Date of Award

8-2008

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Legacy Department

Industrial Engineering

Advisor

Taaffe, Kevin M.

Committee Member

Kurz , Mary E.

Committee Member

Mayorga , Maria E.

Committee Member

Sridharan , V. `Sri'

Abstract

We consider a firm that delivers its products across several customers or markets, each with unique revenue and uncertain demand size for a single selling season. Given that the firm experiences a long procurement lead time, the firm must decide, far in advance of the selling season not only the markets to be pursued but also the procurement quantity. In this dissertation, we present several operational scenarios in which the firm must decide which customer demands to satisfy, at what level to satisfy each customer demand, and how much to produce (or order) in total.
Traditionally, a newsvendor approach to the single period problem assumes the use of an expected profit objective. However, maximizing expected profit would not be appropriate for firms that cannot afford successive losses or negligible profits over several consecutive selling seasons. Such a setting would most likely require minimizing the downside risk of accepting uncertain demands into the production plan. We consider the implications of such competing objectives.
We also investigate the impact that various forms of demand can have on the flexibility of a firm in their customer/market selection process. a firm may face a small set of unconfirmed orders, and each order will often either come in at a predefined level, or it will not come in at all. We explore optimization solution methods for this all-or-nothing demand case with risk-averse objective utilizing conditional value at risk (CVaR) concept from portfolio management.
Finally, in this research, we explore extensions of the market selection problem. First, we consider the impact of incorporating market-specific expediting costs into the demand selection and procurement decisions. Using a lost sales assumption instead of an expediting assumption, we perform a similar analysis using market-specific lost sales costs. For each extension we investigate two different approaches: i) Greedy approach: here we allocate order quantity to market with lowest expediting cost (lowest expected revenue) first. ii) Rationing approach: here we find the shortage (lost sale) then ration it across all the markets. We present ideas and approaches for each of these extensions to the selective newsvendor problem.

Share

COinS