modelling electronic data interchange through ... - Semantic Scholar

1 downloads 27655 Views 34KB Size Report
simulation models for business process analysis. INTRODUCTION. A multitude of change management concepts, methodologies, and tools have been ...
MODELLING ELECTRONIC DATA INTERCHANGE THROUGH SIMULATION: AN INDUSTRY-WIDE PERSPECTIVE George M. Giaglis Department of Computer Science and Information Systems Brunel University Uxbridge, Middlesex UB8 3PH United Kingdom E-mail: [email protected] KEYWORDS Business Process Modelling (BPM), Business Process Re-Engineering (BPR)

• Continuous Process Improvement (CPI) (Harrington 1991) • Total Quality Management (TQM) (Oakland 1993) • Organisational Transformation (OT) (Adams 1984)

ABSTRACT Business process modelling (BPM) is an emerging field of simulation application. Simulation of business processes is increasingly being recognised as an extremely useful tool for assessing the feasibility, efficiency, and effectiveness of process re-engineering decisions. However, almost all of the reported experience is confined to re-engineering initiatives within a single organisation. The problems that seem to arise in more complex situations where re-engineering extends beyond the boundaries of a single organisation to include multiple businesses in the value chain, seem to be quite different and more difficult to overcome. In this paper, we present the development of a simulation model for the assessment of the expected benefits from the implementation of EDI across an industrial sector. The model simulates trading between a number of companies at three sequential stages of production. The simulation keeps track of inventory levels in order to assess the efficiency gains on inventory management introduced by EDI. Based on this study of interorganisational process simulation, we identify a set of requirements for the effective application of simulation as a decision making tool for inter-organisational BPM. Finally, we derive some general conclusions that were obtained from the particular study and indicate some areas for further research on the development of simulation models for business process analysis. INTRODUCTION A multitude of change management concepts, methodologies, and tools have been developed during recent years by management scientists and consultants to help enterprises in their efforts to adapt in a constantly changing business, social, and technological environment. The most popular of these approaches include: • Business Process Re-engineering (BPR) (Hammer 1990, Davenport and Short 1990)

Each approach differs significantly in the scope and range of the anticipated changes, the management tools utilised to achieve change, and the business contexts in which they can be used. However, they all have in common that they require businesses to model the ways in which they currently operate, to identify opportunities for change, and to design and implement alternative ways of carrying out business processes. In view of the above, Business Process Modelling (BPM) has recently received widespread attention and has been acknowledged as an integral part of any change management project. Different tools and techniques have been proposed for BPM, and simulation has been identified as an extremely useful tool for this purpose (see for example Giaglis and Paul 1996, Giaglis et al 1996, Tumay 1995, Swami 1995, Bhaskar et al 1994). Despite the availability of these tools, it has been reported that companies are usually facing significant practical problems when trying to model in detail the way they operate (Hansen 1994) or to implement changes in existing environments (Galliers 1994). These problems can become very significant in large, complex organisational settings, especially in cases where more than one business is involved (interorganisational systems). For example, Business Process Re-engineering projects, where inter-organisational processes almost always play an important role (see for example Riggins and Mukhopadhyay 1994), have been reported to have a large proportion of failures in practice. In this paper, we start by presenting a case study, in order to clarify some of the issues that may arise in inter-organisational business modelling. The case study refers to the modelling of Electronic Data Interchange (EDI) adoption patterns across a whole industrial sector and the impact these patterns have on inventory levels in each individual company and the

sector as a whole. Based on this study, we derive some general conclusions for inter-organisational business systems simulation and indicate some areas for further research. THE CASE STUDY: MODELLING EDI IMPACT ACROSS AN INDUSTRIAL SECTOR Electronic Data Interchange (EDI) can be defined as the inter-organisational, application-to-application exchange of business documentation in a standard automated way (Emmelhainz 1993). It is increasingly becoming a popular way of carrying out business transactions, due to the wide range of benefits that are ascribed to it, at both operational and strategic levels. The use of EDI is expected to be the dominant form of business communication between companies in several markets during this decade (Nygaard-Andersen and Bjorn-Andersen 1994). However, few and limited investigations of the conditions under which an EDI system is worth adopting have been reported (Premkumar et al 1994). EDI is one of the most widely used examples of Inter-Organisational Information Systems (IOS), i.e. systems that link different companies into a single information infrastructure. Due to its significance, EDI is an excellent example to illustrate the issues that might arise during the simulation modelling of inter-organisational business processes. The simulation-based technique that is briefly presented here aimed at assessing the expected benefits of EDI in three major Greek industries, namely pharmaceuticals, supermarkets and textile/clothing. This was part of a wider project the goal of which was to evaluate the impact of EDI in these sectors and to propose appropriate strategies for adoption. The main purpose of the simulation approach was to provide tangible and quantitative measures of the potential operational benefits of EDI on inventory reduction achieved by companies that use it. In general, the simulation system consists of: a. An abstract model of the industry structure and the trading rules between a number of companies in three sequential stages of the value chain for each of the three sectors. b. A set of operational research models tailored to the conditions and assumptions for each industry and each scenario. The major models used concern: • Inventory management • Production planning • Materials requirements planning • Demand forecasting Figure 1 depicts the model structure of one of the three industries (textile and clothing industry) to which this technique was applied as well as the operations carried

out by each type of participant. The model represents only the basic structure of a typical industry value chain. This abstraction was deemed necessary in order to reduce the analytical and computational complexity of the model. However, great effort was exerted in clearly specifying the distinct assumptions built into the model and in investigating their expected impact on the results. The basic assumptions made during model development are: • The different types of internal organisational structures have not been taken into account. • Only one kind of product is traded between the companies in each stage of the value chain. • The number and the market shares of the companies simulated follow the actual distribution of the number and the market shares of the companies in each industrial sector examined. • Time is measured in days. For simplicity reasons it is assumed that each month consists of thirty days, giving a total of 360 days per year. The simulation was run for a period of four years. • All companies in the model are assumed to use the same method for estimating the optimal level of inventory for a future period. This estimation is based on the Economic Order Quantity (EOQ) model for a fixed reorder cycle inventory with backordering. In the following paragraphs, the exact behaviour of the companies at each stage of the textile/clothing industry value chain (retailers, manufacturers, material suppliers) is described. For a more detailed description of the simulation model, see Mylonopoulos et al 1995. Retail Companies Each retail company faces a daily customer demand. This is derived from a daily normal distribution which is estimated on the basis of the real demand in the sector in conjunction with each company’s market share. Each retail company satisfies customer demand in full or in part, according to its current level of product inventory. Retail companies estimate the optimal level of product inventory for the current year in regular periods, based on the total expected customer demand. When the actual inventory level falls below a safety point, the company places an order to its suppliers (manufacturers) to refill the optimal inventory level. This placement is done by means of a merged plan containing all expected future order dates and quantities. It is assumed that each retailer buys merchandise from all available manufacturers at proportions commensurate to their market shares. Of course, the actual times and the respective quantities of the realised orders are not eventually the same as those contained in the plan due to stochastic consumer demand.

Materials Suppliers

Inventory Management (Materials)

Production Planning

Inventory Management (Materials)

Materials Requirements Planning

Production Planning

Retailers Inventory Management (Products)

Demand Forecasting

Figure 1: The Textile/Clothing Industry Value Chain and the models used

Manufacturing Companies Based on the order plans and on the actual orders sent by the retail companies, each manufacturing company schedules its optimal inventory levels. All plans received by retailers are merged into a single plan which represents the expected demand for each manufacturer. According to this plan each company calculates the optimal inventory level. When servicing actual orders, if product inventory is not enough to satisfy the whole demand, the remaining orders are placed into back-orders. Priority is given to older orders and, among orders of the same day, to bigger ones. Production scheduling is based on inventory scheduling. Two basic assumptions are made regarding the production process. First, there is not a constant flow of products, but all units comprising a production bunch become available simultaneously when their production is completed. Second, the productive capacity allocated to each production bunch is equal to its size for the whole duration of its completion. The production plan serves as the basic input for scheduling the materials inventory. Based on the assumption that a certain quantity of products needs an equal quantity of materials to be produced, optimal inventory levels and reorder points are calculated for the materials inventory. Order plans for materials are formed and sent to the suppliers in the same way as order plans for products were received by the retailers.

The distribution of the total order quantity to the various suppliers is again based on their relative market shares. Every time inventory levels fall below the reorder point, actual orders are also distributed to materials suppliers according to the same rule. Material Suppliers Material suppliers operate similarly to manufacturers. The only difference is that the model does not simulate the material inventory for these companies (i.e. the way material suppliers may communicate with their own suppliers). Because of that, the length of production for material suppliers is specified not only by their individual rates of production, but also by a random variable which expresses the materials’ lead times. Implementation and Experimentation The models were calibrated with statistical estimates of the actual values of various exogenous parameters such as final demand, market shares and so on. All these elements were encapsulated into a simulation programme written in Pascal programming language. The final deliverables of the programme are large amounts of data which are subsequently loaded into a spreadsheet for further processing and presentation. In brief, the program keeps record of variables such as inventory level, demand, sales and others for each simulated company over time. As part of a pilot implementation, this programme contains only the basic model structure and a number of measurements. Visual interactive capabilities and further data

processing and presentation tools were not a priority at that stage. Two basic scenaria were considered in the experimentation phase: one with all companies in the industry not using EDI and one with all companies being EDI users. The first scenario is the actual current situation in the sectors studied. The reason for including the real world situation in the scenarios was to provide an indication of the validity of the simulation model, as well as to serve as a basis for comparison with any other scenario. Of course, any intermediate scenario can be employed by simply changing the initial conditions of a simulation session. Results By running the simulation model, it was observed that the mean as well as the deviation of inventory levels were significantly reduced in all cases when companies in all industrial sectors were using EDI (Scenario B). Furthermore, it was clear that an overall reduction in inventory levels in the examined industry value chain was introduced. The actual reduction varied according to the type of the company as well as the company’s market share (bigger companies seemed to achieve relatively larger reductions in inventory levels). The average percentages of reduction were as shown in Table 1. Category Retailers’product inventory Manufacturers’product inventory Manufacturers’materials inventory Suppliers’materials inventory

Reduction 25-40% 20-30% 5-15% 10-15%

Table 1. Typical Levels of Inventory Reduction CONCLUSIONS: REQUIREMENTS FOR INTERORGANISATIONAL SYSTEMS SIMULATION Based on the experience gained by simulating business processes across an industry value chain (as opposed to modelling processes of a single organisation), we can derive the following general conclusions about requirements of inter-organisational BPM: • The degree of uncertainty is substantially increased with possible implications for the validity of the derived models. Modelling of inter-organisational processes must be exercised with great care to avoid such pitfalls and sensitivity analysis becomes an extremely important issue in this case. • Modelling requires extensive data collection and organisation in order to understand and structure the behaviour of the system under study. In the case of inter-organisational modelling, such data might not be available or easy to collect, so modellers usually have to rely on additional assumptions that might further jeopardise the validity of the business

model. • The possible multiplicity of decision making levels might be another source of additional difficulties for inter-organisational BPM. When for example, a simulation model is jointly developed by more than one companies (as was the case presented above), users will want to assess both their individual performance (i.e. single-site level output analysis), as well as the performance of the system as a whole (i.e. aggregate output analysis). This calls for the development of simulation models that can readily satisfy this requirement and therefore be helpful to users during the decision making process. • To add to this, decision making for individual firms becomes extremely difficult in situations where multiple players are involved, since the decisions made by one firm are affected by the (uncontrollable) behaviour of outside parties and the relevant model interactions. Furthermore, the performance indicators for the whole system are not necessarily the same for every entity (individual firm). This highlights the need for both ‘global’ and ‘local’ output analysis in inter-organisational business simulation models. Summarising, we can conclude that a holistic modelling approach is necessary to identify implicit interdependencies among processes, even when they are performed by different organisations. On one hand, different parties should be able to use suitable submodels to assess their own performance. On the other hand, this should be done while at the same time keeping an eye on the influence of ‘local’ changes on ‘global’ performance, as these may have serious implications for the future behaviour of other parties. There is clearly a need for modelling conventions that will allow for modular model implementation and for experimentation with selected sub-models. Finally, implementation of modular models should be achievable even if this is not the initial target of the modelling exercise. For example, two firms might develop models independently of each other and at a later stage wish to link these models into a concrete inter-organisational model. To enhance model reusability, individual models should be easy to link, without extensive modifications. Perhaps the only way to achieve problem-free model decomposition and integration, is by defining standard interfaces between models. At the current status of nonexistence of industrial standards to define the interconnectivity issues between simulation model components, this requirement cannot be easily satisfied. AVENUES FOR FURTHER RESEARCH

A way to overcome the modelling problems would be to use a general purpose programming language to implement a modular model (the approach we had to follow in the case study presented above). An approach is to implement each player (company) as an independent sub-model. For example, in the model we developed for the case study, each sub-model communicates with others when necessary (for example when an order is placed) via a message passing mechanism which initiates some action in the TO model, while it places the message sending entity in an idle state in the FROM model. For example, when retailers place orders to manufacturers, orders remain idle in the retailer sub-model, while a ‘new’ order is created in the manufacturer sub-model. When the order is fulfilled, it is ‘sunk’ in the manufacturer sub-model and the respective entity in the retailer sub-model becomes busy again (forwarded to be processed). Although this message passing mechanism facilitates ‘independent’ modelling of the various levels of decision making of the system and allows for output analysis of sub-models, the whole process is clearly not user-friendly and cannot provide the necessary degree of adaptability and reusability required for business models. A far better solution might be to use a userfriendly simulation package for model development, provided that a package to satisfy the aforementioned requirements actually exists on the market. Based on the above analysis, we can identify the following research directions regarding interorganisational BPM: • A thorough analysis of the distinct characteristics of inter-organisational business process modelling is needed, in order to fully comprehend the differences from ‘traditional’ modelling of internal business systems. • The above analysis should also help towards identifying requirements for inter-organisational BPM and desirable characteristics for simulation software packages to be used for that purpose. • The use of object oriented techniques to facilitate modular model development might be a method worthwhile investigating to address some of the modelling requirements identified. • Finally, the issue of developing standard simulation model interfaces which could be incorporated in simulation packages in order to assist model integration and/or decomposition, might be another possible direction for further research on the issue of business process modelling through simulation. REFERENCES Adams, J.D. (1984) Transforming Work. Miles River Press, Alexandria, VA, USA. Bhaskar, R., Lee, H.S., Levas, A., Petrakian, R., Tsai, F. and

Tulskie, B. (1994) Analysing and Re-engineering Business Processes Using Simulation. In the Proceedings of the 1994 Winter Simulation Conference, Lake Buena Vista, FL, USA, December 1994, pp. 1206-1213. Davenport, T.H. and Short, J.E. (1990) The new industrial engineering: information technology and business process redesign, Sloan Management Review, Summer, pp.11-27. Emmelhainz, M. (1993) EDI: a total management guide. Van Nostrand Reinhold, USA Galliers, R.D. (1994) Information Systems, Operational Research and Business Reengineering, International Transactions of Operational Research, vol.1, no.2, pp.19. Giaglis G.M., Paul R.J., ‘It’s Time to Engineer Reengineering: Investigating the Potential of Simulation Modelling in Business Process Redesign’. In the Proceedings of the International Symposium on Business Process Modelling, Cottbus, Germany, October 1996. Giaglis G.M., Paul R.J., Doukidis G.I., ‘Simulation for intraand inter-organisational business process modelling’. In the Proceedings of the 1996 Winter Simulation Conference, San Diego, California, December 1996 (forthcoming). Hammer, M. (1990) Re-engineering work: don't automate obliterate, Harvard Business Review, July/August, pp.104-112. Hansen, G.A. (1994) Automating Business Process Reengineering: Breaking the TQM Barrier, Prentice-Hall, Englewood Cliffs, New Jersey. Harrington, H.J. (1991) Business Process Improvement: The Breakthrough Strategy for Total Quality, Productivity and Effectiveness, McGraw-Hill, New York. Mylonopoulos N.A., Doukidis G.I., Giaglis G.M., ‘Assessing the expected benefits of EDI through simulation modelling techniques’. In the proceedings of the 3rd European Conference on Information Systems, Athens, Greece, June 1995, pp.931-943. Nygaard-Andersen S., Bjorn-Andersen N.: To join or not to join: a framework for evaluating electronic data interchange systems. Journal of Strategic Information Systems 3 3 (1994) 191-210. Oakland, J.S. (1993) Total Quality Management: The route to improving performance, 2nd ed., Nichols Publishing, New Jersey. Premkumar G., Ramamurthy and Nilakanta S.: Implementation of electronic data interchange: an innovation diffusion perspective. Journal of Management Information Systems 11 2 Fall (1994)157-186. Riggins, F.J. and Mukhopadhyay, T. (1994) Interdependent Benefits from Interorganisational Systems: Opportunities for Business Partner Reengineering, Journal of Management Information Systems, vol.11, no.2, pp.3757. Swami, A. (1995) Building the Business Using Process Simulation. In the Proceedings of the 1995 Winter Simulation Conference, Arlington, VA, USA, December 1995, pp. 1081-1086. Tumay, K. (1995) Business Process Simulation. In the Proceedings of the 1995 Winter Simulation Conference, Arlington, VA, USA, December 1995, pp. 55-60.

BIOGRAPHY

GEORGE M. GIAGLIS is a doctoral candidate and teaching fellow in the Department of Computer Science and Information Systems, at Brunel University, UK. He received his B.Sc. in Informatics from the Athens University of Economics and Business. He has published in the areas of simulation modelling for business systems, Business Process Re-engineering, Electronic Data Interchange, and Information Systems Investment Evaluation. His current research interests include Dynamic Modelling of Business Systems, InterOrganisational Information Systems (IOS), BPR, and EDI. He is the acting manager of the Electronic Commerce/EDI research group in the Athens University of Economics and Business, he has participated in various EC funded research projects, and has also acted as a consultant for a number of companies in Greece.

Suggest Documents