以灰預測來提升零售業銷售預測準確度的可行性研究

Abstract

銷售預測的準確度影響到客戶滿意度及存貨水位。自2009年九月後，ST公司的存貨計畫轉由SSC(Shared service center)運作，原來的預算皆以正成長為目標，在大環境影響下，年營業額連續兩年負成長，但存貨金額卻急遽增加，在在凸顯銷售預測準確性的重要。 因傳統人工預測方法，是以前一年的實際銷售量為基礎，再以未來在不同月份預計增加或減少的行銷活動作加權，因此人工判斷的主觀性及資訊的掌握極為重要。本方法的缺點為非系統性，不易複製與操作，及事後亦較難提出改善準確度的具體方案。 本論文嘗試採用灰預測所推估的銷售預測值，與前述的人工預測，對已知的過去推測已知的未來，做準確度的檢測，以驗證灰預測方法對改善零售業銷售預測準確度的可行性。 實驗結果證明，因促銷量為已知的增加量，會影響波動劇烈的銷售歷史所推算出灰預測值的準確性，因而推論出對不定期強力促銷重點產品(Push)，以人工方法有較佳的準確性;而灰預測方法對主要來自客戶需求(Pull)的產品，反而有較佳的準確性 。

The customer satisfaction and inventory level were affected by the accuracy of sales forecast. Since September 2009, ST company's inventory management planning moved to Shared Service Center in Manila. The growth rate of the budget plan was positive. Under the poor environment, sales growth rate became negative in the past two years but the inventory increased sharply. It could be said the sales forecast accuracy is significant. Traditional manual forecasting method bases on the previous year's actual sales, and adjust the monthly forecast by increasing or decreasing marketing activities vs the same month of the previous year. Therefore, manual judgment and information collection are extremely critical. The disadvantage of this method is non-systematic, difficult to copy and operate, and it’s not easy to take action plan to improve the forecast accuracy . This paper attempts to compare the sales forecast using gray prediction value with the mentioned manual forecasting method by using the known past data to project the known future forecast in the past and to evaluate the feasibility that the gray prediction method can improve retail sales forecast. Experimental results show that the promotional amount will affect the accuracy of the gray predictive value calculated from volatility of sales history data, and thus infer that promoted products (Push) have accurate predict results by using manual forecasting method; but gray prediction method has better accuracy on the natural customer demand’s main product (Pull).