| This dissertation is a collection of three essays in Operations Management and Management Science.;Chapter 2 studies the problem of constructing balanced work groups (i.e., containing approximately equal proportion of members with different gender, age, cultural backgrounds, and other relevant attributes), based on a practical problem of assigning MBA students to study groups. We view balancing requirements as constraints, develop efficient user-friendly software, discuss its implementation and report major improvements in all aspects of students' group work. We also discuss a problem of creating multiple lists of non-overlapping groups, which is unique to our work.;Chapter 3 explains the empirical phenomenon that the constraint programs resulting from the real-world problems in Chapter 2 always had solutions. From the worst-case perspective this need not be the case for in most cases there exist instances with few (e.g., three) attributes where balanced groups cannot be constructed. However, via a variety of techniques (dynamic programming combined with simulation, analytical upper bound and empirical lower bound) we find that the probability that a random instance similar to those observed in practice can be partitioned into balanced groups is effectively 100 percent.;Chapter 1 considers a problem where consumers learn about the possibility of "last-minute" discounts (typical of travel industries) and strategically wait for them. We present a stylized model of aggregate consumer behavior where the firm puts a number of units on sale to maximize its current and future revenue, given that the fraction of customers waiting, and, hence, the revenue in the future, changes depending on the firm's decisions. We formulate the problem as a dynamic program and develop a novel solution approach. We consider several model variations and show that the firm's optimal policy depends on the learning behavior: it is either "passive", where the firm puts some units on sale and allows consumers to "self-regulate" future waiting, or it is an "active" "bang-bang" policy, where the firm intermixes the periods with many units on sale with those with none and thus manages consumer waiting. We discuss managerial insights and show that the firm can strategically allow overbooking to increase its revenue. |
