Essays on Designing Two-Sided Decisions in Operations: Online Fulfillment with Replenishment and Pricing with Information

This dissertation studies how firms coordinate two-sided decisions in operations under uncertainty. It consists of three essays examining how multiple decision levers, such as replenishment, fulfillment, pricing, and information policy, should be jointly designed rather than in isolation. 
In e-commerce operations, firms must coordinate inventory replenishment and real-time fulfillment. While each component has been studied separately, their interaction remains underexplored. The first essay investigates which lever plays a more decisive role: replenishment or fulfillment. We model a one-location online fulfillment problem with lost sales and heterogeneous stochastic arrivals. Replenishment follows either a base-stock or constant-order policy, while fulfillment decisions are made online. The core metric is expected average profit per replenishment cycle, evaluated across policy-algorithm combinations. The main theoretical result shows when the replenishment cycle is long, the cumulative regret of online fulfillment remains of the same order as in a corresponding single-cycle problem, even under repeated replenishment, revealing regret stability. This phenomenon extends to multi-location settings. We further develop a regret-based framework to compare the value of improving replenishment versus fulfillment and characterize regimes in which optimizing one lever yields a larger revenue impact than refining the other. We also introduce a look-ahead online algorithm that anticipates future replenishment and demand. Numerical experiments verify it outperforms myopic baselines.

Beyond this interaction, omnichannel fulfillment creates another two-sided tension: the same store inventory must support both online and offline customers. The second essay studies how a retailer should balance online access to store inventory against protection for future offline demand. Online orders may use store inventory, while offline customers can only be served by local stores usually with higher value. The retailer observes arrivals sequentially and decides immediately how much inventory to fulfill each online order. We use competitive analysis to evaluate policies. In the continuous setting, we show that static booking limits attain the tight competitive ratio when inventories are identical, but generally cannot attain the tight bound when inventories are non-identical, even with only two stores. To address this, we develop an adaptive booking-limit algorithm that updates cumulative limits as online demand is realized and achieves the tight competitive ratio. We also extend the analysis to the discrete setting through randomized rounding algorithms whose expected profit matches the corresponding deterministic fractional algorithms.

In markets where customers learn socially by observing others' purchases, sellers must manage customer beliefs. The third essay investigates how a monopoly seller can use dynamic pricing and strategic information policy to influence purchases and enhance revenue. We present a model where customers update beliefs after an initial public signal from the seller and observed purchases over time. Comparing dynamic and static pricing, we find that under dynamic pricing, the seller can maximize revenue by providing the most informative signals. In contrast, under static pricing, additional signals offer mixed benefits, as purchasing responses are more complex. This highlights how dynamic pricing can facilitate information policies in managing customer beliefs and optimizing revenue. Overall, this dissertation provides theoretical and managerial insights into designing operational market decisions under uncertainty.