未來代工價格具隨機特性之產能預約定價

Abstract

客戶的需求預測通常都會被扭曲，過去許多的研究皆著重在如何鼓勵客戶說實話，本研究認為客戶可能無法完全的掌握需求預測。對於客戶而言，未來需求本質上為一隨機變數，所以客戶預測此一隨機變數的平均值未必能代表據實的告知實際需求。本研究假設產能購買者和產能出售者對於未來市場的總合需求有ㄧ致的共識，並提出ㄧ方法來重新解決客戶需求預測的問題。產能購買者欲預約未來產能，需支付產能預約金。產能預約金的定價，對於顧客透過預約做出合理的需求預測扮演著相當重要的角色。產能出售者可以透過合宜的定價決策，使得本身的利益最大化。本研究以半導體代工廠為例，來說明所提出的產能預約金定價方法。

Demand forecast from customers are usually boosted. Most previous research focused on how to encourage customers to “tell the truth”. This paper argues that customers may not fully “know the truth”. The future demand to customer is generally modeled as a random variable. Telling the mean of the random variable does not imply the truth (the realized demand) has been declared. This paper presents an approach to resolve the “truth-telling” issue of demand forecast. We assume that the capacity-buyers and capacity-sellers have a consensus about aggregate future demand, which is a random variable. This approach suggests that customers pay a booking fee to reserve future capacity. Pricing the booking fee plays a role to motivate customers to make rational demand forecast through booking. Capacity-sellers can thus optimize their profits by make an appropriate pricing decision. This research presents a method for pricing the capacity-booking fee, which is illustrated by a case in semiconductor foundry industry.