Breakthroughs in Decision Science and Risk Analysis

Chapter 4Optimal Learning in Business Decisions

Ilya O. Ryzhov

Robert H. Smith School of Business, University of Maryland, College Park, MD, USA

Introduction

We know that business decisions are made under uncertainty. The demand for a product at a retail store varies from one week to the next, and the exact weekly sales cannot be known in advance. Customer response to a new product or service, offered, for example, through a website, is similarly uncertain. Even if a reliable forecast of sales is available, such a forecast will typically model some sort of average or aggregate behavior across a population of customers. Any individual customer is unlikely to behave exactly according to the forecast. Even the supply of the product may be uncertain. For example, a small electricity operator may generate electricity from a wind farm and sell it back to the grid; the firm’s revenue thus depends on volatile wind speeds. A manufacturer may contract with suppliers that experience occasional shortages.

Management science and business analytics offer many ways to deal with uncertainty. Chapter 3 showed how simulation–optimization can be deployed to optimize decisions when the causal relation between controllable inputs and probabilities of outcomes can be quantified and simulated. Sashihara (2011) describes many case applications in such business problems as logistics, pricing, marketing, and new product design, combining (i) statistical forecasts of the future, based on historical data collected ...