Open Configuration: a New Approach to Product Customization Linda L. Zhang1 and Xiaoyu Chen*1, 2 and Andreas Falkner3 and Chengbin Chu2 Abstract.1 State-of-the-art product configuration enables components in addition to the predefined ones. In another word, it companies to deliver customized products by selecting and cannot configure customized products in a true sense, i.e., to the assembling predefined configuration elements based on known full extent that it covers all reasonable and unforeseen customer relationships. This paper introduces an innovative concept, open requirements. configuration, in order to assist companies in configuring products This study proposes an innovative concept ‘open configuration’ that correspond exactly to what customers want. Superior to product configuration, open configuration involves both predefined in order to help companies configure such products that can meet configuration elements and new ones in configuring customized both predefined and unforeseen customer requirements, that is, to products. As a first step, this study explains the concept of open meet customer requirements as complete as possible without configuration and the basic principles. It also discusses in detail the making too much compromise (see Section 2). In this regard, in challenges involved in open configuration, such as conceptual configuring customized products, open configuration deals with model development, open configuration optimization, and open not only the addition of new configuration elements, such as configuration knowledge representation. functions, components, but also the modification of existing configuration elements, more specifically components. Existing component modification is to accommodate the integration of new 1 INTRODUCTION components with the predefined ones. With the advancement of design and manufacturing technologies, In the rest of this paper, Section 2 uses a fridge configuration customers are no longer satisfied with standardized products. They example to illustrate the limitation of product configuration, i.e., increasingly demand products that could satisfy their individual the product configured lie in a known range in accordance with the needs. As a result, companies need to timely offer customized predefined components. Section 3 introduces the concept of open products at affordable costs to survive [1]. With traditional design configuration, its basic principles, and its process. Section 4 sheds approaches, companies cannot efficiently develop customized lights on the challenges involved in open configuration. We end the products [2, 3]. Product configuration has been proposed to enable paper in Section 5 by pointing out the ongoing research that we are companies to deliver customized products at low costs with short working on. delivery times. Product configuration has been widely applied to a variety of industries, including computer, telecommunication systems, transportation, industrial products, medical systems and 2 PRODUCT CONFIGURATION services [4]. It brings companies a number of advantages in As a special design activity, product configuration capitalizes on delivering required products. These advantages include managing design results, such as components, attributes and their product variety [5], shortening delivery time [6], improving relationships [9, 10]. It entails such a process that based on given product quality [7], simplifying order acquisition and fulfilment customer requirements, suitable components are selected from the activities [8], etc. set of predefined component types; the selected components are Product configuration has received much attention from evaluated and further arranged into products according to the industrial and academia alike. Researchers have approached configuration constraints and rules. product configuration from different perspectives and have Take fridge configuration as an example. Assume in this developed diverse methods, methodologies, approaches, and example, there are 6 component types, including Refrigerator (R), algorithms to solve different configuration issues and problems. In Freezer (F), Freezer drawer (Fd), Variable compartment (V), Base spite of the diversities among these solution tools, they are (B), Outer casing (O). Each component type is defined by a set of developed based on a common assumption: the configuration attributes (number, size, price) and each attribute can assume a elements, such as components, modules, attributes, functions, and number of values. Table 1 summarizes these component types, the their relationships are predefined. In relation to this assumption, the attributes, and attribute values. products that can be configured are known in principle even if not For example, N R : (1, 2) represents the number of Refrigerators explicitly listable [2]. In this regard, product configuration cannot in one fridge can be 1 or 2; SR : (small, medium, large, extra-large) deal with such products that demand new functions and indicates the component Refrigerator has four different sizes: small, medium, large, extra-large. Price mentioned hereinafter 1 IESEG School of Management (LEM-CNRS), Lille-Paris, France states the price of the configured fridge. 2 Ecole Centrale Paris (Laboratoire Genie Industriel), Paris, France 3 Siemens AG Österreich, Vienna, Austria * Corresponding author: x.chen@ieseg.fr Table 1. The attributes of the fridge components. {R :1,large; NF  1;min P} . The configured fridge must satisfy these Component types Number Size Price constraints and additional rules mentioned earlier while fulfilling small, medium, the customer requirements. In this regard, the constraints Refrigerator 1-2 depending on size large, extra-large {R :1,large} and {N F  1} limit the possible choices to: {FCc , FCe } , small, large, i.e., the configuration solutions shown in Figs. 1.c and 1.e. The cost Freezer 0-1 depending on size extra-large constraint {min P} indicating the minimal price results in the final P(Fd) ( i.e., a fixed Freezer drawer 0-2 small solution to be FCc  {R :1,large ; F :1,small ; B :1,standard ; O :1, price) standard } . Variable P(V) (i.e., a fixed 0-1 small As only predefined elements are involved, product compartment price) Base 1 standard, wide depending on size configuration fails to provide customized products in a true sense Outer casing 1 standard, wide depending on size or provides these products which can meet unforeseen customer requirements. Take the above fridge configuration as an example. Suppose that the requirements from another customer include any There are relationships among components, among attributes, of the following: and between components and attributes. For examples,  a fridge consisting of only one medium refrigerator, {SR  large, NF  1}  {SF  small} means if one large sized  a fridge consisting of 2 freezers, Refrigerator and one Freezer are selected, the size of the Freezer is  an outer casing with a special color, and small; NFd  0  {NR  2, SR  medium} states that if the component  a cheaper fridge to be moved easily and with at least one Freezer drawer is selected then two medium Refrigerators are freezer drawer. required. The other relationships include: {SR  medium, NF  0}  In general, the first two requirements violate some predefined N R  2 ; {SF  small, SR  small}  NV  1 ; {SR  extra-large, NF  1}  constraints (although the first one requires a new - lower - type of {SF  extra-large} ; {SF  extra-large}  {SB  wide, SO  wide} ; outer casing as a side-effect); the last two introduce new concepts. {NV  1,NF  0}  {NR  1, SR  large,SV  small} ; {S R  small, N F  1}  In more detail, the third requirement requires a new attribute value {S F  large} . for the component outer casing. The last one is more complex. A There are four additional rules, including (1) ( NR  NV  NF )  3 , part of it, i.e., being cheaper and with one freezer drawer, can be meaning the total number of Refrigerator, Variable compartment, fulfilled by the predefined functions and components, while the and Freezer in one fridge should be no more than 3, (2) rest cannot be fulfilled by the available functions, thus calling for a NR  2  NV  NF  0 , indicating if two Refrigerators are selected, new function: ‘to be movable’. This new function, in turn, needs the number of Freezer and Variable compartment is zero, (3) new components, such as ‘wheels’, ‘brakes’, etc., which are N Fd  N F  0 representing that Freezer cannot be selected together necessary for delivering this function. Because of the lack of these with Freezer drawer, and (4) N Fd  NV  0 indicating that Freezer components, product configuration can provide the customer with drawer cannot be selected together with Variable compartment. one of the fridges shown in Fig. 1 without satisfying all his According to the above pre-defined components and their requirements. The customer, thus, has to accept this fridge by relationships, only 17 fridge configurations are available as making compromise (e.g., accept a cheapest fridge with a freezer possible solutions. While Fig. 1 shows 8 fridge configurations due drawer, which cannot be moved easily). to the space issue, different positions of components in Fig. 1.c, Fig. 1.d, Fig. 1.e, Fig. 1.f, and Fig. 1.g lead to the other 9 fridge configurations. All customized fridges to be configured based on 3 OPEN CONFIGURATION customer requirements fall into this range of configuration solutions. (Note: Fridges from the left to the right are arranged In order to help companies configure customized products that based on the increase of price.) Take fridge f in Fig. 1 as an correspond exactly to what a customer requires, this paper puts example to explain the components and their attributes in the forward the concept of open configuration. The basic principle and configuration solution. This fridge configuration is represented as general process of open configuration are introduced below. FC f  {R :1,small ; V :1,small ;F :1,small ;B :1,standard ; O :1,standard} . It has one small Refrigerator on top, one small Variable compartment in the middle, one small Freezer at the bottom, one 3.1 Open configuration concept standard Base, and one standard Outer casing. Built on top of product configuration, open configuration is to configure customized products to meet customer requirements in a true sense. Similar as product configuration, it utilizes design Variable Refrigerator Refrigerator Freezer compartment Freezer Refrigerator results, selects components, and arranges the selected components Freezer Refrigerator Variable Refrigerator according to constraints and rules. In extension to product compartment Refrigerator Refrigerator Refrigerator Refrigerator configuration, it involves new component design, more specifically Refrigerator Freezer the specification of functions and the selection of the Freezer drawer corresponding components. In addition, it deals with the a b c d e f g h modification of the predefined components, which allows the integration of new configuration elements. Figure 1. Fridge configuration solutions Suppose the requirements from a customer include a cheaper 3.2 Open configuration overview and process fridge with a freezer and a large refrigerator. In accordance with Open configuration involves two types of knowledge: predefined these requirements, the constraints can be modeled as knowledge and dynamic knowledge. Predefined knowledge relates to predefined functions, components, and relationships; dynamic carried out. The selected components are arranged into product knowledge is associated with newly defined elements. In relation configuration alternatives by following the product structure to these customer requirements, which can be fulfilled by the described in the dynamic and predefined knowledge. These predefined functions (i.e., Type Ⅰ requirements in Fig. 2), the configuration alternatives are further evaluated under certain corresponding components are selected, while for these criteria. Based on the evaluation results, the optimal one or requirements, which cannot be fulfilled by the predefined functions multiple are suggested to customers. (i.e., Type Ⅱ requirements in the figure), new functions and corresponding components are specified. The specification of these new configuration elements contributes to the extension of the 4 CHALLENGES INVOLVED IN OPEN dynamic knowledge. The relationships among the predefined CONFIGURATION elements and the newly defined elements are specified as well. In accordance with the involvement of new configuration elements, This specification contributes to the interaction between the open configuration changes the basic assumptions and reasoning predefined knowledge and the dynamic knowledge. By respecting processes of product configuration. In this regard, there are a the constraints embedded in both the predefined and dynamic number of potential challenges involved in open configuration. knowledge, all necessary components are selected, modified, and Due to the page limitation, this paper discusses five of these arranged into a customized product. challenges, including open configuration modeling, system design and development, open configuration solving, open configuration Type Ⅰ Predefined optimization, and open configuration knowledge representation. requirements knowledge Customer Customized requirements products 4.1 Open configuration modeling Type Ⅱ Dynamic requirements knowledge Open configuration modeling addresses the modeling of open configuration knowledge and the reasoning mechanism for using the configuration knowledge. The modeling of open configuration Figure 2. Open configuration overview knowledge is to model configuration elements, constraints, and rules. It involves two kinds of knowledge: predefined knowledge In more detail, suppose that given customer requirements are and dynamic knowledge. A product model and corresponding valid, complete and do not conflict with one another. These functional architectures should be developed for defining and requirements are evaluated first to determine whether or not they further classifying the two different types of knowledge. The can be fulfilled by the available configuration elements (i.e., modeling of the reasoning mechanism is to shed light on (1) how functions and components). According to the evaluation results, new functions are specified, (2) how new components are these requirements are classified into Type Ⅰ and Type Ⅱ determined, and (3) how components are selected and arranged requirements. Fig. 3 summarizes this process. into products. Components In open configuration modeling, the components and functions Type Ⅰ requirements selection are characterized by their attributes, while the inter-connections Components Yes modification Final components selection Components arrangement among the components are represented by connections and ports. Customer All required The modeling of the dynamic knowledge needs to take into requirements evaluation configuration elements available Configured alternatives evaluation account the fact that new functions and components are added No based on the unforeseen customer requirements. Thus, its modeling New New Customized Type Ⅱ requirements functions components specification products involves newly-added concepts, constraints, and rules. The specification modeling of the predefined knowledge needs to consider these predefined components, modified components, and their Figure 3. Open configuration process relationships. The interaction between predefined knowledge and dynamic knowledge needs to be modeled as well. For Type Ⅱ requirements, new functions are specified and all Open configuration modeling is more sophisticated than possible components which can realize these functions are configuration modeling due to the involvement of the dynamic subsequently determined. Also specified are the relationships knowledge. In this regard, it is interesting to see whether or not among functions, among components, and between functions and these techniques which are suitable for modeling product components. This process contributes to the extension of the configuration (e.g., Unified Modeling Language (UML), Alloy, dynamic knowledge. For Type Ⅰ requirements, all possible and generative Constraint Satisfaction Problem (CSP) [11]) can be components are selected from the predefined ones. In addition, to used to model open configuration. If these techniques are feasible, be compatible with the newly introduced components, some how can they be modified or adjusted to model open configuration. predefined components are modified by respecting constrains and If these techniques are not feasible, new modeling formalisms and rules embedded in the predefined and dynamic knowledge. This constructs are to be developed. process reflects the interaction between the dynamic and predefined knowledge. From the modified components, newly introduced components, and selected predefined components, 4.2 System design and development suitable components are further selected for forming configuration System design and development for open configuration refers to alternatives, which can meet customer requirements. In the the design and development of the computer information system to selection, consistency and compatibility evaluations might be implement open configuration, i.e., open configurators. Open configurators consist of a customer input module which deals with and configuration rules. To solve this model, algorithms need to be customer requirements evaluation, open configuration knowledge developed subsequently. bases, reasoning and evaluation mechanisms, optimization and In the situation that customer requirements demand new diagnosis mechanisms, and an output module which communicates functions, the dynamic knowledge will be specified. The modeling the configuration results with users. Different from product of open configuration problem will associate with the interaction configurators, open configurators involve two knowledge bases: a between the customer requirements and two types of knowledge knowledge base for the predefined knowledge and the other for the (predefined knowledge and dynamic knowledge). The main dynamic knowledge. Joint reasoning mechanisms between the two difficulties are (1) the modeling of new function specification, (2) knowledge bases are required, which mainly associate with the modeling of new components selection according to the interacting and integrating elements from the two knowledge bases. customer requirements, (3) and the modeling of the interaction For the dynamic knowledge base, new elements design modules between new components and selected existing components. After are needed to develop and maintain this knowledge base. The new modeling an open configuration problem, suitable algorithms need elements design modules include the module for specifying new to be developed to solve the model. Because of the differences functions with respect to the requirements, the module for selecting between product configuration and open configuration and the new components to fulfill new functions and the module for corresponding differences between a product configuration model interfacing with the predefined elements. For the predefined and an open configuration model, these algorithms, which are knowledge base, different from product configurators, there need suitable for product configuration solving, may not be applicable to be a modification module for modify existing components to be for open configuration solving. Thus, new algorithms are to be compatible with the new ones. developed. In designing and developing open configurators, the techniques should have the ability to model dynamic knowledge and the interaction between dynamic knowledge and predefined 4.5 Open configuration optimization knowledge. In this regard, the available system design techniques During each step of open configuration, optimal functions, for product configuration may need to be modified in designing components and structures need to be specified from a number of and developing open configurators. alternatives. The dynamic feature of open configuration increases the degree of difficulty in optimizing the new functions, new components, and the interaction between new components and 4.3 Open configuration knowledge predefined ones. In this regard, an explicit optimization mechanism representation needs to be developed. Open configuration knowledge representation entails the effective In accordance with the open configuration process discussed organization of open configuration knowledge, including the earlier, the optimization mechanism should evaluate the predefined and dynamic knowledge. It logically uniforms the open configuration elements at three levels. In the first level, the configuration knowledge and enables the utilization of the mechanism should evaluate all the possible function alternatives knowledge in different configuration tasks. for fulfilling Type II requirements and decide on the optimal ones. The representation of open configuration knowledge includes This optimization might be based on, e.g., the performance and the representation of predefined components, relationships, completeness of these function alternatives. In the second level, the constraints and rules; the representation of newly-added mechanism should evaluate all the possible component alternatives components, relationships, constraints and rules; and the for delivering the determined new functions and decide on the representation of the constraints and relationships between optimal ones. This optimization may take into account, e.g., the predefined knowledge and newly added knowledge. From the compatibility among the new components and the interaction with experience of the knowledge representation for product predefined components. In the third level, the mechanism should configuration, open configuration should be considered as both a evaluate all the product configuration alternatives and decide on classification problem (i.e., capturing the aspects of taxonomy and the optimal ones. This optimization may consider, e.g., product topology) and a constraint satisfaction problem (i.e., capturing the reliability. aspects of constraints and resource balancing). Considering the dynamic and indeterminate feature of open configuration, it might be potentially challenging to capture different aspects of open 5 CONCLUSION configuration knowledge (e.g., taxonomy, topology, constraints, In response to the limitation of product configuration, this paper and resource balancing) in one model. Further studies may try to proposed open configuration to help design customer-driven design new models (or sub models to be embedded in the available product in a true sense. It introduced the concept and process of tools) separately on each aspect and joint them together to open configuration. It also discussed several challenges involved in represent the knowledge. open configuration. Currently, we are working on the formulation of open configuration. In the formulation, new components, relationships among new components, and relationships between 4.4 Open configuration solving new components and existing components will be defined and Open configuration solving relates to the development and modeled. This formulation is to rigorously define open application of algorithms or other tools to solve open configuration configuration and shed light on the reasoning behind open problems. In solving an open configuration problem, the problem configuration. needs to be modeled first with respect to customer requirements REFERENCES [1] M. Heiskala, K.S. Paloheimo, and J. Tiihonen, Mass customization of services: benefits and challenges of configurable services, 206-221, Proceedings of Frontiers of e-Business Research, Tampere, Finland, 2005. [2] D. Sabin and R. Weigel, ‘Product configuration frameworks-a survey’, IEEE Intelligent Systems and Their Applications, 13(4), 42- 49, (1998). [3] S. Schmitt and R. Bergmann, Applying case-based reasoning technology for product selection and customization in electronic commerce environments, 42-48, Proceedings of the 12th International Bled Electronic Commerce Conference, Bled, Slovenia, 1999. [4] A. Trentin, E. Perin, and C. Forza, ‘Product configurator impact on product quality’, International Journal of Production Economics, 135(2), 850-859, (2012). [5] C. Forza and F. Salvador, ‘Managing for variety in the order acquisition and fulfilment process: the contribution of product configuration systems’, International Journal of Production Economics, 76(1), 87-98, (2002a). [6] A. Haug, L. Hvam, and N.H. Mortensen, ‘The impact of product configurators on lead times in engineering-oriented companies’, Artificial Intelligence for Engineering Design, Analysis and Manufacturing, 25(2), 197-206, (2011). [7] A. Trentin, E. Perin, and C. Forza, ‘Overcoming the customization- responsiveness squeeze by using product configurators: Beyond anecdotal evidence’, Computers in Industry, 62(3), 260-268, (2011). [8] C. Forza and F. Salvador, ‘Product configuration and inter-firm co- ordination: an innovative solution from a small manufacturing enterprise’, Computers in Industry, 49(1), 37-46, (2002b). [9] S. Mittal and F. Frayman, Towards a generic model of configuration tasks, 1395-1401, Proceedings of the 11th International Joint Conference on Artificial Intelligence, Detroit, USA, 1989. [10] D. Brown, ‘Defining configuration’, Artificial Intelligence for Engineering Design, Analysis and Manufacturing, 12(4), 301-305, (1998). [11] A. Falkner, A. Haselböck, G. Schenner, and H. Schreiner, ‘Modeling and solving technical product configuration problems’, Artificial Intelligence for Engineering Design, Analysis and Manufacturing, 25(2), 115-129, (2011).