We study the multi-period container leasing and load acceptance problem in intermodal freight transportation, where the container capacity could be replenished before allocated to load orders with different profit levels. The container leasing decision is exogenously determined by the initial container inventory left unused from last period and the returned containers in the current period. The load acceptance decision involves trade-offs between accepting the order now or reserve the container for potential high profit level order in the future. Our objective is to characterize the optimal policies in order to maximize the total profit. The two-stage sequential decision problem is formulated as a stochastic dynamic programming problem. We show the optimal leasing quantity follows a base-stock policy and the optimal allocation for different class of demand follows a rationing policy. We further study the impact of demand variability on the optimal policy. Given stochastically higher demand in the future, both total profit and marginal value increase. Such change in demand also results in a higher leasing threshold and a higher rationing level for each demand class.

This dissertation consists of three essays that explore dynamic information and incentives management. Chapter 1 serves as an introduction. In Chapter 2, I study how a manufacturer should sell products through an online retail channel. It is a new class of channel contracting problem, where the retailer can privately observe and control the evolving market conditions. I characterize the optimal contract which unifies the classic first- and second-best policies: it resembles the classic second-best in the short run, but converges to the dynamic first-best in the long run. The result highlights the dual role of network effects: although network effects can improve channel surplus by expanding market size, they can also exacerbate information friction by enhancing the retailer's ability to manipulate the market. Furthermore, I provide new practical guidance: the private information per se need not hurt channel efficiency.

Chapter 3 focuses on a new class of product line design problems, inspired by the growing popularity of personalized subscriptions. Here, consumers' future preferences are determined endogenously by past purchases, current valuation, and random shocks. The optimal design resolves a dynamic tradeoff between preventing cannibalization, extracting surplus, and exploiting consumer habituation. The results shed new lights on product line design: the classic downward distortion principle may no longer work, firms can practice first-degree price discrimination after initial sales. This chapter sheds new light on the increasing demand for personalized subscriptions.

In Chapter 4, I investigate the joint design of compensation and self-directed training schedules, motivated by the trend of self-directed learning in training practice. In the model, the salesperson privately observes his skills, exert effort in selling season, and he can self-invest hiddenly to enhance skills when training; the firm can learn from the salesperson's choice, update the training schedules and revise the sales targets over time. I present a simple implementation of this complex design, and emphasize how training exacerbates the agency problem by providing extra opportunities for the salesperson to manipulate their skills hiddenly. Therefore, I recommend that the firm downgrade training, particularly in the early stages, and implement a penetration-skimming training schedule, where the optimal training level initially increases and then declines towards the end.