunter-
`stützung und Diagnostik. Dissertation, University of Kaiserslautern, Germany. 1995.
`12.W. Wilke. Knowledge Management for Intelligent Sales Support in Electronic Commerce.
`DISKI 213, Infix Verlag, 1999.
`13.W. Wilke and R. Bergmann. Techniques and knowledge used for adaptation during case-based
`problem solving. In 11th International Conference On Industrial and Engineering Applications
`of Artificial Intelligence and Expert Systems, IEA-98, 1998.
`
`
`
`
`
`9
`
`Page 9 of 38
`
`FORD 1106
`
`

`
`THIRTEENTH BLED
`ELECTRONIC
`COMMERCE
`CONFERENCE
`
`“ELECTRONIC COMMERCE:
`THE END OF THE BEGINNING”
`Bled, Slovenia, June 19 - 21, 2000
`
`CONFERENCE
`Bled, June 19 - 21
`THE GRADUATE STUDENTS CONSORTIUM
`AND EDUCATIONAL SYMPOSIUM
`Otoèec, June 22 - 23
`
`Additional information on the program
`and registration is available at:
`http://eCom.fov.uni-mb.si
`
`Page 10 of 38
`
`FORD 1106
`
`

`
`CONFERENCE SPONSORED BY
`
`University of Maribor, Faculty of Organizational Sciences
`Slovenia’s Electronic Commerce Project
`Republic of Slovenia, Ministry of Science and Technology
`Chamber of Commerce and Industry of Slovenia
`
`
`
`
`
`
`
`Page 11 of 38
`
`FORD 1106
`
`

`
`PREREGISTRATION, Sunday, June 18
`
`15.00 – 18.00 Convention Centre
`14.00 – 18.00 IFIP TC8 Workshop, Hotel Toplice
`19.00
`
`Hotel Golf
`Dean's Reception and Dinner (for Conference Speakers only)
`Jože Florjančič, Professor and Dean
`Faculty of Organizational Sciences, University of Maribor
`
`CONFERENCE, June 19-21
`
`08.00 – 17.30 Registration at Convention Centre
`07.30 – 18.30 Internet and E-mail Service, Convention Centre
`Sponsored by: SRC.SI - Systems Integration
`
`
`Compaq and Telekom Slovenije
`08.00 – 09.00 Refreshments, Informal Discussion, Convention Centre
`12.45 – 13.45 Lunch, Hotel Park
`
`Monday, June 19
`
`
`MEETINGS
`
`
`Centers for the Study & Research in Electronic
`Commerce - Directors Meeting
`Co-chairs:
`Donald J. McCubbrey, Director
`Center for the Study of Electronic Commerce
`University of Denver, USA
`Beat F. Schmid, Director
`Institute for Information Management
`University of St. Gallen, Switzerland
`
`9.00 - 12.30
`
`PARK
`
`
`
`
`
`Page 12 of 38
`
`FORD 1106
`
`

`
`14.00 - 17.30
`
`PARK
`
`
`KOMPAS
`(JEZERSKA
`HALL)
`
`
`Graduate Programs in Electronic Commerce
`Chair: Paula M.C. Swatman, Professor of Electronic
`Commerce & Innovation Leader in Electronic Commerce
`RMIT University, Australia
`Co-chairs:
`Claudia Loebb