「長鞭效應」會降低供應鏈的效率,因此有必要探討長鞭效應的因應對策。本研究基於「需求訊號處理」的觀點,試圖從末端需求過程和廠商訂單過程來探討長鞭效應的影響因子。在研究中,考慮一個末端需求符合ARIMA(p,0,q)模式的三階段供應鏈系統,並利用Microsoft Excel模擬之。ARIMA(p,0,q)模式中的自迴歸係數、移動平均係數、常數值和誤差項的標準差等四個參數便是末端需求過程參數;廠商訂單過程的參數則包括配銷商的前置時間、預測方法、服務水準、以及需求資訊等。本研究根據這些參數設計一個因子實驗,然後以完全實驗的方式來收集各種因子組合情境下的長鞭效應數據,並透過統計軟體STATISTICA對這些數據作GLM分析,進而可得知各實驗因子對長鞭效應的影響效果。 本研究考慮三種低階ARIMA(p,0,q)模式和三種預測參數組合,各ARIMA(p,0,q)模式皆分別搭配各種預測參數組合進行完全實驗,共進行九次完全實驗,並比較各實驗結果之異同。研究結果簡述如下: (1) 自迴歸係數對長鞭效應有負向影響; (2) 移動平均係數對長鞭效應有正向影響; (3) 當自迴歸係數比移動平均係數小時,並未出現「反長鞭效應」; (4) 以指數平滑法來預測需求幾乎都會引起較劇烈的長鞭效應; (5) 使用較多的歷史需求資訊來預測需求可使長鞭效應降低; (6) 長鞭效應會隨著廠商的前置時間增加而顯著加劇; (7) 只要供應鏈成員必須作需求預測,就無法避免長鞭效應發生。
It’s necessary to pay attention to countermeasures against the bullwhip effect because the efficiency of supply chain can be reduced by the phenomenon. According to the viewpoint of “demand signal processing”, this study attempted to investigate factors affecting the bullwhip effect from the demand process and ordering process. A three-echelon supply chain system was considered and it was simulated on the worksheets of Microsoft Excel. Assuming that customer demands were described by a ARIMA(p,0,q) process, and the demand process consists of four parameters including the autoregressive coefficient, the moving-average coefficient, the constant, and the standard deviation of the error term. The ordering process also consists of four parameters, and they are the lead times of the wholesaler, the forecasting methods, the service levels, and the demand information. Based on these parameters, this study designed a full-factor experiment for the collection of the bullwhip data and then employed a statistical software named STATISTICA to analyze these data. We can then understand how the factors affect the bullwhip. Three models of low-level ARIMA(p,0,q) and three combinations of forecasting parameters were considered. In every full experiment, one ARIMA(p,0,q) model was arranged in pairs with one combination of forecasting parameters, hence the operations of full experiment totaled to 9 times. The analytic results of these experiments were compared with each other. The conclusions could be stated as follows briefly: (1) The autoregressive coefficient affects the bullwhip negatively. (2) the moving-average coefficient affects the bullwhip positively. (3) The “anti-bullwhip effect” doesn’t appear when the autoregressive coefficient is smaller than the moving-average coefficient. (4) The bullwhip effect will be amplified more when we forecast future demands with exponential smoothing rather than moving average. (5) The more demand information used to construct the forecast, the smaller the bullwhip effect. (6) The increase in variability will be greater for longer lead time. (7) The bullwhip effect always happens if the supply chain members had to forecast future demand.
為了持續優化網站功能與使用者體驗,本網站將Cookies分析技術用於網站營運、分析和個人化服務之目的。
若您繼續瀏覽本網站,即表示您同意本網站使用Cookies。