2. 1 Simulation in the Supply Chain Domain : Evaluating modelling approaches Simulation in the Supply Chain Domain : Evaluating modelling approaches Mr. Chris Owen Dr. Pavel Albores Dr. Doug Love Operations and Information Management Group Aston Business School Aston University, Birmingham

3. Supply chains are dynamic and complex, they involve the integration, coordination and synchronisation of activities between different business entities. Supply chains require the transmission of both material and information between those business entities. The characteristics of supply chains mean that simulation is a useful technique. their complexity obstructs analytic evaluation (Van der Zee and Van der Vorst, 2005) There are many simulation approaches used in the supply chain, however … There is little advice for practitioners on when to use each method. Much of the advice that does exist relies on custom and practice. There is little back to back modelling of the same problem using more than one technique. Background and Context VAN DER ZEE, D. J. & VAN DER VORST, J. G. A. J. 2005. A Modeling Framework for Supply Chain Simulation: Opportunities for Improved Decision Making*. Decision Sciences, 36, 65-95.

4. Research Questions 1.What are the main methods of simulation used to improve supply chain performance? 2.What are the theoretical building blocks and assumptions that lie behind these techniques? 3.How does this illuminate the supply chain problem types for which certain techniques might be better suited than others? 4.What are the relative strengths and weaknesses of different techniques in simulating certain supply chain problem types? 5.What experiments can be done to test and compare alternative approaches? 6.How can these conclusions be used to generate recommendations for practitioners on how they should deploy these tools in achieving their supply chain objectives?

5. Research Questions 1.What are the main methods of simulation used to improve supply chain performance? 2.What are the theoretical building blocks and assumptions that lie behind these techniques? 3.How does this illuminate the supply chain problem types for which certain techniques might be better suited than others? 4.What are the relative strengths and weaknesses of different techniques in simulating certain supply chain problem types? 5.What experiments can be done to test and compare alternative approaches? 6.How can these conclusions be used to generate recommendations for practitioners on how they should deploy these tools in achieving their supply chain objectives?

6. There are three main methods of simulation in the supply chain context Discrete Event Simulation System Dynamics Agent Based Modelling Classification of 100 papers from literature search on simulation and the supply chain

7. Proposition 1 : Discrete methods of simulation can be useful in investigating strategic problem types as well as System Dynamics; Proposition 2 : Discrete methods of simulation can represent feedback effects in models; Proposition 3 : System Dynamics can model supply chain problem types at the operational end of the spectrum as well as the strategic; Proposition 4 : The nature and role of decision makers in the problem may influence the selection of simulation technique; Proposition 5 : The purpose of the modelling (exploratory, problem solving or explanatory) may influence the selection of simulation technique. Propositions were formulated from the literature review and theoretical analysis

8. The case study method was selected to test the propositions The key areas for investigation were : the level of the supply chain problem on a scale from strategic, tactical and operational; the importance of feedback as a feature of the problem; the role of decision makers in the problem and how they can be represented; the purpose of the modelling itself, whether exploratory, descriptive or explanatory. Case Study Approach However, there are times when little is known about a phenomenon, current perspectives seem inadequate because they have little empirical substantiation, or they conflict with each other or common sense. (Eisenhardt, 1989). EISENHARDT, K. M. 1989. Building Theories from Case Study Research. The Academy of Management Review, 14, 532-550.

9. A multiple case approach was used Develop theory Select cases Conduct 2 nd case study Write individual case report Draw cross-case conclusions Design data collection protocol Conduct 1 st case study Conduct remaining case studies Write individual case report Modify theory Develop policy implications Write cross-case report YIN, R. K. 2003. Case Study Research - Design and Methods, London, Sage.

10. Case Studies Case StudyCase characteristicsSimulation Approaches Used Strategic Purchasing CaseModelling the centralisation of procurement in a large construction company System Dynamics Agent Based Modelling Bullwhip CaseDemand amplificationSystem Dynamics Agent Based Modelling Coffee pot ProblemGlobal logistics problem balancing inventory with customer service Discrete Event Simulation System Dynamics Operational Scheduling ProblemDetailed scheduling of job shopDiscrete Event Simulation System Dynamics

11. Case Studies Case StudyCase characteristicsSimulation Approaches Used Strategic Purchasing CaseModelling the centralisation of procurement in a large construction company System Dynamics Agent Based Modelling Bullwhip CaseDemand amplificationSystem Dynamics Agent Based Modelling Coffee pot ProblemGlobal logistics problem balancing inventory with customer service Discrete Event Simulation System Dynamics Operational Scheduling ProblemDetailed scheduling of job shopDiscrete Event Simulation System Dynamics Strategic Purchasing SD Model Strategic Purchasing Agent Model

12. Case Studies Case StudyCase characteristicsSimulation Approaches Used Strategic Purchasing CaseModelling the centralisation of procurement in a large construction company System Dynamics Agent Based Modelling Bullwhip CaseDemand amplificationSystem Dynamics Agent Based Modelling Coffee pot ProblemGlobal logistics problem balancing inventory with customer service Discrete Event Simulation System Dynamics Operational Scheduling ProblemDetailed scheduling of job shopDiscrete Event Simulation System Dynamics Bullwhip SD Model Bullwhip Agent Based Model

13. Case Studies Case StudyCase characteristicsSimulation Approaches Used Strategic Purchasing CaseModelling the centralisation of procurement in a large construction company System Dynamics Agent Based Modelling Bullwhip CaseDemand amplificationSystem Dynamics Agent Based Modelling Coffee pot ProblemGlobal logistics problem balancing inventory with customer service Discrete Event Simulation System Dynamics Operational Scheduling ProblemDetailed scheduling of job shopDiscrete Event Simulation System Dynamics Coffee Pot DES Model Coffee Pot SD Model

14. Case Studies Case StudyCase characteristicsSimulation Approaches Used Strategic Purchasing CaseModelling the centralisation of procurement in a large construction company System Dynamics Agent Based Modelling Bullwhip CaseDemand amplificationSystem Dynamics Agent Based Modelling Coffee pot ProblemGlobal logistics problem balancing inventory with customer service Discrete Event Simulation System Dynamics Operational Scheduling Problem Detailed scheduling of job shopDiscrete Event Simulation System Dynamics Operational SD Model Operational DES Model

15. Findings Case Study 1 – Strategic Purchasing Case Study 2 – Bullwhip and SD Archetypes Case Study 3 – Coffee Pot Problem Case Study 4 – Operational Scheduling Proposition 1 - Discrete methods of simulation can be useful in investigating strategic problem types as well as System Dynamics. Supported The purchasing case clearly demonstrates that ABM can be used very effectively to investigate a strategic problem type. In fact, it could be argued that it identified key decision making in the system which was not identified by the SD approach. Supported The Bullwhip case clearly demonstrates that discrete methods, and in this case Agent Based Modelling (ABM) can be used effectively to investigate a strategic problem type. Supported Discrete Event Simulation (DES) is found to be very effective in investigating this supply chain problem which spans strategic to operational issues. Not applicable Proposition 2 - Discrete methods of simulation can represent feedback effects in models. Supported The agent based model does incorporate feedback effects. Supported Both the Bullwhip case and the SD archetypes demonstrated that discrete methods can represent feedback effects in models. Supported This case demonstrates that both feedback and feed forward of information can be effectively modelled by DES. Not applicable Proposition 3 - System Dynamics can model supply chain problem types at the operational end of the spectrum as well as the strategic. Not applicable Not applicable Contradicted This case demonstrates that SD cannot model classes of supply chain problem where discrete behaviour or measures are important. Contradicted This case demonstrates that there are practical limits to the applicability of SD to discrete problem types. Proposition 4 - The nature and role of decision makers in the problem may influence the selection of simulation technique. Supported This case demonstrates that ABM can be very effective in identifying the key decision makers in systems. It also provides some evidence that the SD approach may contain risks that it may ignore key decision makers embedded in the system. Supported ABM is found to be very effective at locating the decision making process in a given agent. This means that the representation of decision making by individuals is perhaps more accurate and easier to model than in SD, where the decision making process is detached from the individuals in the system. Supported DES seems to be suited to problems where decision making is quite mechanistic. If more complex decision making processes are required, ABM may be needed. Supported DES appears to be suited to modelling fairly simple, mechanistic decisions. Proposition 5 - The purpose of the modelling (exploratory, problem solving or explanatory ) may influence the selection of simulation technique. Supported Both SD and ABM appear to be well suited to exploratory or explanatory modelling. Supported Both SD and ABM are found to be useful in exploratory and explanatory modes. The transparency of SD models may lend the approach to a more interactive process than ABM. Supported DES seems well suited to a problem solving approach. Supported DES appears to be suited to problem solving.

16. An initial guidance framework for practitioners Approach Aspect System DynamicsDiscrete Event SimulationAgent Based Modelling Purpose of the enquiry Characteristics of the problem Problem Level – Strategic Problem Level – Operational Discreteness of measures, entities, resources, process is important Decision making process Physical Space Characteristics of the approach Perspective Feedback Extending ideas from LORENZ, T. A. J., A. 2006. Towards an orientation framework in multi- paradigm modelling. 23rd International Conference of the System Dynamics Society. Nijmegen.

17. An initial guidance framework for practitioners Approach Aspect System DynamicsDiscrete Event SimulationAgent Based Modelling Purpose of the enquiry Exploration and explanation of dynamic relationships Interactive investigation of policies with client Problem solving Optimisation Can be used for exploration and explanation, but transparency and client involvement become key Exploration Understanding of agent behaviours Investigation of unexpected consequences Characteristics of the problem Problem Level – Strategic Policy investigation and evaluations Can be used at the strategic level but transparency of models and involvement of clients becomes key Useful when understanding of individual agent behaviours are the focus of enquiry. Can be used at the strategic level but transparency of models and involvement of clients becomes key Problem Level – Operational May be vulnerable to missing important decision making at this level Cannot be used for certain discrete problems at this level Process level investigation into inventory levels, customer service, physical logistics Investigation into behaviour of individual agent behaviours at the operational level. Discreteness of measures, entities, resources, process is important Aggregation of measures, entities, resources and processes is acceptable Discrete measures or behaviour of discrete entities, resources and processes is an important aspect of the problem Decision making process Not important to link decision process in the model directly to decision makers in the real world Important to link decision process in the model directly to decision makers in the real world Physical Space Not important to the problemMay be important to the problem Characteristics of the approach Perspective Top down, development of dynamic hypothesis Process perspective (material and information flows) Agent perspective (essentially bottom up) Feedback Modelled explicitlyCan be modelled but is hidden Extending ideas from LORENZ, T. A. J., A. 2006. Towards an orientation framework in multi- paradigm modelling. 23rd International Conference of the System Dynamics Society. Nijmegen.

18. Limitations and Further Work This research has used a limited number of cases for the analysis. One practical case has been used and three typifications or secondary cases. More cases would clearly lead to a higher level of confidence in the generalisability of the findings. Another potential limitation has been that the modelling has been carried out by this researcher. This means that there is some risk of bias, since this researcher may have conscious or unconscious preferences or attitudes that are leading to the focusing on certain issues and the biasing of results. More real life cases would lead to more confidence in the findings.

19. Limitations and Further Work This research has used a limited number of cases for the analysis. One practical case has been used and three typifications or secondary cases. More cases would clearly lead to a higher level of confidence in the generalisability of the findings. Another potential limitation has been that the modelling has been carried out by this researcher. This means that there is some risk of bias, since this researcher may have conscious or unconscious preferences or attitudes that are leading to the focusing on certain issues and the biasing of results. More real life cases would lead to more confidence in the findings.

20. Summary / Conclusions The three main methods of simulation (SD, DES and ABM) in the supply chain domain have been compared through back to back modelling of case studies Discrete methods (DES and ABM) have been found able to model all problem types and in particular to be useful in modelling strategic problem types System Dynamics has been found to have hard limits to its application in more operational and discrete problems System Dynamics (as a top down approach) may be vulnerable to missing decision making at the operational level A framework to assist practitioners in the selection of the appropriate method has been presented

21. Summary / Conclusions The three main methods of simulation (SD, DES and ABM) in the supply chain domain have been compared through back to back modelling of case studies Discrete methods (DES and ABM) have been found able to model all problem types and in particular to be useful in modelling strategic problem types System Dynamics has been found to have hard limits to its application in more operational and discrete problems System Dynamics (as a top down approach) may be vulnerable to missing decision making at the operational level A framework to assist practitioners in the selection of the appropriate method has been presented

22. Questions ?

23. Findings PropositionFindings Proposition 1 : Discrete methods of simulation can be useful in investigating strategic problem types as well as System Dynamics; Supported both by case study findings and enfolding literature review. Proposition 2 : Discrete methods of simulation can represent feedback effects in models; Supported both by case study findings and enfolding literature review. Proposition 3 : System Dynamics can model supply chain problem types at the operational end of the spectrum as well as the strategic; Rejection of the original proposition supported by case study findings and refined by enfolding literature review. System Dynamics (SD) cannot model problems where the performance measure or the behaviour of discrete entities, resources or processes is the focus of interest. Proposition 4 : The nature and role of decision makers in the problem may influence the selection of simulation technique; Decision making is modelled differently by all three approaches. Each approach has limitations. Proposition 5 : The purpose of the modelling (exploratory, problem solving or explanatory) may influence the selection of simulation technique. Some insights into the potential uses for all three approaches in all different modes of enquiry depending on the philosophical stance of the modeller and the client. Proposition 6 : System Dynamics (SD) is vulnerable to the risk of overlooking important decision making processes at the operational level and of making a Type III error. The purchasing case has identified this potential risk. The literature appears to support the possibility of this error, although proposes a methodological solution (calibration).