ieepapersoftdrink

download ieepapersoftdrink

of 6

Transcript of ieepapersoftdrink

  • 8/6/2019 ieepapersoftdrink

    1/6

    2004 IEEE InternationalC o n f e r e n c eon S y s t e m s , Man and Cyberne t i cs

    A Decision Support System forDesigning Business Strategies

    *- An Application to Soft Drink Business -

    Itsuo HATONO Kenichi KUROTANI Kenya MURAKAMIInformation Science and Fuji Electric Systems Fuji Electric Advanced

    Technology Center, Co., Ltd., Technology Co., Ltd.Kobe University Tokyo, Japan Tokyo, Japan

    Kobe, 657-8501, [email protected] p

    Nobutada FUJII, Kanji UEDAResearch into Artifacts,

    Center for EngineeringThe University of Tokyo

    Tokyo, Japan

    Abstract - mispaperdea l s wi th a decision su pportsystem 2 Virtual Market Model~ ~~

    of planning business strategies for demand chains. To de -scribe the complex behavior of ea ch element in the demandchain, an agent based modeling approach and methodologyof i'irtual marker are introduced to develop the simulationmode l. The proposed model is applied to so@ drink vendingmachine busiriess as an example. Demand characteristics ofsop drinks through vending machines are investigated basedon the m erhodology by simulation sriidies.

    A virtual market model in this paper con sistsof the agentscorresponding to manufacturers of products, dealers, andcustomers in a virtual market. We assume that each agentdetermines own behavior autonom ously based on the inter-action between the other agents. In the virtual market, cus-tomers' selection of products effect on the future new prod-ucts. T he introduction of new products to the market maychange the behavior of the customers. This means that themodel might be simulate the au tonomous evolution processof the demand chain and there is possiblity that we canfind the effective business strategies in the simulationDro-

    Keywords: Decision support, Demand chain, Simulation,Busin ess Strategy. -

    cess. Figure 1 llustrates the evolution processof the de-man d chain. Furthermore, we canalso analyze the effectsof the business strategies, and can adjust those by tuning theParameters in the simulation models.

    1 IntroductionT~ cope with recent globalizationofeconomy and diversi.

    fied customers' request, enterprises have required to estimatethe change of customers' preference immediately and drawup the effective business strategies inorder to overcome thecompetitors in the market. However, it is not easy to predictthe customer's and competitors' behavior, bothof which are

    2.1 Autonomous behavior of agents

    Each agent has an evaluation function to determine ow nimpomnt th e effectivestrategies. In th e behaviors, such as buying 01 ot, selectionof types, volume,

    and price of the product, and so on , In th e evaluation func.resence, many enterprises analyzethe Of the

    tomers by the conventional statistical methods and reflectsthe resu lts to own strategies.

    simula-and support top1an the business

    strategies based on the research onthe biological manufac-turing systems[*, 21. As the first step Of the research, wedevelop a Prototype system of demand chain simulator forsoft drin k business through vending machines.

    tion, we of the evaluation functionaredecided by the internal information in the agent and theex-tema l one obtained through the interaction between agents.As shown in Figure 2, the behavior of each agent happensthrough the production How from the manufacturers cus-tomers, and the monetary feed back in the direction fromcustomers to Furthem ore, theflow and thefeed back cause the change of the internal informationofeach agent.

    that the

    In Ibis paper3 wesystem that can

    to realize a demand

    *0-7S03-8566-7/04/$20.00 @ 2004 IBEE

    1475

    mailto:[email protected]:[email protected]
  • 8/6/2019 ieepapersoftdrink

    2/6

    Figure 1: Emergence of demand chain.

    Flow of productsAroducer

    Customer

    Monetary flow

    Figure 2: Material and monetary flows between the interact-ing agents.

    2.2 Soft drink business through v ending ma-

    chinesIn this paper, we deal with thesoft drink business throughvending machines as a case study of demand chain simula-tion. The soft drink business has the characteristics as fol-lows:

    Th e variety of the products is relatively sm all,

    the market is relatively simple hut the market size islarge,

    the products are very familiar to all the customers.

    In this paper, we deal with only sof t drinks such coffee, cola,each of which is often sold through vending machines.

    The demand of soft drinks through vending machines havethe characteristics a s follows:

    Character is t ics of products: Soft drinks are consumerproducts, small varieties(10-100 kinds of products)and the volume of each productis large.

    De ma nd fluctuation: The demand is much depend on theseasons, change of temperature, the number of cus-tomers that pass in front of each vending machine.

    Dem and control : Inorder to increase the demand, the op-erators of vending machines (dealers) takes measures,such as campaign, promotion, change the arrangementof products, change of the price, andso on .

    For example, a typical business office of a n operator dealwith about 250 kinds of products by 10 manufacturers, has

    600 vending machines an d several staffs (called root m an) fo rfilling up, chan ge of arrange, and change of price of productsat each vending machine. Each root man handles80 vendingmachines, and visits each vending machinein a week.

    3 Modeling vending machine demandchain

    In this paper, we develop a simulation model of the de-mand chain shown in Fig ure3. The simulation model con-sists of 4 agents; (1 ) soft drink manufacturers,(2) operatorsof vending machines, (3) vending machines, (4) customers.Furthermore, to represent the difference of the behavior ofthe customersat the various places that the vending mac hines

    are located, we introduce an attribute location, each ofwhich corresponds to the places such as schools, buildings,parks, and so on. A location can contains several vendingmachines of different operators. We describe the behavior ofeach agen t as follows.

    1. Soft d r i n k m a n u f a c t u r e rs

    Take measures, such advertisement, campaign f or

    Accept the orders from the operators and sh ip the

    Produce the products to keep the amount of the

    the products based on ow n strategies.

    products from the stock.

    stock.

    2. O p e r a t o r s of vending machines

    Order the product to keep the amount of the own

    Select and arrange products in each vending ma-

    Assign the vending machines to each root man an d

    Change the price of each product based on the

    stock.

    chine based on own strategies.

    the traveling schedule.

    strategies.

    3. Locat ion

    Vending machines:e Show the products to customers.

    Sell the product to customers.

    Buy the product based on the own preference

    In this paper, we use OMN et++[3], which is a kind ofdiscrete event simulator. Figure5 an d 6 show the snapshotsof the demand chain simulation.

    Cus tomer :

    for each product.

    1476

  • 8/6/2019 ieepapersoftdrink

    3/6

    Figure 3: An example of demand chain of soft drink vending machine business.

    Figure 4 Flow of planning p rocess of the strategies.

    1477

  • 8/6/2019 ieepapersoftdrink

    4/6

    Figure 6: A snapshot of location in demand chain simula-tion.

    Figure 5 : A snapshotof product flow in demand chain sim-

    ulation.

    Table 1 : Example of simulation scenarios

    7 1 1ssumed situations Conditions and results to be simulated

    * Their effective timing ?

    * Their influences on sales?

    Their influences on sales?

    1478

  • 8/6/2019 ieepapersoftdrink

    5/6

    3.1 Simulation scenariosIn this paper, we assume that each agent behaves in the

    virtual market accordingto som e scenarios.In the scenarios,we try to draw up the strategies of manufacturer or operatoragents. T he strategies are describedas the parameters in theagents in the simulation model. Th e parameters are evaluated

    and adjusted based on the simulation results. By repeatingthe process, we try to obtain so me good strategies.Table 1shows an example of the simulation scenariosi n this paper.

    Figure 4 shows the flow of planning processof the strat-egy.

    4 Numerical examplesTo evaluate the behavior of the simulator developed in this

    paper, we show the simulation results under the simple sim-ulation conditionsas follows:

    3 soft drink m anufacturersexist in the market.

    3 operators exist in the market, each operator has30vending machines, and3 root men.

    e 5 locations exist. In each location, the average arrivalinterval of customers is 0.004 simulation unit time.

    Under th e above conditions,

    Operator 3 will down the price at300 time unit from120 Japanese yen to 1IO yen.

    The other operators down by 10 Japanese yen if thesales account decreases15% after AT simulation unittime.

    Figure 7,s nd 9 show the simulation results whenAT is m,

    15, and 40, respectively. When AT of an operator is 00, heoperator never change the price. Th e simulation results showthat the strategy considering only the sales account leads theexcessive competition.

    Figure 10 shows the simulation result under the condi-tions as follows:

    The price of eac h product i s 12 0 Japanese yen (fixed).

    The initial, standard, and minimum stock of each softdrink manufacturer are 10,000, 7,000, and 10,000, re-spectively.

    Figure 11 shows the simulation results when only the initialand standard stock of manufacturer 2 are changedto 20,000.In these simulations, each manufacturer produces to increasethe stock until the standard one, when the stockis less thanthe minimum.

    In Fig. 10, the sales account of manufacturer 2 oscillate.We can interpret the simulation result that the dem and fore-casting of manufacturer 2 is not accurate. O n the other hand,Fig. 11, the sales account manufacturer 2 does not. Thi smeans that standard stock 20,000is adequate based on thecorrect demand forecasting.

    I I'0 100 200 300 do0 500 600 7W 800 9W

    Simviafion xime

    Figure 7: Simulation result i n the case thatAT is m.

    Figure 8: Simulation result in the case thatAT is 15.

    Figure 9: Simulation result in the cas e thatAT is 40.

    1479

  • 8/6/2019 ieepapersoftdrink

    6/6

    ? A 3 8 1 1

    Slmvlaln" time

    Figure IO : Simulation result in the c ase thatsi, "', an d ss re100 00,7 000 , and 10000, respectively.

    450WO I

    soood lb 0 200 360 400 GO 600 7W 8bO 810Slm"1aIlOn m e

    Figure 1 1 : Simulation result in the case thats' an d ss are20000, and 20000 only in maker2, respectively.

    5 ConclusionIn this paper, we develop a d ema nd chain simulation sys-

    tem for soft drink business through vending machines inor-de r to analyze and draw up the bus iness strategies. Th e sim-ulation results suggest that the sim ulation system can be ap-plicableto the real market.

    Further research might be focused on the application tomore realistic simulation case studies and the implementa-tion of adaptation m echanism of ag ents to obtain the effec-tive strategies. Therefore, the agents predict the chang eofthe external environment including the behavior of the otheragents, and adapt the own strategies. By o bservingthe adap-tation process, we might be able to obtain the knowledgeabou t the effective strategies in the virtual market.If the vir-tual market has enough similarityto the real market, we canestima te the obtained the strategies are also effective in the

    real market.

    AcknowledgementThis research is supported by IM S/NGM S PhaseII (Intel-

    ligent Manufacturing Systems/ Next Generation Manufac-turing Systems) international research program and Grant-in-Aid for Scientific Research(C) (No. 15560346) by Min-isiry of Education, Culture, Sports, Scienc e and Technology,Japan.

    References[ I ] K. Ueda, H. Vaario, and N. Fujii: Interactive manufac-

    turing: H uman aspects for biological m anufacturingsys-tem, Annals o f f he U R P , Vol. 47, No. 1, pp. 389-392,1998.

    [2] K. Ueda, J. Vaario, T. Takeshita, and I. Hatono: Anemergent synthetic approach to supply network,Annalsofthe U R P, Vol. 48, No . 1, pp. 377-380, 1999.

    [3] A. Varga: OMNeT++ - Discrete event sirnulator -,http://www.hit.bme.hu/phdvargaa/omnetpp. htm,002.

    1480

    http://www.hit.bme.hu/phdvargaa/omnetpp.htmhttp://www.hit.bme.hu/phdvargaa/omnetpp.htm