A Multi-Agent Simulation of Retail Management Practices Peer-Olaf Siebers Working for Uwe Aickelin In co-operation with Leeds University Research Seminar 15/03/2007 Content 1. Research Field 2. Project Aim 3. Simulation 4. Focus 5. Doing the Job 6. Conclusions 15/03/2007 A Multi-Agent Simulation of Retail Management Practices http://www.cs.nott.ac.uk/~pos/ 2/12
1. Research Field Operational Research Synonyms: Operations Research; Systems Analysis Definition: The discipline of applying advanced analytical methods to help make better decisions. Analytical methods used (examples): Linear Programming Network Analysis Meta Heuristics Queuing Theory Game Theory Simulation 15/03/2007 A Multi-Agent Simulation of Retail Management Practices http://www.cs.nott.ac.uk/~pos/ 3/12 2. Project Aim Project Aim Research Tool? Understand Predict Retail Planning - supply chain dynamic models - staffing - shelf optimisation - resource management - space allocation Economics - efficiency evaluation - self organising markets - artificial laboratories Understand and Predict the Impact of Different Management Practices on Retail Store Productivity Operational Research and Management Science Modelling Techniques - linear programming - network analysis - meta heuristics - queuing theory - game theory - simulation People Movement - crowd behaviour - egress analysis - artificial intelligence modelling techniques Social Sciences Psychology - organisational psychology - models of consumer behaviour - impact of advertisement Social Science Simulation - system dynamics - micro simulation - multilevel simulation - queuing models - cellular automata - multi-agent systems - artificial neural networks - evolutionary computation 15/03/2007 A Multi-Agent Simulation of Retail Management Practices http://www.cs.nott.ac.uk/~pos/ 4/12
3. Simulation What is Simulation? 15/03/2007 A Multi-Agent Simulation of Retail Management Practices http://www.cs.nott.ac.uk/~pos/ 5/12 4. Focus Our Simulation Study Goal: Develop a simulation model that helps to understand and predict the impact of different management practices on retail store productivity Our Focus: Individual departments within department store Incorporating variables from different levels of analysis Simulation study supported by case studies Using a relatively new technology: Agent-Based Simulation 15/03/2007 A Multi-Agent Simulation of Retail Management Practices http://www.cs.nott.ac.uk/~pos/ 6/12
Modelling Concept Shopping need, attitudes, demographics etc. Visual Dynamic Stochastic Simulation Model Sales Staff Agent Sales Agent Attitudes, length of service, competencies, training etc. Emergent behaviour on macro level Understanding about interactions of entities within the system Manager Agent Leadership quality, length of service, competencies, training etc. Identification of bottlenecks Global Parameters Number of customers, sales staff, managers etc. Interface for User Interaction during Runtime Performance Measures Staff utilisation, average response time, customer satisfaction etc. Data Envelopment Analysis Inputs for DEA Outputs for DEA Relative efficiency of different simulated scenarios 15/03/2007 A Multi-Agent Simulation of Retail Management Practices http://www.cs.nott.ac.uk/~pos/ 7/12 Case Study Research Questions Management Practice: Training Staff at different training levels Product oriented vs. customer oriented Management Practice: Empowerment Refunds Remain when other staff assist (learning) Other Aspects Customer population 15/03/2007 A Multi-Agent Simulation of Retail Management Practices http://www.cs.nott.ac.uk/~pos/ 8/12
State Charts if(uniform(1)>=0.8) Customer Fire: After timeout Timeout: triang(1,5,7) Staff 15/03/2007 A Multi-Agent Simulation of Retail Management Practices http://www.cs.nott.ac.uk/~pos/ 9/12 Simulation Execution 15/03/2007 A Multi-Agent Simulation of Retail Management Practices http://www.cs.nott.ac.uk/~pos/ 10/12
5. Conclusions First Experiments: Test of simulator investigating training and empowerment impact So far - so good Conclusions Stuff on the To Do List: Validation, service level index refinement Adding variable arrival rates Costs (e.g. for training, new tills) Agent memory (to model things like intentions and motivation) More diverse populations, e.g. customer and staff stereotypes 15/03/2007 A Multi-Agent Simulation of Retail Management Practices http://www.cs.nott.ac.uk/~pos/ 11/12 Thanks! Thanks to Will, Phil, Jan and Helen for their support! 15/03/2007 A Multi-Agent Simulation of Retail Management Practices http://www.cs.nott.ac.uk/~pos/ 12/12