Retailer's ordering decision under demand uncertainty and capacity constraint

• Main Author: Ganesh Abhishek
• Other Authors: School of Mechanical and Aerospace Engineering
• Format: Final Year Project
• Language: English
• Published: 2014
• Subjects: DRNTU::Engineering::Industrial engineering::Operations research
• Online Access: http://hdl.handle.net/10356/60433
• Institution: Nanyang Technological University

Obtaining accurate demand information is a great motivator for companies trying to optimize their overall operational and supply chain costs. Several modes of investigation have been conducted to study the impact of demand forecasting on profit maximization. In this study, the author investigated the effect of demand forecast evolution under capacity and cost constraints and its outcome on the overall ordering decision for a retailer. This study was based on a conventional newsvendor model with a retailer who has a single inventory system and single ordering opportunity over multiple periods which are equally spread out over a fixed planning horizon. The author also incorporated a linearly increasing procurement cost function and non-linearly decreasing production quantity limit to study the trade-off a retailer has to make between the evolution of demand forecast updating, procurement cost and available production quantities to maximize profits. The author encompassed a Markovian Model of demand forecasting called the Martingale Model of Forecast Evolution in this study. The normal and log-normal distribution systems of this model were studied as part of this investigatory project. Analysis was conducted by simulating the model on MATLAB and then conducting a sensitivity analysis on the results obtained. This analysis involved understanding the effect of varying variables from the newsvendor equation and the exogenous variable – the manufacturer’s production quantity. Special focus was given to the newsvendor critical ratio and the variability of demand forecast updating. The general results obtained leaned towards the log-normal distribution system giving out a better set of results.